Transitioning to Physics-of-Failure as a Reliability Driver in Power Electronics

Transcription

1 2014 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. Digital Object Identifier (DOI): /JESTPE IEEE Journal of Emerging and Selected Topics in Power Electronics, accepted on 01 November 2013 Transitioning to Physics-of-Failure as a Reliability Driver in Power Electronics Huai Wang Marco Liserre Frede Blaabjerg Peter de Place Rimmen John B. Jacobsen Thorkild Kvisgaard Jørn Landkildehus Suggested Citation H. Wang, M. Liserre, F. Blaabjerg, P. P. Rimmen, J. B. Jacobsen, T. Kvisgaard and J. Landkildehus, "Transitioning to physics-of-failure as a reliability driver in power electronics, IEEE Journal of Emerging and Selected Topics in Power Electronics, accepted.

2 1 Transitioning to Physics-of-Failure as a Reliability Driver in Power Electronics Huai Wang, Member, IEEE, Marco Liserre, Fellow, IEEE, Frede Blaabjerg, Fellow, IEEE, Peter de Place Rimmen, John B. Jacobsen, Thorkild Kvisgaard, and Jørn Landkildehus Abstract Power electronics has progressively gained important status in power generation, distribution and consumption. With more than 70% of electricity processed through power electronics, recent research endeavors to improve the reliability of power electronic systems to comply with more stringent constraints on cost, safety and availability in various applications. This paper serves to give an overview of the major aspects of reliability in power electronics and to address the future trends in this multidisciplinary research direction. The ongoing paradigm shift in reliability research is presented first. Then the three major aspects of power electronics reliability are discussed, respectively, which cover from physics-of-failure analysis of critical power electronic components, state-of-the-art design for reliability process and robustness validation, and intelligent control and condition monitoring to achieve improved reliability under operation. Finally, the challenges and opportunities for achieving more reliable power electronic systems in the future are discussed. Index Terms Power electronics, design for reliability, physics-of-failure, robustness validation, IGBT modules, capacitors. P I. INTRODUCTION OWER electronics enables efficient conversion and flexible control of electric energy by taking advantage of the innovative solutions in active and passive components, circuit topologies, control strategies, sensors, digital signal processors and system integrations. While targets concerning efficiency of power electronic systems are within reach, the increasing reliability requirements create new challenges due to the following factors: Mission profiles critical applications (e.g. aerospace, military, more electrical aircrafts, railway tractions, automo- Manuscript received April 22, 2013; revised July 25, 2013 and October 9, 2013; accepted November 1, H. Wang and F. Blaabjerg are with the Department of Energy Technology, Aalborg University, DK-9220 Aalborg, Denmark ( hwa@et.aau.dk; fbl@et.aau.dk). M. Liserre is with the Faculty of Engineering, Christian-Albrechts-University of Kiel, Kiel, Germany ( ml@ tf.uni-kiel.de). P. P. Rimmen and J. Landkildehus are with Danfoss Power Electronics A/S, R&D Design Center, DK-6300 Gråsten, Denmark ( pr@danfoss.com; jla@danfoss.com). J. B. Jacobsen and T. Kvisgaard are with GRUNDFOS Holding A/S, DK-8850 Bjerringbro, Denmark ( jbjacobsen@grundfos.com; tkvisgaard@grundfos.com). Color versions of one or more of the figures in this paper are available online at Digital Object Identifier /JESTPE tive, data center, medical electronics). Emerging applications under harsh environment and long operation hours (e.g. onshore and offshore wind turbines, photovoltaic systems, air conditions and pump systems). More stringent cost constraints, reliability requirements and safety compliances, e.g. demand for parts per million (ppm) level failure rates in future products. Continuous need for higher power density in power converters and higher level integration of power electronic systems, which may invoke new failure mechanisms and thermal issues. Uncertainty of reliability performance for new materials and packaging technologies (e.g. SiC and GaN devices). Increasing complexity of electronic systems in terms of functions, number of components and control algorithms. Resource constraints (e.g. time, cost) for reliability testing and robustness validation due to time-to-market pressure and financial pressure. Table I illustrates the industrial challenges in a reliability perspective of yesterday, today and tomorrow. To meet the future application trends and customer expectations for ppm level failure rate per year, it is essential to have better understanding of failure mechanisms of power electronic components and to explore innovative R&D approaches to build reliability in power electronic circuits and systems. From this perspective, opportunities exist for power electronics to expand its role in dealing with efficient and reliable power processing in different kinds of applications. Nearly four decades ago, the scope of power electronics was defined by William E. Newell as three of the major disciplines of electrical engineering shown in Fig. 1(a) [1]. Likewise, the future reliability research in power electronics involves multidisciplinary knowledge as defined here shown in Fig. 1(b). It covers the following three major aspects: analytical analysis to understand the nature of why and how power electronic products fail; design for reliability and robustness validation process to build in reliability and sufficient robustness in power electronic products during each development process; intelligent control and condition monitoring to ensure reliable field operation under specific mission profiles. Robustness validation is a process that is widely accepted and implemented in the automotive sector, which is to demonstrate that a product performs its intended functions with sufficient margin under a defined mission profile within its specified lifetime [2]. Mission profile

3 2 TABLE I THE RELIABILITY CHALLENGE IN INDUSTRY SEEN BEFORE, TODAY AND IN THE FUTURE Yesterday Today Tomorrow Customer expectations Replacement if failure Years of warranty Low risk of failure Request for maintenance Peace of mind Predictive maintenance Reliability target Affordable market returns (%) Low market return rates ppm market return rates R&D approach Reliability test Robustness tests Design for reliability Avoid catastrophes Improve weakest components Balance with field load / mission profile Product operating and Testing at the limits Understanding failure mechanisms, field load, root cause R&D key tools function tests Multi-domain simulation TABLE II TYPICAL LIFETIME TARGET IN DIFFERENT POWER ELECTRONIC APPLICATIONS (a) Fig. 1. Defined scope in (a) power electronics by William E. Newell in 1970 s [1] and (b) power electronic reliability research needs seen from today. is a representation of all of the relevant operation and environmental conditions throughout the full life cycle [2] through production process, test, shipping, service to end of life. The robustness validation process involves the activities of verification, legal validation, and producer risk margin validation. This paper gives an overview of power electronics reliability from the three respective aspects defined in Fig. 1(b) and addresses the future trends in this multidisciplinary research direction. II illustrates the ongoing paradigm shift in reliability research in power electronics. III presents the Physics-of -Failure (PoF) analysis of reliability-critical components to provide a basis for system level design. IV discusses the state-of-the-art Design for Reliability (DFR) and robustness validation process to build-in reliability through design. V presents the control and monitoring (including fault-tolerant strategies) methods to improve reliability of power electronic systems under field operation. Finally, the future challenges and opportunities in reliability of power electronics are discussed. II. ONGOING PARADIGM SHIFT IN RELIABILITY RESEARCH IN POWER ELECTRONICS Reliability is defined as the ability of an item to perform the required function under stated conditions for a certain period of time [3], which is often measured by probability of survival and failure rate. It is relevant to the durability (i.e. lifetime) and availability of the item. The essence of reliability engineering is to prevent the creation of failures. The deficiencies in the design phase have effect on all produced items and the cost to correct them is progressively increased as the development proceeds. (b) Applications Aircraft Automotive Industry motor drives Railway Wind turbines Photovoltaic plants Typical design target of Lifetime 24 years (100,000 hours flight operation) 15 years (10,000 operating hours, 300,000 km) 5-20 years (60,000 hours in at full load) years (10 hours operation per day) 20 years (24 hours operation per day) 5-30 years (12 hours per day) A. Reliability in Typical Power Electronic Applications The performance requirements of power electronic products are increasingly demanding in terms of cost, efficiency, reliability, environmental sustainable materials, size, and power density. Of which, the reliability performance has influences on the safety, service quality, lifetime, availability and life cycle cost of the specific applications. Table II summarizes the typical design target of lifetime in different applications. To meet those requirements, paradigm shift is going on in the area of automotive electronics, more electrical aircrafts, and railway tractions by introducing new reliability design tools and robustness validation methods [2], [4]-[5]. With the increasing penetration of renewable energy sources and the increasing adoption of more efficient variable-speed motor drives [6]-[8], the failure of power electronic converters in Wind Turbines (WTs), Photovoltaic (PV) systems and motor drives are becoming an issue. Field experiences in renewables reveal that power electronic converters are usually one of the most critical assemblies in terms of failure level, lifetime and maintenance cost [9]. For example, it shows that frequency converters cause 13% of the failure level and 18.4% of the downtime of 350 onshore wind turbines in a recent study associated with 35,000 downtime events [10]. Another representative survey in [11] concludes that PV inverters are responsible for 37% of the unscheduled maintenance and 59% of the associated cost during five years of operation of a 3.5 MW PV plant. It should be noted that such statistics always look backwards as those designs are more than 10 years old. The present technology could have different figures. B. Ongoing Paradigm Shift in Reliability Research The reliability engineering has emerged as an identified discipline since 1950s with the demands to address the reliability

