Authors:
Harry Beyel
1
;
Omar Makke
2
;
Fangbo Yuan
2
;
Oleg Gusikhin
2
and
Wil van der Aalst
1
Affiliations:
1
Chair of Process and Data Science, RWTH Aachen University, Aachen, Germany
;
2
Global Data Insight & Analytics, Ford, Dearborn, U.S.A.
Keyword(s):
Connected Car, Continuous Data, Sensor Data, Process Mining, Conformance Checking, Case Study.
Abstract:
Cyber-physical systems connected to the internet are generating unprecedented volumes of data. Understanding cyber-physical systems’ behavior using collected data is becoming increasingly important. Process-mining techniques consider sequences of events and thus can be used to check and verify how such cyber-physical systems operate. The data captured by cyber-physical systems are typically noisy and are not readily suitable for process mining. In this work, we present how a stream of connected-vehicle data can be transformed into an event log suitable for process mining. By applying different process-discovery techniques, we discover de-facto models that capture the behavior of an assistance system embedded in cars. We apply conformance-checking techniques and consult domain experts to find the best de-facto model. In addition, we apply conformance-checking methods to a preexisting, de-jure model that we transformed into a Petri net. We compare both models and point out differences.
In this process, we show how we overcome challenges and highlight why applying process-mining techniques in the cyber-physical systems domain is valuable.
(More)