Available online at ScienceDirect. Procedia CIRP 61 (2017 ) The 24th CIRP Conference on Life Cycle Engineering

Transcription

1 Available online at ScienceDirect Procedia CIRP 61 (2017 ) The 24th CIRP Conference on Life Cycle Engineering Internet-of-Things Enabled Real-Time Monitoring of Energy Efficiency on Manufacturing Shop Floors Yee Shee Tan a, *, Yen Ting Ng a, Jonathan Sze Choong Low a a Singapore Institute of Manufacturing Technology, 2 Fusionopolis Way, Singapore * Corresponding author. Tel.: ; fax: address: tanys@simtech.a-star.edu.sg Abstract Energy efficiency (EE) has become an important indicator in manufacturing industry due the rising concerns about climate change and tightening of environmental regulations. However, most manufacturing companies today are only able to monitor aggregated energy consumption and lack the real-time visibility of EE on the shop floors. The ability to access energy information and effectively analyze such real-time data to extract key indicators is a crucial factor for successful energy management. Therefore, in this paper, we introduce an internetof-things (IoT) enabled software application for real-time monitoring of EE on manufacturing shop floors. While enabling real-time monitoring of EE, it also applies data envelopment analysis (DEA) technique to detect abnormal energy consumption patterns and quantify energy efficiency gaps. Through a case study of a microfluidic device manufacturing line, we demonstrate how the application can assist energy managers in embedding best energy management practices in their day-to-day operations and improve EE by eliminating possible energy wastages on manufacturing shop floors The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license 2017 The Authors. Published by Elsevier B.V. ( Peer-review under responsibility of the scientific committee of the 24th CIRP Conference on Life Cycle Engineering. Peer-review under responsibility of the scientific committee of the 24th CIRP Conference on Life Cycle Engineering Keywords: Energy Efficiency; Real-Time Monitoring; Data Envelopment Analysis; Best Practices; Internet-of-Things 1. Introduction wadays, efficient usage of energy is becoming more of priority due to the rising concerns about climate change and regulatory requirements. Among all end-user sectors that are targeted for achieving energy efficiency (EE) improvement, manufacturing industry is deemed one of the high potentials as it is the largest consumers among all end-user sectors. According to Singapore s energy statistics, the energy consumption by industry sector was 42.6% of total energy consumed, followed by the commercial and household sectors, which are 36.5% and 14.9%, respectively in 2015 [1]. To improve EE, the role of energy management has greatly expanded in manufacturing industry as it has also resulted in reducing operating costs in long term. As defined by ISO 50001:2011, energy management is a comprehensive and systematic approach for energy conservation efforts in an industry. It is judicious and effective use of energy to maximize profits and to enhance competitive positions through industrial measures and optimization of EE in the process. For energy management, the first step is usually the energy monitoring. Industry can only manage their energy when they initiate to measure and understand their current energy performance. However, industry today is only able to monitor aggregated energy consumption, but unable to visualize real-time EE at shop floor. The ability to access energy information and effectively analyze such real-time data to extract key indicators is a crucial factor for successful energy management. The new emerging technology, Internet-of-Things (IoT), which connects physical objects using electronic sensors and internet is drawing attention nowadays. IoT technology promotes the heightened level of awareness about the world and a platform from which to monitor changing conditions and react to those changes [2]. IoT is expanding to many other interesting application domains while energy is one of the The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ( Peer-review under responsibility of the scientific committee of the 24th CIRP Conference on Life Cycle Engineering doi: /j.procir