4 3 issues in electronic products for military applications [12]. Since then, much pioneer work has been devoted to various reliability topics. One of the main streams is the quantitative reliability prediction based on empirical data and various handbooks released by military and industry [13]. Another stream of the discipline focuses on identifying and modeling of the physical causes of component failures, which was the initial concept of PoF presented in 1962 [14]. However, until the 1980s, the handbook based constant failure rate models (e.g. Military-Handbook-217 series [15]) have been dominantly applied for describing the useful life of electronic components. Since 1990s, with the increased complexity of electronic systems and especially the application of integrated circuits (ICs), more and more evidences were suggesting that constant failure rate models are inadequate [16]. The Military-Handbook-217F is therefore officially cancelled in PoF approach has started to gain its more and more important role in reliability engineering. In recent years, the initiatives to update the Military-Handbook-217F have turned to a hybrid approach, which is proposed for the planned version of Military-Handbook-217H [17]. During the stage of program s acquisition-supplier selection activities, updated empirical models will be used for comparing different solutions. During the actual system design and development stage, scientific based reliability modeling together with probabilistic methods will be applied. Intensive PoF research has been continuously conducted since 1990s in microelectronics and the state-of-the-art results are presented in [16] and [18]. With the transition from pure empirical based methods to more scientific based approaches, the paradigm shift in reliability research is going on from the following aspects: 1) From Components to Failure Mechanisms PoF approach is a methodology based on root-cause failure mechanism analysis and the impact of materials, defects and stresses on product reliability [19]. It changes the analysis of system from a box of components to a box of failure mechanisms. The traditional handbook based reliability prediction provides failure rate models for various components. PoF approach analyzes and models each failure mechanism induced by environmental and usage stresses. For a given component, there could be multiple failure mechanisms which should be identified individually. Moreover, failure mechanisms are not limited to the component level. As discussed in the standard ANSI/- VITA 51.2 [18], there are various failure mechanisms in component level (i.e. single transistor level), package level, and Printed Circuit Board (PCB) level. From this prospective, it is challengeable to apply PoF to a complex system of which limited number of models and their associated parameters are available [18]. Therefore, it is important to identify and to focus on the critical failure mechanisms in specific applications. 2) From Constant Failure Rate to Mean Cumulative Function (MCF) Curve The conventional reliability metrics constant failure rate (defined as λ) and the corresponding Mean-Time-Between- -Failures (MTBF) (defined as 1/λ) are found to be inappropriate Fig. 2. Example of MCF or M (t) curve for explaining and measuring reliability. to most practical cases as discussed in [9], [13] and [20]. Therefore, it is discouraged the indiscriminate use of these metrics. The failure rate over operational time is not constant. An alternate technique to present the failure level and time is the MCF curve [21]. When analyzing repairable systems, it graphs the number of failures versus time (i.e. since installation). It is also possible to represent the behavior of the group of systems by an average number of failures versus time, which is known as MCF. As shown in Fig. 2, it can be broken down to the main functions which can be described as following parameters: Zero time failures occur from lack of robustness to transportation or installation, indicated by the red line. Early failures come from lack of production capabilities. Some few products are slipped through the control parameters in the production, which are marked as the green Weibull curve (β << 1, where β is defined as the shape parameter in Weibull distribution [22]). If the product is not robust to Catastrophic stress, the product might fail. This weakness is designed in and the time when failure occurs has nothing to do with the age of the product. The only way such accident can be shoved as random in the operation time (β = 1), marked in orange curve. This has nothing with fate rate-values or MTBF to do. The last dominant curve (i.e. the blue one) is the Lack of lifetime. This is the accumulated degradation for all parameters which are able to degrade as a function of operational time. The customer will be the person who sees the accumulated failure level of all these weaknesses in the purple curve in Fig. 2. This figure is also an integration of the bathtub model, but here it is possible to operate with quantitative figures, which can be broken down in budgets (e.g. the degradation budget). 3) From Reliability Prediction to Robustness Validation Conventional empirical methodologies mainly attempt to determine the feasibility in fulfilling certain reliability goals and to predict the warranty-costs and maintenance-support requirements [12]. They provide limited insights in the design of the systems themselves to eliminate failures within targeted service life. Compared to them, the concept of PoF is to identify the root causes of different types of failure under environmental and operation stress conditions. Therefore, it helps locate the weak-links and formulate the corresponding guidelines on robustness design, process control, validation testing and filed operation. Take the design as an example, Fig. 3 describes how

5 4 Fig. 3. Concept for robustness design. the designer shall understand degradation. Products should be designed by considering the degraded parameters at the end-of-life with certain level of design margins. It is also show that it makes only sense to measure performance inside the design specifications. The optimal situation is that the design is so good that no weakness can be found inside the customer or design specifications. All validations to reliability, lifetime and robustness have to be demonstrated outside or at design specifications under relative high stress levels. 4) From Microelectronics to also Power Electronics The PoF approach has been extensively applied to microelectronic systems in the last two decades. Different failure mechanisms, lifetime models, and equivalent damaged circuit simulation models of electron devices are well presented in [16]. More and more new models are under development. Several PoF based industry standards or guidelines have been released (e.g. [2] and [18]). One of the common driven factors from industry, academia and military behind this is the demanding for more reliable commercial-of-the-shell devices and systems. In power electronic applications, reliability has been and will continue to be one of the important performance aspects in many applications as discussed in II Part A. To address the challenges discussed in I, power electronic engineers and scientists have started to apply various reliability tools for reliability prediction and reliability-oriented design of power electronic converters or systems. Several literature reviews on field experiences [23], strategies to improve reliability of power electronic systems [24] and DFR for power electronic systems [9] have been presented in the last two years. Respective research in different applications is also discussed in various literatures, such as three-phase converters for aircrafts [25], power inverters for railway tractions [26], inverters for hybrid electric vehicles [27], high power variable-speed motor drives [28], and pulsed power converters for industrial process control [29]. Besides these applications, last decade also saw much pioneering work on the reliability of power converters for WTs [30] [32] and inverters for PV systems [33] [50]. It reveals that, unlike the case in microelectronics, conventional handbook methods are still dominantly applied nowadays for the reliability prediction in those studies. While the pace of power electronics toward PoF approach is relatively slower than that of microelectronics, the need for this paradigm shift has been well recognized in automotive industry [2] and then in other sectors. Especially, much interesting work from the semiconductor side investigates the failure mechanisms of IGBT modules [51] and physical based lifetime models [52]. More realistic thermal stress analysis of Si and SiC based devices under long term mission profile are also studied in [49] and [50], respectively. The level of technology and scientific understanding are still highly evolving. The research in microelectronics could provide a very important foundation for the ongoing and future work in power electronics, especially from the methodologies point of view. Nevertheless, it should be noted that most of the physical based models are not scalable for power electronic components. System level reliability problems (e.g. active thermal stresses, interconnections among components, interaction of different components) are still of interest to be investigated. Therefore, the following three sections intend to provide a basic framework of the future reliability research in power electronics relevant to the ongoing paradigm shift. III. POF ANALYSIS OF RELIABILITY CRITICAL COMPONENTS IN POWER ELECTRONICS As shown in Fig. 1(b), understanding of the reliability physics of components applied in power electronics is the most fundamental aspect. The PoF approach is based on analyzing and modeling each failure mechanism under various environmental and usage stresses. In practice, the PoF analysis focuses

6 5 TABLE III FOCUS POINTS MATRIX (FPM) IN RELIABILITY OF POWER ELECTRONIC COMPONENTS Ambient Load Climate + Design => Stressor Product design Active power components Stressors Die LASJ Wirebond Passive power components Focus points Cap. Ind. Solder Joint Control circuitry, IC, PCB, connectors MLCC IC PCB Connectors Relative humidity -RH(t) Temperature -T(t) -thermal system -operation point -ON/OFF -power P(t) Temperature swing ΔT X X X X Average Temperature T X X X X X X x x x dt/dt x x x x Water X X x Relative x x x Humidity X x x x X X x Pollution Tightness Pollution x x Mains Circuit Voltage x x x X X x x x x Cosmic Circuit Voltage x Mounting Mechanical Chock /vibration x x x x x x LASJ - Large Area Solder Joint, MLCC - Multi-Layer Ceramic Capacitor, IC- Integrated Circuit, PCB Printed Circuit Board, Cap. - Capacitor, Ind. - Inductor, Level of importance (from high to low): X-X-X-x. on critical components under critical stress conditions. Among other components, switching devices and capacitors are two of the most vulnerable components in terms of failure level and time as analyzed in [23], [34], [35] and [53]. They are considered as the reliability critical components in power electronic converters, especially the IGBT modules in medium to high power applications and capacitors for DC-link applications. Therefore, in the following parts, the critical stressors for different power electronic components is firstly discussed. Then, the PoF analysis of IGBT modules and DC-link capacitors is given. A. Critical Stressors for Different Power Electronic Components Focus Point Matrix (FPM), as suggested in [2], is a useful way to analyze the critical stressors that will kill the components. Based on the accumulated industrial experiences and future research needs, Table III shows the critical stressors for different components in power electronic systems. It can be noted that steady-state temperature, temperature swings, humidity, voltage and vibrations have different level of impact on semiconductor devices, capacitors, inductors and low power control boards. It provides the information on determining the critical failure mechanisms. The interactions among different stressors are also of interest to be explored. Fig. 4. Structural details of an IGBT module (connections that are relevant to module lifetime are marked red) [54]. B. PoF Analysis of IGBT Modules Fig. 4 shows a typical structure of IGBT modules [54]. There are three dominant wear out failure mechanisms for IGBT modules due to cyclic thermal stress: baseplate solder joints cracking, chip solder joint cracking, and the wire bonds lift-off. The cyclic thermal stress is a response to the converter line and Fig. 5. A typical stress-strain (σ-ε) curve for a material [55]. loading variations as well as periodically commutation of power switching devices. It will induce thermal cycling on different layers of materials used for fabrication of power electronic components.

