SDM4Service banner
 
 

Data Mining for Service and Maintenance

We invite you to attend the SDM 13 Workshop on Data Mining for Service and Maintenance, to be held in conjunction with SDM 13, on May 4, 2013 in Austin, Texas.

Background

The maintenance and servicing of system components, machines, or instruments are a crucial part of business for many industries for systems ranging from photocopiers to power plants and from trucks and locomotives to airplanes. It is a common industry standard to provide system service as part of the contractual agreement, and the quality of service is increasingly becoming the key differentiator between competing products. Appropriate and timely maintenance can increase the safety and reliability of the system, detect failing components to reduce collateral damage to other healthy subsystems, and ultimately increase revenue and improve customer satisfaction.

Instead of standard practices such as fixed maintenance schedules (by usage time) or performing reactive diagnostics maintenance (i.e., fixing the problem after it happens), increasingly industries are moving toward a so-called proactive service mode (which is often called condition monitoring or preventive maintenance, among other names). This is also enabled by the sensor data collected by most of the systems mentioned above. Hence, these systems can be monitored and serviced more efficiently by employing intelligent service data analytics solutions that lead to a proactive (act before the problem has occurred) timely maintenance schedule. Several such simple solutions have been already deployed but, based on the increased amounts of service data generated by all these systems, better and improved solutions are possible. The service data includes sensor data, system error messages, event log files, service & maintenance data, and other types data on machine/system status. The key questions here are (1) how to analyze this data effectively and efficiently in order to provide necessary service to the systems, while minimizing service cost and (2) how to employ this data for online decision support. Most of the data is time-series or event sequence data but these systems and processes also generate unstructured or semi-structured data such as text. Therefore the data analytics processes are challenging, as they require dealing with noise in data, unaligned multivariate data, missing and compressed data, and a multitude of data types. The complexity and dynamics of the systems that generate the data (e.g., aircraft engines) also lead to system-to-system variations and a large number of failure modes (i.e., thousands of categories of classes).

Because of its unique characteristics mentioned above, service data analytics is a domain in which data mining and knowledge discovery techniques just started to be applied and it requires the attention of both researchers and practitioners. Novel innovative methods in time series analysis and prediction, relational learning, multi-view learning, transfer learning or incomplete data analysis can have a profound effect.

We believe that this new domain can benefit from increased attention from our SIAM data mining community. There are various other communities that deal with systems health management such as the Prognostics and Health Management society and the DX Workshop community, but they tend to deal more with individual components or systems, and therefore have more of an engineering perspective that is reflected in their emphasis on model-based or physics-based approaches, even though they are increasingly investigating more data-driven approaches.

Objectives

The purpose of this multi-disciplinary workshop is to bring together researchers and industry experts in the fields of data mining, machine learning, text analysis and signal processing who share an interest in problems and applications of system service and maintenance. We believe that this is an important application domain for the data mining community. There has been significant interest displayed recently in this area, as evidenced by the previous workshop and the recent text on "Machine Learning and Knowledge Discovery for Engineering Systems Health Management". For this workshop, the goal is be to discuss recent progress, to frame and further clarify the relevant research questions, and to bridge the gap between data mining research and industry needs on certain concrete problems. We will also encourage submissions describing open research problems and challenges in this area, which will help the community to identify future directions.

The workshop will provide a platform for exchange of ideas, identification of important and challenging applications, and discovery of possible synergies. The difference between service and maintenance will be highlighted in multiple industries. It is our hope that this will spur vigorous discussions and encourage collaboration between the various disciplines, potentially resulting in collaborative projects. We will particularly emphasize the mathematical and engineering aspects of service data analysis.

We will address many of these topics through both invited and contributed talks and a combination of position papers (describing research ideas or new challenges) and full papers (describing more mature research and practical results). The workshop program will consist of presentations by invited speakers of both industry and academia, and by the authors of the abstracts and papers submitted to the workshop. In addition, there will be a slot for a panel discussion to identify important problems, applications, and synergies between the two disciplines. Topics of interest include but are not limited to:

  • Time series and data stream classification for prognostics
  • Feature extraction from time series, event data, and text
  • Semi-automated trend analysis
  • Prognostics modeling
  • Survival analysis
  • Regression and ranking from time series
  • De-noising and handling missing data in service data
  • Rule based systems for service
  • Combining multiple data sources for prediction
  • Classification with imperfect class labels (e.g., noisy service notifications)
  • Performance measures for preventive maintenance
  • Knowledge representation for service
  • Risk-sensitive data mining based decision support for prognostics tasks
  • Empirical preventive maintenance data sets and comparison
  • Testing and validation of data mining algorithms for service
  • Cost-sensitive data mining for service and maintenance
  • Other service applications
Workshop length: One Day - on May 4, 2013.

Related Event: The 2013 Annual Conference of the Prognostics and Health Management Society.