2 Yee Shee Tan et al. / Procedia CIRP 61 ( 2017 ) application areas where IoT technology plays a major role [3]. Energy management is integrated with IoT technology to provide the ideal solution for monitoring real-time energy consumption while providing the level of awareness of energy performance [4][5]. With the support of IoT technology, i.e. energy sensor, energy consumption data can be collected in real-time at different levels, such as machine, production line or facility level [6]. However, collection of these data without the production or operating data will not be sufficient to understand EE [7][8]. Thus, this paper aims to bridge the gap by introducing an approach that uses both the energy and production data for EE assessment. Coupled with the IoT technology, the approach is able to provide real-time EE monitoring in manufacturing shop floors. Besides, data envelopment analysis (DEA) technique is applied to identify abnormal energy consumption patterns and quantify energy efficiency gaps. Through a case study of a microfluidic device manufacturing line, we demonstrate how the application can assist energy managers in embedding best energy management practices in their day-today operations and improve EE by eliminating possible energy wastages on manufacturing shop floors. 2. Internet-of-Things enabled software application for real-time energy efficiency monitoring 2.1. Concept Overview The proposed IoT enabled software application helps energy managers in achieving better EE by understanding both the energy consumption patterns and production data while eliminating possible energy wastages in the manufacturing operation. It works in a simple manner as shown in Fig. 1. The application users, e.g. energy manager, monitor the energy performance for each machine at shop floor using tablet while data such as power consumption and process operating parameters, e.g. temperature, pressure, etc. are captured via sensor or controller. Production data are provided by existing software such as manufacturing execution system, work order tracking system, etc. The server stores the data as well as the energy performance status and respective analysis results in the common repository. The three components, i.e. data acquisition, server and energy manager, interact with each other via wireless network. Shop Floor Real-time EE monitoring Data Capturing via Sensors/Meters Energy consumption Waste generation Data Capturing via Controller Material input rate Throughput Power consumption Process condition Software Architecture Fig. 1. Concept overview of Internet-of-Things enabled software application DAQ node Standard Data Product type Throughput Common Repository LAN Server 2.2. Software application architecture A software application is developed and structured as a multi-layered application consisting of user experience, business logic and data layers [9]. As Fig. 2 illustrates, it consists of (1) Presentation layer This layer contains the components that implement and display the user interface and manage user interaction. A set of user interface components such as the dashboard, notification pages and reports are design to provide a way for users to interact with the application. User interface can make use of controllers to communicate with the back-end and to navigate or process the interface components. Mobile Application Presentation Layer Business Logic Layer Data Layer UI Components - Energy efficiency dashboard - Alert notification Monitoring Engine Data Access Components User Benchmarking Engine Service Agents Database Cross-Cutting Security Configuration Communication/ Connectivity Fig. 2. Software application structure for real-time EE monitoring (2) Business logic layer This is the layer where all the engines in the application reside. It contains all the processing logic to make the application possible. The application consists of two parts, i.e. a) monitoring algorithm and b) benchmarking engine. a) Monitoring algorithm Monitoring is a process for metering the energy consumption and collecting real-time energy data. Data collected from monitoring is served as a basis to understand current level of energy use. With the basis, we can identify patterns and obtain information that can further be used to implement corrective and preventive action. However, as mentioned, in order to improve EE, integrating energy and production data is important. The algorithm flow chart to show how we correlate energy and production data is presented as Fig. 3. Inputs considered are 1) power consumption data (P) captured via sensor or controller and 2) production data (PV) tracked by production monitoring software, such as work order tracking system. The interval data (i) comes in increments of 1-minute granularity. With the acquired data, the energy consumption and production volume involved with respect to different reporting period r,

3 378 Yee Shee Tan et al. / Procedia CIRP 61 ( 2017 ) i.e hourly, daily, weekly and monthly are evaluated. By correlating the evaluated energy consumption (E r ) to the key drivers such as the production volume (PV r ) at respective period, we can then identify the machine status as no operation, idling, or operating. While if the machine is operating, specific energy consumption, i.e. energy consumption per production volume, will then be calculated to understand how effectively energy is used to produce given amount of production or to deliver certain work by the machine. considered as historical best. By comparing each DMU with the historical best practice, the relative efficiency is evaluated by dividing the respective E r of the historical best and evaluated DMUs. For instance, DMU P 2 which is one of the historical best is the efficiency reference set for DMU P 6 as well as P 2 is performing better than P 6 as it is consuming less energy E r while producing the same output, PV r. Thus relative efficiency for P 2 is evaluated as P 6 (E r )/ P 2 (E r ) while relative efficiency for P 6 will be 1 as it is the historical best. Start P 4 Sampling interval, i Power consumption at time t, P t Production volume at time t, PV t E r P 8 P 6 P 7 Calculate total energy consumption for different reporting period r Calculate total production volume for different reporting period r P 5 P 2 P 3 P s T t where s: start of reporting period T: end of reporting period r: {hourly, daily, weekly, monthly} where s: start of reporting period T: end of reporting period r: {hourly, daily, weekly, monthly} P 1 PV r Show Operation for entity status Show Idling for entity status End Fig. 3. Flow chart for monitoring algorithm b) Benchmarking engine Benchmarking is the next process taken into consideration after the understanding of current energy performance via monitoring. Benchmarking is a process of searching for those practices which lead to the excellent performance. This help to establish baseline, and hence highlight the problem area as well as the potential for improvement in comparing with the best practices. DEA, which is a very powerful benchmarking technique is selected and applied. It is a non-parametric method for evaluating relative efficiency of decision making units (DMU) based on multiple inputs and outputs. An input oriented Banker, Charnes and Cooper (BCC) DEA models that assumes variable returns to scale while minimizing inputs (i.e. energy consumption) and keeping outputs (i.e. production volume) at current levels are considered [10]. DMUs defined in this study have the granularity of time and spatiality, for instance DMU is the hourly energy (E r ) and production (PV r ) performances. As a result from linear programming, which used to construct a non-parametric piece-wise surface over the data, by minimizing inputs E r and keeping outputs PV r at current levels (equation in Fig. 5), those DMUs that lie on the envelope (e.g. P 1, P 2, P 3 and P 4 in Fig. 5) are identified and Calculate specific energy consumption for different reporting period r Fig. 4. Illustrative example for DEA The obtained relative efficiency is then clustered into three categories using quantile classification. Three categories are selected for showing the alerts based on the red, amber and green colours used in a traffic light rating system. Top 25% of DMUs ( >Q 3 ) are considered as good as historical best, showing the machines with normal energy performance. While the lowest 25% ( <Q 1 ) will be alerted for further investigation as the machines are performing in abnormal status, this showing a potential gap for EE improvement. Those DMUs that lie in between first quartile (Q 1 ) and third quartile (Q 3 ), the machines are considered in warning status, which need to be aware of. The details flow chart of the benchmarking engine is illustrated in Fig. 5. (3) Data layer This layer contains the functionality for creating, transforming, updating and deleting items into the database. The data access components abstract the logic necessary to access the underlying data stores. (4) Cross-cutting functionality Besides the major three layers, the cross-cutting functionality (i.e. security, configuration and communication) are required to support the application. Security consists primarily of data protection. Configuration considers how to handle device resets while device communication includes wireless communication and wired communication with a host computer.