7 6 Fig. 6. Typical catastrophic failure of IGBT modules [60]. Thermal cycling is found to be one of the main drivers for failure of IGBT modules. The effect of the temperature cycling can be explained by the typical stress-strain curve [55] shown in Fig. 5. σ is defined as the cyclic stress (e.g. temperature cycling) and ε is defined as the deformation. With a low cyclic stress below σ yield, no damage occurs and the material is in the elastic region. When the stress is increased above σ yield, an irreversible deformation is induced and the material enters into the plastic region. The coefficients of thermal expansion of different materials in the IGBT modules are different, leading to stress formation in the packaging and continuous degradation with each cycle until the material fails. As derived in [9], the number of cycles to failure under thermal cycling can be obtained as 0 N k T - T m (1) where k and m are empirically-determined constants and N is the number of cycles to failure. T is the thermal cycling range and T 0 is the portion of T that in the elastic strain range. If T 0 is negligible compared to T, it can be dropped out from the above equation, which then becomes the Coffin-Manson model discussed in [56]-[58]. The model shown in (1) considers the influence of thermal cycling only. It does not take into account the effect of steady- -state temperature, thermal cycle time, and geometry. In [59], an empirical model is developed for bond wire fatigue of IGBT modules, which tends to treat all the above factors as well as failure of the diodes in parallel with the IGBT switches. In [52], a physical based model for wire bond fatigue has been developed which could analyze the cycle-to-failure under different steady-state temperature and thermal cycle time. Although it may be difficult to obtain some of the parameters required by the model, it is a promising model in its kind for the PoF analysis. Besides wear out failure discussed above, different types of catastrophic failure could also occur triggered by single-event overstress. Unlike the wear out failure, the catastrophic failure is difficult to be predicted and thus may lead to serious consequence to the power electronic converters. Fig. 6 classifies the IGBT catastrophic failure into open-circuit mode and short- -circuit mode induced by different failure mechanisms. It should be noted that both wear out failure and catastrophic failure may have the same failure mechanisms (e.g. bond wire lift-off) but the former one is due to long term degradation (see the blue curve in Fig. 2) and the latter one is due to single-event overstress within short time duration (see the orange curve in Fig. 2). C. PoF Analysis of DC-Link Capacitors DC-link capacitors contribute to cost, size and failure of power electronic converters on a considerable scale. To address the issue, research efforts can be divided into two directions: a) advance the capacitor technology with improved and predetermined reliability built in and b) a proper and optimal DC-link design based on the commercially available capacitors to ensure reliable field operation. The latter one is more relevant from the perspective of power electronic designers, which is discussed here. Three main types of capacitors are available for DC-link applications, which are the Aluminum Electrolytic Capacitors (Al-Caps), Metallized Polypropylene Film Capacitors (MPPF- -Caps) and high capacitance Multi-Layer Ceramic Capacitors (MLC-Caps). The DC-link design requires the matching of available capacitor characteristics and parameters to the particular application needs under specific environmental, electrical and mechanical stresses. Table IV summarizes the failure modes, critical failure mechanisms and corresponding stressors. More detailed discussions on them have been given in [61]. Table V gives the wear out failure criterion and typical electrical parameters for the condition monitoring of capacitors. Lifetime prediction of capacitors is mainly based on empirical models as physical based models are still not available. The most widely used empirical model for capacitors is shown in (2) which describes the influence of temperature and voltage stress. n V E a 1 1 L L0 exp V0 K B T T0 where L and L 0 are the lifetime under the use condition and testing condition, respectively. V and V 0 are the voltage at use condition and test condition, respectively. T and T 0 are the temperature in Kelvin at use condition and test condition, respectively. E a is the activation energy, K B is Boltzmann s constant ( ev/k), and n is the voltage stress exponent. Therefore, the values of E a and n are the key parameters to be determined in the above model. (2)

8 7 TABLE IV OVERVIEW OF FAILURE MODES, CRITICAL FAILURE MECHANISMS AND CRITICAL STRESSORS OF THE THREE MAIN TYPES OF DC-LINK CAPACITORS (WITH EMPHASIS ON THE ONES RELEVANT TO DESIGN AND OPERATION OF POWER CONVERTERS) Cap. type Failure modes Critical failure mechanisms Critical stressors Al-Caps MPPF-Caps MLC-Caps Open circuit Self-healing dielectric breakdown Disconnection of terminals V C, T a, i C Vibration Short circuit Dielectric breakdown of oxide layer V C, T a, i C Wear out: electrical parameter drift (C, ESR, tanδ, I LC, R p) Open circuit (typical) Short circuit (with resistance) Wear out: electrical parameter drift (C, ESR, tanδ, I LC, R p) Short circuit (typical) Wear out: electrical parameter drift (C, ESR, tanδ, I LC, R p) Electrolyte vaporization Electrochemical reaction (e.g. degradation of oxide layer, anode foil capacitance drop) Self-healing dielectric breakdown Connection instability by heat contraction of a dielectric film Reduction in electrode area caused by oxidation of evaporated metal due to moisture absorption Dielectric film breakdown Self-healing due to overcurrent Moisture absorption by film Dielectric loss Dielectric breakdown Cracking; damage to capacitor body Oxide vacancy migration; dielectric puncture; insulation degradation; micro-crack within ceramic T a, i C V C V C, T a, dv C/dt T a, i C Humidity V C, dv C/dt T a, i C Humidity V C, T a, i C, humidity V C, T a, i C Vibration V C, T a, i C, vibration C -capacitance, ESR-equivalent series resistance, tanδ -dissipation factor, R p -insulation resistance, V C -capacitor voltage stress, i C -capacitor ripple current stress, i LC -leakage current, T a -ambient temperature. TABLE V WEAR OUT FAILURE CRITERION AND PARAMETERS FOR CONDITION Wear out failure criterion (typical) Condition monitoring parameters MONITORING OF CAPACITORS Al-Caps MPPF-Caps MLC-Caps CC0 80% ESR ESR0 200% CC0 95% tan tan 3 Rp Rp0 1.5% C and ESR C R p, C and tanδ C 0 -initial capacitance, ESR 0 initial equivalent series resistance, tanδ 0 -initial dissipation factor, R p0 -initial insulation resistance. However, the voltage dependency of lifetime for Al-Caps quite depends on the voltage stress level. In [62], instead of a power law relationship, a linear equation is found to be more suitable to model the impact of voltage stress. In order to obtain 0 7 R 10 p CC0 90% tan tan 2 0 the physical explanations of the lifetime model variants from different capacitor manufacturers, a generic model is derived in [9] as shown in (3). Where a 0 and a 1 are constants describing the voltage and temperature dependency of E a. E a0 is the activation energy under test. It can be noted that the influence of voltage stress is modeled as linear, power law, and exponential relationship, respectively for low-voltage stress, medium-voltage stress and high-voltage stress. Another important observation is that the activation energy E a is varying with voltage and temperature, especially under high-voltage stress condition. It is still a challenge to determine the value of the parameters and the boundaries of low-voltage stress, medium-voltage stress and high-voltage stress in (3). D. Cases Studies on the Application of IGBTs and Capacitors in Power Converters 1) IGBT Modules in a 2.3 MW Grid-Side Wind Power Converter V0 E 1 1 a exp (low voltage stress) V K B T T0 n L V E a 1 1 exp ( medium voltage stress) L0 V0 K B T T0 Ea0 a0v Ea0 a0v 0 exp a1 V 0 V exp (high voltage stress) KBT KBT 0 (3)

