Pragmatic Fleet Condition Monitoring of Gearboxes

Size: px
Start display at page:

Download "Pragmatic Fleet Condition Monitoring of Gearboxes"

Transcription

1 Pragmatic Fleet Condition Monitoring of Gearboxes Gerrit van Middelkoop and Floris Manni DEKRA Rail 2017 Introduction Fleet condition is nowadays an essential insight to achieve optimal maintenance and lifespan for asset components. If the wear and failure risks are managed transparently and balanced against cost reduction, traditional maintenance concepts can be replaced by modern predictive variants. This presentation explains how actual fleet gear box condition of extensive asset fleets can be monitored in a pragmatic and cost effective manner. The most examples shown are rail applications, but the method is generally applicable and used for wind turbines, bridges or seawalls assets. Individual component condition modeling data acts as ammunition for data analysis and modeling of damage behavior that can predict current and future fleet condition.

2 This system only works well if all building blocks function pragmatically and effectively, and irrelevant data is filtered out as much as possible. We will focus on condition monitoring of lubricated systems in the rest of this presentation. Individual Components Monitoring Method: Analysis of oil, magnetic plug and oil filter samples of lubricated systems Gear boxes Compressors Engines Hydraulics Benefits: Prevent - unexpected break down - time table delays - unscheduled maintenance Limit - follow up costs In order to detect abnormal wear, effective diagnostic methods are required. Most abnormal wear processes generate significant "big" wear particles with alarm levels strongly dependent on application and component type. Therefore DEKRA Rail analyses metal content by the XRF method. Particle size distribution in samples is determined by a house method: the wear scan coefficient.

3 Principles of wear scan: wear particles are separated from oil, magnetic plug, oil filter or grease sample and investigated by microscope number of particles is evaluated en rated (0-9 classification) in 5 diameter classes Wear scan coefficient is calculated by sum of weighted class ratings Interpretation guidelines and rejection limits are component/machine dependent Black Holes Several asset types contain Black Holes for wear particles: Magnetic plugs Full flow filters (wind turbines!) Filters in engines Additional filtration for oil conditioning Assets with these black holes cannot be effectively monitored by oil analysis alone, but need additional diagnosis methods, e.g.: Magnetic plug analysis by swiping off the plugs and wear scan analysis in the lab Oil filter analysis by disassembling the filter in the lab, gathering the wear particles and wear scan analysis

4 Crucial: interpretation of machine condition Translation of laboratory analysis data into machine condition requires expertise: And of course a lot of additional information is relevant: Dependent on component type Specific properties sample location and component type/size Service life of lubricant and component Aging of lubricant Abnormalities and downtime Relubrication interval and history Run in after revision Seasonal effects Trend and recidivism Rejection limits After all this information has been processed correctly, a conclusion about the component's condition can be drawn and any necessary maintenance recommended for this individual component. Next step is high value fleet condition monitoring.

5 Fleet condition monitoring Method: Benefits: Centralization of condition information individual components per fleet Data analysis Construction defects Wear versus service life Predict remaining useful life Improved risk management Refurbishment planning Major cost reduction Extend revision terms Responsible control in case of anomalies The bathtub curve is commonly used as a theora\etical model for asset degradation risk: In order to make this model practical and manageable based on measurable parameters, some preconditions are essential: - The failure rate on the Y axis needs to be translated into a single measurable metric parameter that is representative for the asset risk - The time on the X axis needs to be translated to real service life or mileage to compensate down time or intensity of use - A reliable y axis reference framework that represents the normal situation in region B

6 The bathtub live After the essential data is gathered this can result in pictures like this: The X axis shows the mileage (x 1000 up to 4 million km) of the gearboxes, the Y axis shows the analyzed wear scan coefficients. The left graphics show individual data points, the right hand graphics show the mean (blue) and 25 and 75% population levels. So in fact the right hand graphics are bathtub curves based on analyzed magnetic plug samples. These pictures are presented on line as live dashboard information. The upper graphics prove that an initial scheduled revision period of 3 million km can safely be extended with another km and may be even further while controlling the failure risk based on condition monitoring of the individual gear boxes. The upper and lower graphics represent two very similar fleets of gearboxes, but the bath tubs show a slightly different behavior. Between 1 and 1.5 million kms both fleets showed increased wear as result of a minor construction problem. After refurbishment the wear levels decreased as a consequence. Live dashboards look like this:

7 Disaster control tool Fleet condition monitoring can also be a very effective disaster control tool. The accompanying example shows a fleet of gearboxes in four different positions. Each bar represents an individual gearbox. The two groups at the right side show a construction problem, that generates abnormal wear (red color). Based on this information repair of damage can be prioritized while the fleet stays in operation. Actual on line dashboards support this process effectively. Degradation Modeling Feed back and evaluation on in practice detected damage details of predicted failures is a vital quality loop to improve interpretation skills and develop degradation models. This example shows typical degradation patterns for different failure types. The Y axis shows wear scan coefficients. The X axis shows incubation period in days until the moment of predicted failure and advice on necessary repair.

8 Predicting the future Currently we are modeling hundreds of known and evaluated failure cases from the recent 10 years by means of modern data mining software. In the near future this will result in a predictive system by which measured laboratory results will be translated into quantified predictive failure risk of single components and fleets. Benefits and restrictions of fleet condition monitoring In Summary: Benefits in practice: - Improved risk management: known actual fleet wear condition - Major cost reduction: minimize spare parts and assets - Extend revision terms: e.g. gearboxes 3.5 -> 4 million km - Responsible control in case of disaster: support warranty claims prioritize worst wear cases Restrictions: - Applied diagnosis methods must be effective - Some wear processes develop rapidly: sample intervals must be balanced - Extend revision terms: phasing interval prolongation crucial Conclusion fleet condition monitoring Effective maintenance tool for asset managers Provides improved risk management High potential for maintenance cost reductions and extended revision periods Setup at acceptable cost possible by data analysis of individual component condition monitoring Gerrit van Middelkoop Specialist - Tribology DEKRA Rail bv PO Box 8125, 3503 RC Utrecht, The Netherlands Office address Concordiastraat 67, 3551 EM Utrecht gerrit.vanmiddelkoop@dekra.com