4 Yee Shee Tan et al. / Procedia CIRP 61 ( 2017 ) Start device. The device is then forwarded to functional test and packaged at the end of the process. Total energy consumption for different reporting period r Total production volume for different reporting period r Data Envelopment Analysis where : Relative efficiency Data Classification using Quantile Frequency Relative Efficiency Fig. 6. An example of microfluidic device Relative Efficiency Show ABNORMAL as machine performance Show WARNING as machine performance Show NORMAL as machine performance End Fig. 5. Algorithm flow chart for benchmarking engine 3. Case study A case study based on the in-house learning factory on microfluidic device manufacturing line is conducted to illustrate the key features of real-time EE monitoring application. The device is made of microfluidic chip, electronic part (i.e. printed circuit board (PCB)) and casing. The PCB and casing are outsourcing to third party maker while the microfluidic chip is manufactured in house. The microfluidic chip is made of a set of micro-channel moulded into a material such as polycarbonate. Fluids are pumped in the device in order to achieve the desired features (mix, biochemical control). The fluids used in microfluidic devices could be blood samples, bacterial cell, protein solution or even essential oil. It is commonly used in clinical diagnostics. A sample of microfluidic device is shown in Fig. 6, which it consists of cover, base and membranes for filtration purpose. Fluids for sampling will be pumped through the inlet and trapped in the device, and then the analysed sample is connected to outside by output pierced through the device. The manufacturing line of microfluidic device is presented in Fig 7. The production line starts with moulding of cover and base using injection moulding machine, followed by cutting of membrane into the desired shape using laser trimming machine. The fabricated ship surface shall be flat and the dimension of the micro-channels shall be accurate before they are welded together with membrane using ultrasonic welding machine. The chip is then going through bonding inspection, repeated printing of various materials and curing, then coating to protect the surface. After the firmware installs on PCB, microfluidic chip will be assembled into a Fig. 7. Manufacturing line of microfluidic device The key features of the software application are demonstrated through the case study of a microfluidic device manufacturing line as following: (1) Real-time EE monitoring dashboard The application provides the real-time EE monitoring dashboard at the shop floor level. All the processes involved in the microfluidic device manufacturing line are monitored and the energy performances are presented in the main screen of the application (Fig. 8). As a result of real-time benchmarking using DEA and considering hourly energy performance as DMUs, the processes involved are shown in different colours to indicate different EE levels. Grey indicates no operation, while red, amber and green indicate operations that are abnormal, warning and normal respectively. The system will update the EE statuses of all the processes every hour. This is where the energy manager understand the current level of energy use and detect the abnormal energy consumption patterns at a glance. Those processes with warning and abnormal EE occurrences will be captured and alerted. For example, plastic injection moulding, ultrasonic welding and oven are alerted in this case.