9 8 (a) Simplified structure. Fig. 7. Two-level back-to-back converter for a 2.3 MW wind turbine using a Permanent Magnet Synchronous Generator (PMSG) [9]. TABLE VI CONVERTER PARAMETERS FOR THE CASE STUDY [9] Topology 2L-BTB as shown in Fig. 7 Rated output active power DC bus voltage 2.3 MW 1.1 kv DC * Rated primary side voltage 690 V rms Rated load current Switching frequency 1.93 ka rms 1950 Hz Filter inductance 132 µh IGBT Selection I (grid side) IGBT Selection II (grid side) 1.6 ka/1.7kv/125ºc, two in parallel 2.4 ka /1.7kV/ 150ºC, single switch * Line-to-line voltage in the primary windings of transformer. TABLE VII LIFETIME PREDICTION RESULTS OF THE SELECTED IGBT MODULES Failure mechanisms B 10 lifetime (year) Selection I Baseplate solder joints IGBT chip solder joints Wire bonds Overall (determined by the shortest one) Section II A case study on the lifetime prediction of IGBT modules in a 2.3 MW wind power converter has been studied in [9]. A Two-Level Back-to-Back (2L-BTB) converter is applied in the study as shown in Fig. 7. The technical advantage of the 2L-BTB topology is the relatively simple structure and few components, which contributes to a well-proven robust and reliable performance. Table VI gives the specifications and selections of the IGBT modules. By following the prediction procedure from wind speed profile analysis, case temperature and junction temperature estimation, cycling counting of temperature swings to parameter estimation of lifetime models, the lifetime of two condidates of IGBT modules for the grid-side (b) Instantaneous power flow. Fig. 8. Single-phase grid-connected PV inverter. converter is predicted in [9]. The lifetime prediction is based on each of the three critical failure mechanisms related to thermal cycling discussed in III Part A. The results are shown in Table VII. It should be noted that other failure mechanisms induced by thermal stress or other types of stresses need also to be considered, besides those listed in Table VII. 2) DC-Link Capacitors in a 1 kw PV Inverter Electrolytic capacitors that widely used in PV inverters are considered as the weakest link with respect to the semiconductor devices [34] [35]. Therefore, the case study for DC-link capacitors is performed on a 1 kw 400 V DC-link PV inverter. Fig. 8(a) presents a simplified structure of the inverter. The input power of the PV inverter is assumed constant within one cycle of the grid voltage. Fig. 8(b) shows the instantaneous power balancing function of the input capacitor C. The nominal input voltage of the inverter is 400 V with a maximum voltage ripple of 5% and a maximum input voltage of 600 V. The calculated minimum required capacitance is 398 µf and ripple current stress is 1.8 A. A reliability-oriented design guideline proposed in [63] is applied for the selection of the input capacitor to fulfill 20 years of lifetime. According to the electrical stress analysis, preliminary choices of the capacitors are determined as shown in Table VIII. Then the thermal stresses of those capacitors are estimated based on their specific thermal models. The lifetime of the selected capacitors is therefore can be estimated based on the mission profile, operation mode and specific lifetime model. The applied empirical lifetime models from the respective capacitor manufacturers are consistent with the generic lifetime model shown in (3). Finally, the optimal capacitors can be chosen by comparing different options. Case No. TABLE VIII THREE KINDS OF CAPACITORS FROM DIFFERENT MANUFACTURES CONSIDERED FOR THE PV INVERTER DESIGN Capacitor bank Rated lifetime (85 ) ESR at 100 Hz (mω) four 350 V/470 µf /1.9A 1,000 hours two 315 V/1000 µf /3.63A 2,000 hours two 350 V/1000 µf/ 5.5A 24,000 hours Natural cooling thermal resistance R th ( /W)

10 9 method allows the optimal selection of the input capacitors in terms of both electrical and reliability performance. According to Fig. 9(d), for applications when output power de-rating is allowed, the required 100,000 hours of lifetime could still be fulfilled for selection of Case 1 and Case 2 by load management according to the de-rating curves. (a) Capacitor power losses. (b) Capacitor hotspot temperatures. (c) Capacitor predicted lifetime. (d) Output power de-rating curves. Fig. 9. Simulation results of different capacitors under various ambient temperatures ((a)-(c) are with 1 kw output power and (d) is with minimum lifetime of 20 years). Figs. 9(a)-(c) compare the power loss, hotspot temperature, lifetime prediction with 1 kw output power. Fig. 9(d) plots the power de-rating curve to fulfill the lifetime requirement. If 100,000 hours of lifetime (equivalent to 12 hours/day operation in 20 years with a design margin of 12.4%) is required, it can be noted from Fig. 9 (c) and Fig. 9 (d) that only the Case 3 can fulfill the requirement in a wide ambient temperature range. The selection of Case 2 can only have 20 years of lifetime when the ambient temperature is below 20. Therefore, the proposed IV. DESIGN FOR RELIABILITY AND ROBUSTNESS VALIDATION OF POWER ELECTRONICS The second aspect of power electronics reliability is to build in reliability and sufficient robustness into system design through DFR process. Industries have advanced the development of reliability engineering from traditional testing for reliability to DFR [64]. DFR is the process conducted during the design phase of a component or system that ensures them to be able to achieve required level of reliability. It aims to understand and fix the reliability problems up-front in the design process. In [65], a structured approach to DFR was recommended which include an interactive progression of key design activities by using appropriate analysis tools. Due to the difference in chosen reliability tools and specified requirements of products, DFR process varies with industry sectors, however, the generic form usually covers the process of identify, design, analyze, verify, validate and control [64]. In [66], a DFR process is presented for aerospace systems starting from the concept, planning and requirement, development, test and evaluation, release and evaluate the field operation performance. Statistics is a necessary basis to deal with the effects of uncertainty and variability on reliability, which is also true for PoF based DFR process. Both empirical models and physical based models are subjected to various kinds of uncertainty in material properties, stress conditions, manufacture process and accuracy of the available models [67]. For example, if the variances of material-dependent parameters k, m and T 0, and operational-stress-dependent parameter T are taken into account in (1), the predicted lifetime of the IGBT modules will be distributed with time rather than the single fixed values as shown in Table VII. The 2-parameter Weibull distribution is widely applied to present the cumulative failure distribution function [64] due to its flexibility to model different trends of the failure level with time. In [68], the probabilistic based PoF models have been developed for plastic package corrosion. Moreover, Monte-Carlo simulation is another important statistics tool to handle uncertainties to analyze the robustness margin, confidence level, and so on [69]. However, as the variation is often a function of time and operating condition, statistics itself is not sufficient to interpret the reliability data without judgment of the assumptions and non-statistical factors (e.g. modification of designs, new components, etc.). A systematic DFR procedure specifically applicable to the design of power electronic products is presented in [9] and shown in Fig. 10. By implementing the procedure, reliability is well considered and treated in each development phase (i.e. concept, design, validation, production and release), especially in the design phase. The design of power electronic converters

11 10 Fig. 10. State-of-the-art reliability design procedure for power electronics. is mission profile based by taking into account large parametric variations (e.g. temperature swings, solar irradiance level changes, wind speed fluctuations, load changes, manufacturing process, etc.). Important concepts (except for PoF approach which has been discussed in III) and design tools shown in Fig. 10 are discussed as follows. More detailed discussions are given in [9]. A. Load-Strength Analysis Load-strength analysis is an important method in the first step of the design phase shown in Fig. 10. A component fails when the applied load exceeds the design strength. The load refers to a kind of stress (e.g. voltage, cyclic load, temperature, etc.) and the strength refer to any resisting physical property (e.g. harness, melting point, adhesion, etc.) [64]. Load and strength of power electronic components are allocated within a certain interval which can be described by a specific probability density function (e.g. normal distribution). Moreover, the strength of a material or device could be degraded with time. Theoretically, the probability of failure can be obtained by analyzing the overlap area between the load distribution and the strength distribution. From another prospective, it implies that failure could be reduced or eliminated within service life by either design with an increased strength (i.e. an increased design margin), or with a reduced load by control (i.e. stress control or load management), or both. Practically, the exact distributions of load and strength are very often not available, Monte Carlo simulation [64] can be applied to randomly select samples from each distribution, compare them and thus roughly estimate the probability of failure. B. Reliability Prediction Toolbox Reliability prediction (not based on constant failure rate λ) is an important tool to quantify the lifetime, failure level and design robustness based on various source of data and prediction models. Fig. 11 presents a generic prediction toolbox based on the PoF approach. The toolbox includes statistical models and lifetime models and various sources of available data (e.g. manufacturer testing data, simulation data and field data, etc.) for the reliability prediction of individual components and the overall system. The statistical models are well presented in [64], while the number of physical based lifetime models available for power electronic components is still limited. Research efforts to both accelerated testing and advanced multidisciplinary simulations will be beneficial to obtaining those lifetime models. A more detailed step-by-step procedure for lifetime prediction is presented in [70].

12 11 Fig. 11. Reliability prediction toolbox for power electronic systems. TABLE IX SUMMARY OF SYSTEM LEVEL RELIABILITY PREDICTION METHODS Reliability Block Diagram (RBD) Fault Tree Analysis (FTA) Markov Analysis (MA) Concepts Elements RBD is an analytical technique graphically representing the system components and their reliability-wise connections (from simple series-parallel to complex) by a logic diagram based on the system characteristics. Rectangle blocks Direction lines Failure level and time of the component/subsystem represented by each blocks FTA is an analytical technique using a top-down approach to analyze various system combinations of hardware, software and human failures (i.e. sub events) that could cause the system failure (i.e. top event). Events (i.e. initiating fault events, intermediate events and top event) Logic gates (e.g. AND, OR and more complex ones) Probability of each event Outcome System level reliability System level reliability Identified all possible faults (similar to the results from FMEA) Applications For non-repairable systems Without redundancy With redundancy For non-repairable systems Without redundancy With redundancy Advantages Simplicity and ease of application All factors including human factors could be taken into account Useful also for identifying failure causes and design problems Disadvantages/ Limitations Limitation in considering external events (e.g. human factor) and priority of events Dependencies among components or subsystems are not well treated Dependencies among components/subsystems are not well treated MA is a dynamic state-space analytical technique presenting all possible system states (i.e. functioning or failed) and the existing transitions between these states. States (i.e. functioning or failed) Transitions between states Transition rates based on failure rates and repair rates of components/ subsystems System level reliability System availability Mainly for repairable systems Without redundancy With redundancy Dynamic (i.e. represent state of every component at any time and the dependences among them) Applicable for repairable systems State-based models easily get large (e.g. maximum 2 n states with n components) Primarily applicable for constant failure rate and constant repair rate (which works in theory only) To map the reliability from component level to the system level [3], Reliability Block Diagram (RBD), Fault-Tree Analysis (FTA) and state-space analysis (e.g. Markov analysis) are widely applied as summarized in Table IX. It should be noted that the tabulated three methods are conventionally applicable to constant failure rate cases, which are corresponding to the handbook based reliability prediction methods. The PoF based system level reliability prediction is

13 12 still an open research topic even in microelectronics [16] and [18]. Interactions among different failure mechanisms will bring additional complexity for the analysis. Therefore, it was argued that the PoF approach is not practical for assessing an entire system in [12]. Moreover, it should be noted that the system reliability depends not only on components, but also on packaging, interconnects, manufacturing process, and human errors. The latters need also to be treated properly for a more accurate reliability assessment. V. INTELLIGENT CONTROL AND MONITORING OF POWER ELECTRONIC SYSTEMS After power electronic systems have been designed, their reliability could be further improved through control and condition monitoring. This is the third important aspect shown in Fig. 1(b). Among many options, three main actions can be taken to increase the reliability of power electronic systems: prognostics and health management, active thermal control for reducing temperature and temperature swing that are the main killing factors of power device modules and fault tolerant operation to continue operate the system even in case of failures. The last can be considered as an alternative measure with respect to the first two or like the last attempt to make the system operating if it was not possible to predict failures or to avoid them. Of course all these actions entail important investments in terms of devices, sensors and control actions and even request redundancies. All of them shall be evaluated in terms of cost respect to the specific application. A. Prognostics and Health Management Fig. 12. Prognostics and health management of power electronic systems. The Electronic Prognostics and Health Management Research Center at the University of Maryland has categorized the main approaches as: use of fuses and canary devices, built-in- -test (BIT), monitoring and reasoning of failure precursors, and modeling accumulated damage based on measured life-cycle loads [71]. Apart from the first category that relies on the presence of devices that fail before the main one (like the canary in a mine full of hazardous gas), the three other categories need: 1) sensors to check the functionality of a device or circuit, like in case of BIT, to monitor precursors of the failure or to measure cycles whose number can be correlated to failure and 2) data-logging to store the data coming from sensors. Accordingly, Fig. 12 shows the configurations for the prognostics and health management. The last three categories can in part overlap with the concept of condition monitoring that refers to on-line monitoring of the device. In case of power semiconductor this can be done by modifying the gate driver circuit. Different sensors can be employed and can be classified in two categories: ambient sensors (temperature, humidity and pollution) and internal sensors (module temperature, vibration, electrical parameters). Since the main failure cause for power module is the junction temperature swing, measuring it or estimating by means of Thermo-Sensitive Electrical Parameters (TSEPs) is one of the most interesting challenges [72]. In fact sensing the junction temperature during converter operation is notoriously difficult - direct access to chips is prevented by module packaging and dielectric gel, which therefore limits the optical and physical contact methods such as the use of infrared cameras or optical fibers. Electrical methods allow the measurement of temperature without any physical alteration to a device. However, one disadvantage compared to optical or physical contact methods is that the latter can be used to measure the temperature at specific points in the die or in the module. Electrical methods generally give an average temperature across the die. For instance, the voltage drop across a PN-junction is known to vary with temperature - a measurement of this voltage can therefore be used to derive a temperature solely for the junction. However, this perhaps does not give a reliable enough estimation of temperature elsewhere in the devices, such as in bond wires, solder joints, etc. Data acquisition is very important since once determined the important quantities to be sensed or estimated, there are issues related to the amount of data that is possible to store and how to use those data. Hence developing a data-acquisition system taking into account both the ambient and the internal quantities can be a challenge [73]. The goal is to have the real-time operating characteristics and the health conditions of the components (particularly of the power modules and of the capacitors) and of the overall power converter [74]. This information can be used for two main goals: implement a proactive maintenance plan (i.e. prognostic maintenance) and provide information for proactive control schemes that can be a simple load management (i.e. reduction or sharing of the load among different units) or the more advanced active thermal control, briefly introduced in the following. B. Active Thermal Control The thermal analysis of power converters, especially in case of more complex structures like multilevel or multi-cell ones, reveals that some of the power semiconductor devices can be more stressed with respect to others and this difference can be even more evident in some particular conditions like those caused by system faults [75]. Hence the possibility to modify

14 13 Fig. 13. Active thermal control of the power semiconductor junction temperature T j by means of y (switching frequency, reactive power or any other quantity that can modify the power semiconductor losses). T j is obtained by using an estimator based on TSEP (Thermo-Sensitive Electrical Parameter) or an observer using measured voltages and currents. Fig. 14. Inverter faults considered: (a) single switch short-circuit, (b) phase- -leg short-circuit, (c) single switch open-circuit, and (d) single-phase open- -circuit [83]. the modulation and control of the power converter using as a feedback the junction temperature of the most stressed device is an appealing possibility. The more straightforward approach is manipulating the switching frequency and the current limit to regulate the losses and prevent over-temperature or to reduce temperature swing [76]. In view of controlling the junction temperature, estimating it by means of TSEP, as already mentioned, or using an observer based on FEM modeling of the device [76] are two interesting alternatives to the more expensive ones by using of integrated sensors in the chip [77]. Fig. 13 gives the general block diagram for active thermal control of the power semiconductors once the junction temperature is measured or estimated. The easiest way is to change the switching frequency [78]-[79] to control the junction temperature. In case of parallel power converters that present some redundancies it is also possible to share the load among the different units also in view of controlling the temperature swing [80]. Another alternative is to circulate reactive power among the different power converters connected in parallel in a high power converter or in a wind or photovoltaic park to reduce the temperature swing in the most stressed power semiconductor devices [81]. The idea can be applied only in case of power converters like the neutral-point-clamped inverters where there is an uneven distribution of power losses and as a consequence of the temperature of the power semiconductor devices, being this difference even bigger in some particular stressing circumstances like in the case of sudden power changes (e.g. for a wind gust), grid faults or variable atmospheric conditions. The main drawbacks are: higher losses and higher mean temperature of the most stressed device but also of the other devices. However, one risk is to move the stress from bond wires to solder joints. Fig. 15. Switch redundant topology for fault tolerant control [83]. Fig. 16. Redundancies by connecting IGBTs in parallel or in series [84]. C. Fault Tolerant Operation Working outside the Safe Operating Area (SOA) leads power semiconductors to damage. The main failure causes are: fault currents either over-current, short-circuit current or earth fault current, over-voltages, over-temperature and cosmic radiation [54]. Other problems may arise because of the driver of the power semiconductor: malfunctioning of the driver board, Fig. 17. Fault-tolerant voltage source inverter by adding extra leg for more electrical aircraft application [86].

15 14 auxiliary power supply failure or dv/dt disturbance. As a consequence five main types of faults can be identified: single switch short-circuit (power semiconductor is de-saturated working as current source or it is a physical short-circuit), phase-leg short-circuit, single switch open-circuit, single-phase open-circuit, intermittent gate-misfiring [82]. Fig. 14 describes the first four types of faults. Three levels of protections can prevent failures or limit their effects: fast (in the switch, 10 ns), slow (outside the switch) and very slow (system level). Several diagnostic methods can be used to detect failures and they can be mainly classified in those used for open-switch failure or closed-switch failure and in software one or hardware one. Generally software methods are more suitable in case of open-switch failure while closed- -switch failures need a fast detection because they can lead to destructive failure of the overall system, the desaturation one is the most famous [82]. Once a fault is detected and isolated or on-line repair is implemented, the system can continue to operate safely and fault tolerant operation can be implemented. There are already several simple solutions implemented in industry also without any redundancy if operation with high harmonic content and lower power level can be accepted [83] as shown in Fig. 15. Otherwise redundancies based on paralleling or connecting in series power semiconductor devices [84] as shown in Fig. 16 is the simplest and most adopted solution in industry and use of devices that can continue to operate in short-circuit like press-pack IGCT can help. The next step in using redundancy to improve fault-tolerant operation is in adding extra legs to the power converter. The procedure consists of the following steps: 1) detection of the faulty leg, 2) stop the control signal for the two switching drivers of the faulty leg, 3) trigger the bidirectional switch connecting the new leg, 4) use the control signals of the faulty leg for the redundant one. The use of fast digital computational devices [85] and the integration in the power module of the extra leg and of a thyristor that can survive high energy pulses [86] are the enabling technologies to isolate a faulty leg and on-line substitute it with a healthy one. The latter has been developed within an EU project for more electric aircraft as shown in Fig. 17, and this is an interesting sign of the importance of fault operation capability solutions that should be designed to guarantee high reliability in safety critical applications. The last two solutions to guarantee fault-tolerant operation entail larger investments at system level and in general a significant shift in the power converter design: multiphase power converters and machines [87]-[88] and use of parallel or series connection of power converters [89]-[90]. Fig. 18 and Fig. 19 show the redundancy achieved by connecting converters in parallel and in input-series output-parallel, respectively. Both of the solutions have been proposed and in some cases implemented in the aforementioned more electric aircraft [88]-[89]. Particularly connection in series and/or in parallel power converters is already widely adopted in dc/dc converters and requires the power modules to be identical and capable of working independently as the case shown in Fig. 19. Moreover the faulted module(s) must be quickly isolated from the system, which is not always easy during operation of the power Fig. 18. Redundancy achieved by connecting converters in parallel [89]. Fig. 19. Common duty ratio, automatic master-slave control scheme with identical and independent modules for input-series, output-parallel connection [90]. converter especially taking into account the associated transients that should be minimized to avoid damage of the healthy modules while replacing the faulty ones. VI. CONCLUSION AND OUTLOOK Reliability is an important performance index of power electronic systems. The status and future trends of design for reliability in power electronics are presented in this paper. A paradigm shift in reliability research on power electronics has left simple handbook based on constant failure rate for the PoF approach and DFR process. Accordingly, three major aspects of power electronics reliability are discussed: the PoF analysis of reliability critical components (e.g. IGBT modules and DC-link capacitors) and two associated study cases; the state-of-the-art DFR process and robustness validation for power electronic systems; the prognostics and health management, active thermal control and fault-tolerant strategies for reliable field operation. Joint efforts from engineers and scientists in the multiple