5 380 Yee Shee Tan et al. / Procedia CIRP 61 ( 2017 ) Fig. 8. Real-time EE monitoring dashboard (2) Real-time energy performance monitoring By looking at the details of any process, the real-time energy performance monitoring is illustrated as Fig. 9. The performance indicators monitored include total energy consumption, total production and energy cost involved with respect to different periods, i.e. hourly, daily, weekly and monthly. Fig. 9 showing the hourly energy performance monitored for the plastic injection moulding machine (PIM-1). With the selected time frame, i.e. 15:00 16:00, the total energy consumption is kwh to produce 27 pieces of product (14 covers and 13 bases). The total energy cost, i.e SGD/hour is being spent for value added production (55%) and non-value added production such as idling (45%). Besides, specific energy consumption is provided to understand how effectively energy is consumed. Hourly EE performances throughout the day were monitored and the abnormal EE patterns were detected to provide the level of awareness. For example, energy performances from midnight 12 am until morning 9 am were alerted as the machine was idling without any production. This can bring to the next level of investigation. to understand the improvement potentials. Besides, the factors affecting the abnormal EE occurrences are also presented and the alert message can be notified to respective person for action taken. Fig. 10 details all the essential information regarding the alert investigation using PIM-1 as an example. The hourly energy consumption from 9 to 10 am on 8 Sep 2016 is kwh for producing 27 covers. It can be known that energy reduction of kwh (33.66 %) can be achieved potentially by referring to its frontiers, which are 1) consuming less energy (9.08 kwh) while producing around the same amount of products (20 covers) and 2) consuming around the same quantity of energy (9.08 kwh) while producing more products (40 covers). With these, we can further identify the operating conditions applied in these two days. For instance, the factor that affects the abnormal occurrence is the hourly idle time. The idle time is longer and the production volume is lower. Eventually, this alert message will be sent out automatically by the system to respective person but it can be sent out to external party as well by the energy manager, if necessary. With the understanding of improvement potentials, the energy manager is able to embed best practice energy management in day-to-day operations and achieve better EE by eliminating possible energy wastages in the manufacturing operations. Fig. 10. Real-time alert investigation Fig. 9. Real-time energy performance monitoring (3) Real-time alert investigation For further investigation, it is able to show how bad the current energy performance is as compared to the historical best practice and to provide the efficiency reference set for achieving the historical best. With the information, it is able (4) Report generating In 2013, Singapore National Environment Agency introduced the mandatory energy management requirement under Energy Conservation Act [11]. This requires energy intensive companies that annually consume more than 15 GWh of electricity to submit a periodic reporting of energy use. The information required in the annual report includes 1) days with abnormal EE occurrences together with the reasons affecting the anomaly, 2) breakdown of total energy consumption of each machine, 3) breakdown of total energy cost of each machine and 4) specific energy consumption for different products. With the usage of this software application, the report can be easily generated by just a single click as shown in Fig 11.

6 Yee Shee Tan et al. / Procedia CIRP 61 ( 2017 ) References 4. Conclusion Fig. 11. Report generating An approach and methods for an IoT enabled real-time monitoring of EE on manufacturing shop floors are presented in this study. The application is tested using the microfluidic device manufacturing line. Based on the study, the application has resulted in valuable benefits. Mainly included are enable real-time monitoring while capture the abnormal EE occurrences and quantify the improvement potentials through real-time benchmarking and assist energy managers to embed best practice energy management in day-to-day operations and achieve better EE by eliminating possible energy wastages in the manufacturing operations. [1] The Energy Market Authority, Singapore Energy Statistics, [2] The Big Book of MoT TM - Making the Most of the Monetization of Things, Aria Systems, Inc., [3] Haller, S., S. Karnouskos, and C. Schroth, The Internet of Things in an Enterprise Context, in First Future Internet Symposium, Vienna, Austria, [4] Bhardwaj, A., Leveraging the Internet of Things and Analytics for Smart Energy Management, TATA Consultancy Services, [5] S. Karnouskos, et. al., Towards the Energy Efficient Future Factory, Procedia 7th IEEE International Conference on Industrial Informatics 2009, Jun , 2009, pp [6] Tao, F., et. al., IoT-based Intelligent Perception and Access of Manufacturing Resource Towards Cloud Manufacturing, IEEE Transactions on Industrial Informatics, [7] Shrouf, F., et. al., Smart Factories in Industry 4.0: A Review of the Concept and of Energy Management Approached in Production Based on the Internet of Things Paradigm, In Proceedings of 2014 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM), 2014, pp [8] F. Tao, et. al., Internet of Things and BOM Based Life Cycle Assessment of Energy-saving and Emission Reduction of Products, IEEE Trans. Ind. Informat., 2014, vol. 10, no. 2, pp [9] Meier, J.D., et al., Mobile Application Architecture Guide, Microsoft, [10] Y.S. Tan, T.B. Tjandra, B. Song, Energy Efficiency Benchmarking for Mass and High-Mix Low-Volume Productions, Procedia CIRP 29, 2015, p [11] Mandatory Energy Management Practice, Singapore National Environment Agency, Retrieved from Further cases involving more variations in product types and operating parameters are being studied and necessary enhancement to the application will be made.