16 15 disciplines are required to fulfill the research needs and promote the paradigm shift in reliability research. The major challenges and opportunities in the research on reliability for power electronic systems are addressed as follows: A. Challenges Pervasive and fast implementation of power electronics in a large variation of applications with all kind of environmental exposures. Outdated paradigms and lack of understanding in the design for reliability process in power electronics. Uncertainties in mission profiles and variations in strength of components. Increasing electrical/electronic content and complexity. Lack of understanding in failure mechanisms and failure modes of reliability critical components. Traditional system level reliability prediction methods are based on constant failure rates. However, physics-offailure based component level reliability prediction results in varying failure level with time. Resource-consuming testing for reliability prediction and robustness validation from components to entire systems. End up with ppm level return rates for mass-manufactured power electronic products. Higher operating temperature (e.g. with wide band-gap devices) which challenges the overall reliability and lifetime. Software reliability becomes an issue with more and more digital controllers are introduced in power electronic systems, which should be treated adequately. B. Opportunities The research in microelectronics provides an important foundation for the ongoing and future work in power electronics, especially from the methodologies point of view. More and more mission profiles and on-line monitoring data from the field are available and accessible. Physics-of-failure approach provides insights to avoid failures in power electronic components, circuits and systems. Active thermal control by controlling the power flow in power electronic circuits. Component level and system level smart de-rating operation. Condition monitoring and fault tolerant design which allow extended lifetime and reduced failure rate. Emerging semiconductor and capacitor technologies enable more reliable power electronic components and systems. Computer-aided automated design software to save time and cost in the development process. Trends for modular design of power converters and standardized power electronic components and packaging technologies. With better understanding of failure mechanisms in power electronics, more failure mechanism specific accelerated testing could be designed, leading to improved reliability predictions for targeted applications. REFERENCES [1] W. E. Newell, Power electronics - emerging from limbo, IEEE Trans. Ind. Appl., vol. IA-10, no. 1, pp. 7-11, Jan./Feb [2] ZVEL, Handbook for robustness validation of automotive electrical/electronic modules, revised version, Jun [3] IEEE standard framework for the reliability prediction of hardware, IEEE Std. 1413, IEEE, 2009, New York. [4] T. Jomier, Final public MOET technical report, EU FP6 project on More Open Electrical Technologies, Dec [Online]. Available: [5] X. Perpinya, Reliability and safety in railway, ch7, InTech, 2012, ISBN: [6] F. Blaabjerg, Z. Chen, and S. B. Kjaer, Power electronics as efficient interface in dispersed power generation systems, IEEE Trans. Power Electron., vol. 19, no. 4, pp , Sep [7] F. Blaabjerg, M. Liserre, and K. Ma, Power electronics converters for wind turbine systems, IEEE Trans. Ind. Appl., vol.48, no.2, pp , Mar/Apr [8] S. B. Kjaer, J. K. Pedersen, and F. Blaabjerg, A review of single-phase grid connected inverters for photovoltaic modules, IEEE Trans. Ind. Appl., vol. 41, no. 5, pp , Sep./Oct [9] H. Wang, K. Ma, and F. Blaabjerg, Design for reliability of power electronic systems, in Proc. IEEE Industrial Electronics Society Annual Conference, 2012, pp [10] Reliawind, Report on wind turbine reliability profiles - field data reliability analysis, [Online]. Available: /files/file-inline/110502_reliawind Deliverable_D.1.3ReliabilityProfilesResults.pdf. [11] L. M. Moore and H. N. Post, Five years of operating experience at a large, utility-scale photovoltaic generating plant, Journal of Progress in Photovoltaics: Research and Applications, vol. 16, no. 3, pp , May [12] W. Denson, The history of reliability prediction, IEEE Trans. Reliability, vol. 47, no. 3-SP, pp. SP321-SP328, Sep [13] M. G. Pecht and F. R. Nash, Predicting the reliability of electronic equipment, in Proc. IEEE, vol. 82, no. 7, pp , Jul [14] K. Chatterjee, M. Modarres, and J. B. Bernstein, Fifty years of physics of failure, Journal of the Reliability Information Analysis Center, vol. 20, no. 1, Jan [15] Military Handbook: Reliability prediction of electronic equipment, MIL- HDBK-217F, Dec. 2, [16] M. White and J. B. Bernstein, Microelectronics Reliability: Physics-offailure based modeling and life evaluation, Jet Propulsion Laboratory, [17] J. G. Mecleish, Enhancing MIL-HDBK-217 reliability predictions with physics of failure methods, in Proc. IEEE Annual Reliability and Maintainability Symposium, 2010, pp [18] ANSI/VITA 51.2 Standard, Physics of failure reliability predictions, VEMbus International Trade Association, [19] M. Pecht and A. Dasgupta, Physics-of-failure: an approach to reliable product development, in Proc. International Integrated Reliability Workshop, 1995, pp [20] M. Krasich, How to estimate and use MTTF/MTBF would the real MTBF please stand up? in Proc. IEEE Annual Reliability and Maintenance Symposium, 2009, pp [21] W. B. Nelson, Recurrent-events data analysis for repairs, disease episodes, and other applications, ASASIAM Series on Statistics and Applied Probability, Society for Industrial and Applied Mathematics, Philadelphia, PA, [22] W. Weibull, Statistical distribution function of wide applicability, ASME Journal of Applied Mechanics, vol. 18, no. 3, pp , Sep [23] S. Yang, A. T. Bryant, P. A. Mawby, D. Xiang, L. Ran, and P. Tavner, An industry-based survey of reliability in power electronic converters, IEEE Trans. Ind. Appl., vol. 47, no. 3, pp , May/Jun [24] Y. Song and B. Wang, Survey on reliability of power electronic systems, IEEE Trans. Power Electron., vol. 28, no. 1, pp , Jan.

17 [25] R. Burgos, C. Gang, F. Wang, D. Boroyevich, W. G. Odendaal, and J. D. V. Wyk, Reliability-oriented design of three-phase power converters for aircraft applications, IEEE Trans. Aerospace and Electronic Systems, vol. 48, no. 2, pp , Apr [26] X. Perpiñà, X. Jordà, M. Vellvehi, J. Rebollo, and M. Mermet-Guyennet, Long-term reliability of railway power inverters cooled by heat-pipebased systems, IEEE Trans. Ind. Electron., vol. 58, no. 7, pp , Jan [27] D. Hirschmann, D. Tissen, S. Schroder, and R. W. De Doncker, Reliability prediction for inverters in hybrid electrical vehicles, IEEE Trans. Power Electron., vol. 22, no. 6, pp , Nov [28] P. Wikstrom, L. A. Terens, and H. Kobi, Reliability, availability, and maintainability of high-power variable-speed drive systems, IEEE Trans. Ind. Appl., vol. 36, no. 1, pp , Jan./Feb [29] F. Carastro, A. Castellazzi, J. Clare, and P. Wheeler, High-efficiency high-reliability pulsed power converters for industrial processes, IEEE Trans. Power Electron., vol. 27, no. 1, pp , Jan [30] M. Boettcher and F. W. Fuchs, Power electronic converters in wind energy systems - Considerations of reliability and strategies for increasing availability, in Proc. European Conference on Power Electronics and Applications, 2011, pp [31] S. S. Smater and A. D. Dominguez-Garcia, A framework for reliability and performance assessment of wind energy conversion systems, IEEE Trans. Power Systems, vol. 26, no. 4, pp , Nov [32] K. Fischer, F. Besnard, and L. Bertling, Reliability-centered maintenance for wind turbines based on statistical analysis and practical experience, IEEE Trans. Energy Conversion, vol. 27, no. 1, pp , Mar [33] IEA International Agency, Reliability study of grid connected PV systems Field experience and recommended design practice, Report IEA-PVPS T7-08: 2002, Mar [34] D. Ton and W. Bower, Summary report on the DOE high-tech inverter workshop, US Department of Energy, Jan [35] Navigant Consulting, Inc., A review of PV inverter technology cost and performance projection, National Renewable Energy Lab Report Subcontract Report NREL-SR , Jan [36] S. Atcitty, J. E. Granata, M. A. Quintana, and C. A. Tasca, Utility-scale grid-tied PV inverter reliability workshop summary report, Sandia Report SAND , Jul [37] E. L. Meyer and E. E. Van Dyk, Assessing the reliability and degradation of photovoltaic module performance parameters, IEEE Trans. Reliability, vol. 53, no. 1, pp , Mar [38] C. Rodriguez and G. Amaratunga, Long-lifetime power inverter for photovoltaic AC modules, IEEE Trans. Ind. Electron., vol. 55, no. 7, pp , Jul [39] A. Ristow, M. Begovic, A. Pregelj, and A. Rohatgi, Development of a methodology for improving photovoltaic inverter reliability, IEEE Trans. Ind. Electron., vol. 55, no. 7, pp , Jul [40] G. Petrone, G. Spagnuolo, R. Teodorescu, M. Veerachary, and M. Vitelli, Reliability issues in photovoltaic power processing systems, IEEE Trans. Ind. Electron., vol.55, no.7, pp , Jul [41] F. Chan and H. Calleja, Design strategy to optimize the reliability of grid-connected PV systems, IEEE Trans. Ind. Electron., vol. 56, no. 11, pp , Nov [42] R. S. Balog, Y. Kuai, and G. Uhrhan, A photovoltaic module thermal model using observed insolation and meteorological data to support a long life, highly reliable module-integrated inverter design by predicting expected operating temperature, in Proc. IEEE Energy Conversion Congress and Exposition, 2009, pp [43] F. Chan and H. Calleja, Reliability estimation of three single-phase topologies in grid-connected PV systems, IEEE Trans. Ind. Electron., vol. 58, no. 7, pp , Jul [44] S. V. Dhople, A. Davoudi, A. D. Dominguez-Garci, and P. L. Chapman, A unified approach to reliability assessment of multiphase dc-dc converters in photovoltaic energy conversion systems, IEEE Trans. Power Electron., vol. 27, no. 2, pp , Feb [45] P. Zhang, Y. Wang, W. Xiao, and W. Li, Reliability evaluation of grid-connected photovoltaic power systems, IEEE Trans. Sustainable Energy, vol. 3, no. 3, pp , Jul [46] S. Harb and R. S. Balog, Reliability of candidate photovoltaic module-integrated-inverter (PV-MII) topologies A usage model approach, IEEE Trans Power Electron., vol. 28, no. 6, pp , Jun [47] E. Koutroulis and F. Blaabjerg, Design optimization of transformer-less grid-connected PV inverters including reliability, IEEE Trans. Power Electron., vol. 28, no. 1, pp , Jan [48] S. Saridakis, E. Koutroulis, and F. Blaabjerg, Optimal design of modern transformerless PV inverter topologies, IEEE Trans. Energy Conversion, vol. 28, no. 2, pp , Jun [49] Y. Yang, H. Wang, F. Blaabjerg, and K. Ma, Mission profile based multi-disciplinary analysis of power modules in single-phase transformerless photovoltaic inverters, in Proc. European Conference on Power Electronics and Applications, 2013, pp [50] N. Sintamarean, F. Blaabjerg, and H. Wang, Real field mission profile oriented design of a SiC-based PV-inverter application, in Proc. IEEE Energy Conversion Congress and Exposition, 2013, pp [51] M. Ciappa, Selected failure mechanisms of modern power modules, Journal of Microelectronics Reliability, vol. 42, no. 4-5, pp , Apr./May [52] L. Yang, P. A. Agyakwa, and C. M. Johnson, Physics-of-failure lifetime prediction models for wire bond interconnects in power electronic modules, IEEE Trans. Device and Materials Reliability, vol. 13, no. 1, pp. 9-17, Mar [53] E. Wolfgang, Examples for failures in power electronics systems, presentation at ECPE Tutorial on Reliability of Power Electronic Systems, Nuremberg, Germany, Apr [54] A. Wintrich, U. Nicolai, W. Tursky, and T. Reimann, Application manual power semiconductors, SEMIKRON International, 2011, ISBN: [55] J. W. McPherson, Reliability physics and engineering: time-to-failure modeling, Springer, ch. 8 and ch. 12, 2010, ISBN: [56] L. F. Coffin, A study of the effects of cyclic thermal stresses on a ductile metal, Trans. ASME, vol. 76, no.6, pp , [57] S. S. Manson, Behaviour of materials under conditions of thermal stress, NACA, Washington, DC, Report No. 1170, [58] C. Busca, R. Teodorescu, F. Blaabjerg, S. Munk-Nielsen, L. Helle, T. Abeyasekera, and P. Rodriguez, An overview of the reliability prediction related aspects of high power IGBTs in wind power applications, Journal of Microelectronics Reliability, vol. 51, no. 9-11, pp , [59] U. Scheuermann, Pragmatic bond wire model, Presentation at ECPE workshop on Lifetime Modeling and Simulations, Germany, Jul [60] R. Wu, F. Blaabjerg, H. Wang, M. Liserre, and F. Iannuzzo, Catastrophic failure and fault-tolerant design of IGBT power electronic converters An overview, in Proc. IEEE Industrial Electronics Society Annual Conference, 2013, Accepted. [61] H. Wang and F. Blaabjerg, Reliability of capacitors for DC-link applications - An overview, in Proc. IEEE Energy Conversion Congress and Exposition, 2013, pp [62] Cornell Dubilier, Application guide - aluminum electrolytic capacitors. [Online]. Available: [63] H. Wang, Y. Yang, and F. Blaabjerg, Reliability-oriented design and analysis of input capacitors in single-phase transformer-less PV inverters, in Proc. IEEE Applied Power Electronics Conference and Exposition, 2013, pp [64] P. O'Connor and A. Kleyner, Practical reliability engineering, the 5 th edition, John Wiley & Sons, ch.2, ch. 4 and ch. 7, 2012, ISBN: [65] A. Mettas, Design for reliability: overview of the process and applicable techniques, International Journal of Performability Engineering, vol. 6, no. 6, pp , Nov [66] V. Crk, Reliability engineering challenges in aerospace industry, presentation, [Online]. Available: /11/Reliability%20Engineering%20Challenges%20in%20Aerospace%2 0Industry.pdf [67] P. V. Varde, Role of statistical VIS-À-VIS physics-of-failure failure methods in reliability engineering, Journal of Reliability and Statistical Studies, vol. 2, no. 1, pp , Jun [68] P. L. Hall and J. E. Strutt, Probabilistic physics-of-failure models for component reliabilities using Monte Carlo simulation and Weibull analysis: a parametric study, Journal of Reliability Engineering and System Safety, vol. 80, no. 3, pp , Jun [69] H. Janssen, Monte-Carlo based uncertainty analysis: Sampling efficiency and sampling convergence, Journal of Reliability Engineering and System Safety, vol. 109, no. 1, pp , Jan [70] ZVEL, How to measure lifetime for robustness validation-step by step, Nov [71] N. M. Vichare and M. G. Pecht, Prognostics and health management of electronics, IEEE Trans. Components and Packaging Technologies, vol.

18 17 29, no. 1, pp , Mar [72] Y. Avenas, L. Dupont, and Z. Khatir, Temperature measurement of power semiconductor devices by thermo-sensitive electrical parameters- A review, IEEE Trans. Power Electron., vol. 27, no. 6, pp , Jun [73] E. Koutroulis and K. Kalaitzakis, Development of an integrated dataacquisition system for renewable energy sources systems monitoring, Renewable Energy, vol. 28, no. 1, pp , Jan [74] S. Yang, D. Xiang, A. Bryant, P. Mawby, L. Ran, and P. Tavner, Condition monitoring for device reliability in power electronic converters: a review, IEEE Trans. Power Electron., vol. 25, no. 11, pp , Nov [75] K. Ma, F. Blaabjerg, and M. Liserre, Thermal analysis of multilevel grid side converters for 10 MW wind turbines under low voltage ride through, IEEE Trans. Ind. Appl., vol. 49, no. 2, pp , Mar./Apr [76] D. Murdock, J. Torres, J. Connors, and R. Lorenz, Active thermal control of power electronic modules, IEEE Trans. Ind. Appl., vol. 42, no. 2, pp , Mar./Apr [77] E. R. Motto and J. F. Donlon, IGBT module with user accessible on-chip current and temperature sensors, in Proc. Applied Power Electronics Conference and Exposition, 2012, pp [78] M. Weckert and J. Roth-Stielow, Lifetime as a control variable in power electronic systems, in Proc. Emobility - Electrical Power Train, 2010, pp.1-6. [79] L. Wei, J. McGuire, and R. A. Lukaszewski, Analysis of PWM frequency control to improve the lifetime of PWM inverter, IEEE Trans. Ind. Appl., vol. 47, no. 2, pp , Mar./Apr [80] M. N. Marwali, J. W. Jung, and A. Keyhani, Control of distributed generation systems - Part II: Load sharing control, IEEE Trans. Power Electron., vol. 19, no. 6, pp , Nov [81] K. Ma, M. Liserre, and F. Blaabjerg, Reactive power influence on the thermal cycling of multi-mw wind power inverter, IEEE Trans. Ind. Appl., vol. 49, no. 2, pp , Mar./Apr [82] B. Lu and S. Sharma, A literature review of IGBT fault diagnostic and protection methods for power inverters, IEEE Trans. Ind. Appl., vol. 45, no.5, pp , Sep./Oct [83] B. A. Welchko, T. A. Lipo, T. M. Jahns, and S. E. Schulz, Fault tolerant three-phase AC motor drive topologies: a comparison of features, cost, and limitations, IEEE Trans. Power Electron., vol. 19, no. 4, pp , Jul [84] F. Richardeau, and L. Pham, Reliability calculation of multilevel converters - theory and applications, IEEE Trans. Ind. Electron., vol. 60, no. 10, pp , Oct [85] S. Karimi, P. Poure, and S. Saadate, Fast power switch failure detection for fault tolerant voltage source inverters using FPGA, IET Trans. Power Electron., vol. 2, no. 4, pp , Jul [86] K. Kriegel, A. Melkonyan, M. Galek, and J. Rackles, Power module with solid state circuit breakers for fault-tolerant applications, in Proc. Integrated Power Electronics Systems (CIPS), 2010, pp [87] E. Levi, Multiphase electric machines for variable-speed applications, IEEE Trans. Ind. Electron., vol. 55, no. 5, pp , May [88] R. Bojoi, M. G. Neacsu, and A. Tenconi, Analysis and survey of multi-phase power electronic converter topologies for the more electric aircraft applications, in Proc. International Symposium on Power Electronics, Electrical Drives, Automation and Motion, pp , [89] R. De Maglie, G. Osvald, J. Engstler, A. Engler, and J. P. Carayon, Optimized 70 kw power inverter dedicated to future aircraft applications, in Proc. European Conference on Power Electronics and Applications (EPE), 2009, pp [90] V. Choudhary, E. Ledezma, R. Ayyanar, and R. M. Button, Fault tolerant circuit topology and control method for input-series and output-parallel modular DC-DC converters, IEEE Trans. Power Electron., vol. 23, no. 1, pp , Jan Huai Wang (S 07 M 12) received the B.Eng. degree in Electrical and Electronic Engineering from Huazhong University of Science and Technology, Wuhan, China, in 2007, and the Ph.D. degree in Electronic Engineering from City University of Hong Kong, Kowloon, Hong Kong, in Since 2012, he has been with Aalborg University, Denmark, where he is currently an Assistant Professor in the Department of Energy Technology. He was a Visiting Scientist at Massachusetts Institute of Technology (MIT), USA, during September to November, His industry experience in power electronics includes 6 months work at the ABB Corporate Research Center, Baden, Switzerland, in His research interests include the reliability of DC-link capacitors, reliability of power electronic systems, high-voltage DC-DC power converters, time-domain control of converters, and passive components reduction technologies. On the above research topics, he has contributed over 40 journal and conference papers and filed 3 patents. Dr. Wang is the recipient of 5 paper awards and project awards from industry, IEEE and the Hong Kong Institution of Engineers (HKIE). He serves the guest Associated Editor of IEEE Transactions on Power Electronics Special Issue on Robust Design and Reliability in Power Electronics, and session chair of various conferences in power electronics. Marco Liserre (S 00-M 02-SM 07-F 13) received the MSc and PhD degree in Electrical Engineering from the Bari Polytechnic, respectively in 1998 and He has been Associate Professor at Bari Polytechnic and Professor in reliable power electronics at Aalborg University (Denmark). He is currently Full Professor and Chair of Power Electronics at Christian-Albrechts-University of Kiel (Germany). He has published 168 technical papers (44 of them in international peer-reviewed journals), 3 chapters of a book and a book (Grid Converters for Photovoltaic and Wind Power Systems, ISBN-10: IEEE-Wiley, also translated in Chinese). These works have received more than 6000 citations. He has been visiting Professor at Alcala de Henares University (Spain). He is member of IAS, PELS, PES and IES. He is Associate Editor of the IEEE Transactions on Industrial Electronics, IEEE Industrial Electronics Magazine, IEEE Transactions on Industrial Informatics, where he is currently Co-EIC, IEEE Transactions on power electronics and IEEE Journal of Emerging and Selected Topics in Power Electronics. He has been Founder and Editor-in-Chief of the IEEE Industrial Electronics Magazine, Founder and the Chairman of the Technical Committee on Renewable Energy Systems, Co-Chairman of the International Symposium on Industrial Electronics (ISIE 2010), IES Vice-President responsible of the publications. He has received the IES 2009 Early Career Award, the IES 2011 Anthony J. Hornfeck Service Award, the 2011 Industrial Electronics Magazine best paper award and the Third Prize paper award by the Industrial Power Converter

19 18 Committee at ECCE 2012, He is senior member of IES AdCom. He has been elevated to the IEEE fellow grade with the following citation for contributions to grid connection of renewable energy systems and industrial drives. Frede Blaabjerg (S 86 M 88 SM 97 F 03) was with ABB-Scandia, Randers, Denmark, from 1987 to From 1988 to 1992, he was a Ph.D. Student with Aalborg University, Aalborg, Denmark. He became an Assistant Professor in 1992, an Associate Professor in 1996, and a Full Professor of power electronics and drives in He has been a part time Research Leader with the Research Center Risoe in wind turbines. From 2006 to 2010, he was the Dean of the Faculty of Engineering, Science, and Medicine and became a Visiting Professor with Zhejiang University, Hangzhou, China, in His current research interests include power electronics and its applications such as in wind turbines, PV systems, reliability, harmonics and adjustable speed drives. He received the 1995 Angelos Award for his contribution in modulation technique and the Annual Teacher Prize at Aalborg University. In 1998, he received the Outstanding Young Power Electronics Engineer Award by the IEEE Power Electronics Society. He has received 15 IEEE Prize Paper Awards and another Prize Paper Award at PELINCEC Poland in He received the IEEE PELS Distinguished Service Award in 2009, the EPE-PEMC Council Award in 2010 and the IEEE William E. Newell Power Electronics Award He has received a number of major research awards in Denmark. He was an Editor-in-Chief of the IEEE TRANSACTIONS ON POWER ELECTRONICS from 2006 to He was a Distinguished Lecturer for the IEEE Power Electronics Society from 2005 to 2007 and for the IEEE Industry Applications Society from 2010 to He was a Chairman of EPE in 2007 and PEDG, Aalborg, in Peter de Place Rimmen is today Reliability Advisor at Danfoss Power Electronics A/S in Denmark since Peter has worked with practical approach implementing Reliability during the last 25 years in followed companies: Vestas Wind System R&D from 2004 to 2009, Grundfos Management R&D from 1997 to 2004 and Bang & Olufsen R&D form 1988 to Before that he had careers (14 years) at B&O as Constructor, Test engineer, Plant manager and Project manager. Peter had for some time participated in IEC dependability group. Peter has together with Nokia trained Nokia R&D and Vestas R&D people around the world in Design for Quality and Reliability. Today Peter is participating in CORPE Centre of Reliable Power Electronics at Aalborg University, participating in ZVEI facts sheets group for Robustness Validation, ECPE Course instructor, board member FAST (Danish Society for Applied Statistics) and initiated in 2001 and member of the Danish Six Sigma ERFA-group, subgroup of FAST. Peter holds 1½ patent for Vestas concern Lifetime improvement by thermal control improvements, and for Danfoss he hold 2 patents concern Dehumidifier for enclosures and one 1 patent for Monitoring device usage for stress in the field. John B. Jacobsen was born in Hørning, Denmark, in After practical education to electrician he received B.Sc. El.Eng from Aarhus Teknikum, Aarhus in Main work experience is from Grundfos, Denmark; 7 years in hybrid technology development, 7 years in hybrid production & more than 10 years as chief specialist in integration of power electronics into products, i.e. mechatronic disciplines including thermal management, interconnection, fixation, protection from environment (impact, humidity,...), reliability & cost. Optimization criterion being needed performance & reliability at lowest possible cost. Generalist in understanding value chain and all the disciplines that meet in the physical mechatronic reality, i.e. function trade-offs in design and produce ability in production. Thorkild Kvisgaard was born in Fabjerg Denmark in After the practical education as electrician he received a B.Sc.E.E from Aarhus Teknikum in In addition he received an e-mba in Innovation and Technology Management from Aalborg University in Main work experience is from Scanvaegt International where he in 10 years was Manager of Electronics Development creating solutions for the food industry as well as electronics for offshore applications. In 1994 he joined Grundfos and served as a Product Development Manager in 11 years. After this Thorkild Kvisgaard changed position in Grundfos to Global Technology Manager. In addition he became a Member of the Board in Center for Electrical Energy Systems (CEES) in 2006 and he act as Chairman of the Center for Intelligent and Efficient Power Electronics (IEPE). In the Center Of Reliable Power Electronics (CORPE) Thorkild Kvisgaard has the role as Vice Chairman. He had several patents granted and was nominated for the best Danish patent granted Jørn Landkildehus was born in Ebeltoft, Denmark, in He received the M.Sc.- E.E. from Aalborg University, Aalborg, Denmark, in Since 1995 he is with Danfoss Power Electronics engaged with development of variable speed drives and research in power topologies. He has been specializing in the design for EMC and in recent years been leading reliability engineering department at Danfoss Power Electronics. His technical interest include development processes, multi-disciplinary design techniques and Design for Reliability.