Fault Diagnosis and Prognosis in a Network of Embedded Systems in Automotive Vehicles

Abstract:

The relentless competition among automotive companies and increasing demands from customers for driver assistance functions and dynamically-controlled safety systems in vehicles are creating mounting time-to-market pressures and, consequently, shortened development times. With the increased vehicle complexity and shortened development times, guaranteeing hardware-software integrity and, hence, vehicle performance has become a salient issue. This is because poor vehicle performance increases warranty costs to automotive manufacturers and maintenance costs to customers; these, in turn, result in customer dissatisfaction and in reduced competitiveness of US automotive industry. GM has identified the development of methods to enhance the reliability of embedded systems (i.e., hardware, software and interfaces) designed and implemented by different suppliers, and vehicle diagnostics and prognostics as key research challenges. The objectives of this collaborative project between the University of Connecticut (UConn) and the GM Research and Development R&D) are two-fold: (i) Develop an integrated diagnostic and prognostic (D&P) modeling framework for designing high-integrity heterogeneous network of embedded systems so that (hardware, software and interface) faults can be detected and isolated rapidly, and (ii) Investigate on- line and off-line inference methods for diagnosing faults and for predicting the remaining useful life of (hardware) components based on the inferred failing components and the various tracked paths of degradation. The latter methods can be embedded in the vehicle or accessed over a network. To achieve the objectives, our high-level approach is to identify the potential failure modes associated with software, hardware and HW/SW interfaces, develop functional dependency models between the failure modes of a network of embedded systems and the concomitant monitoring mechanisms (called “tests”), perform testability analysis, generate  model-based failure modes,  effects and criticality analysis (FMECA), and validate  the  diagnostic and prognostic inference methods via fault injection prior to deployment in the field. The first objective is addressed by formulating an integrated D&P process for fault diagnosis and prognosis of embedded systems that combines data-driven techniques, graph-based system-level functional dependency models and mathematical/physical models (for components with well-understood dynamics). The integrated D&P process is applied to two automotive systems, namely, regenerative braking system (RBS) and electric power generation and storage system (EPGS). In the first application: the RBS model consists of a driver model, component (physical system) model, and six electronic control units (viz., battery control, engine control, motor control, brake control, and powertrain controller) communicating through a controller area network (CAN) bus. A variety of physical system faults (parametric and sensor-related faults), software logic faults, and network communication faults (babbling idiot, missing message, too many message, burst loss and outdated message faults) were modeled. Subsequently, a data-driven approach is applied for fault detection and diagnosis and the results demonstrated that the faults can be isolated with a reasonably good accuracy (>90% correct fault isolation). In the second application, a hierarchical diagnosis strategy is developed to diagnose faults in an EPGS system with a plant model (drive belt, alternator and battery), and two ECU’s: an engine control module (ECM) and a body control module (BCM). Here, the faults considered include actuator faults, software within the controllers (ECUs), as well as interactions between hardware and software, i.e., between controllers and plants. A factorial hidden Markov model based inference algorithm was used to infer the root causes based on the observed test outcomes (>97% correct fault isolation). The second objective is addressed by designing inference algorithms suitable for coupled systems with delays. We have made four novel contributions. The first contribution addressed the problem of diagnosing coupled faults (coupled fault diagnosis) in complex cyber-physical systems, such as automotive systems where failure of one component (either hardware or software) may probabilistically trigger the malfunctioning of another component. The formulation involves a mixed memory Markov model within a factorial hidden Markov modeling formalism to represent coupled fault dependencies. The second  contribution formulated and solved the dynamic set covering (DSC) problem, a series of set-covering problems that are coupled over time, using Lagrangian relaxation and a Viterbi decoder-based algorithm. The objective of the DSC problem is to infer the most probable time sequence of a parsimonious set of faults that explains the observed test outcomes over time. The novel feature of the DSC problem is that the test outcomes may be observed with time delays induced by network transmission and sensor processing delays. The third contribution is a delay dynamic coupled fault diagnosis algorithm (DDCFD) to deal with the problem of coupled fault diagnosis with fault propagation/transmission delays and observation delays with imperfect test outcomes, a realistic representation of practical networked embedded systems. We proposed two methods to solve the problem: Partial- sampling method and a method based on block coordinate ascent and the Viterbi algorithm. Finally, we have formalized a unified data-driven prognostic framework that combines failure time data, static parameter data and dynamic (time-series) data. The framework employs Cox proportional hazards model and soft dynamic multiple fault diagnosis algorithm to infer the degraded state trajectories (complementary survival functions) of components and to estimate their remaining useful life.This research enhances the competitiveness of American automotive industry by (i) minimizing life cycle cost of vehicle systems, (ii) enhancing safety and reliability of vehicular systems and (iii) improving customer satisfaction through enhanced vehicle availability. It has utility far beyond the immediate automotive application area being pursued here. Representative applications include aerospace systems, electrification of transportation, medical equipment, smart buildings/smart grid and communication networks, to name a few.

Book Chapter

1. A. Patterson-Hine, G. Aaseng, G. Biswas, S. Narasimhan and K.R. Pattipati, “Diagnostics and Testability, Chapter 16 in Stephen B. Johnson (Ed.), System Health Management with Aerospace Applications, Wiley, 2011, pp. 265-277.

Journal Publications

1. C. Sankavaram, K. Pattipati, Y. Zhang, M. Howell, and M. Salman, “Fault Diagnosis in Hybrid Electric Vehicle Regenerative Braking System using Data-driven Techniques”, IEEE Transactions on Industrial Electronics, 2013 (Under Review).
2. A. Kodali, K. Pattipati, and S. Singh, "A Coupled Factorial Hidden Markov Model (CFHMM) for Diagnosing Coupled Faults", IEEE Transactions on Systems, Man, and Cybernetics: Systems, Vol. 43, No. 3, pp. 522-534, May 2013.
3. A. Kodali, S. Singh, and K. Pattipati, "Dynamic set-covering for real-time multiple fault diagnosis with delayed test outcomes", IEEE Transactions on Systems, Man, and Cybernetics: Systems, Vol. 43, No. 3, pp. 547-562, May 2013.
4. A. Kodali, Y. Zhang, C. Sankavaram, K. Pattipati, and M. Salman, "Fault Diagnosis in Cyber-physical Systems: Application to Automotive Electric Power Generation and Storage System (EPGS)", IEEE/ASME Trans. on Mechatronics, Vol. PP, No. 99, pp. 1-10, September 2012 (early access).
5. S. Zhang, K.R. Pattipati, Z. Hu, X. Wen, “Optimal Selection of Imperfect Tests for Fault Detection and Isolation,” IEEE Transactions on Systems, Man and Cybernetics, part A – Systems and Humans, Vol. PP, No. 99, pp. 1-15, March 2013 (early access).
6. B. Pattipati, C. Sankavaram and K.R. Pattipati, “System Identification and Estimation Framework for Pivotal Automotive Battery Management System Characteristics,” IEEE Trans. on SMC: Part C, Vol. 41, No. 6, November 2011, pp. 869-884.
7. S. Zhang, K. Pattipati, Z. Hu,  X. Wen,  and  C. Sankavaram,  "Dynamic  Coupled  Fault  Diagnosis  with  Propagation and Observation Delays", IEEE Transactions on Systems, Man and Cybernetics: Part A, Vol. PP, No. 99, pp. 1-15, March 2013 (early access).
8. B. Pattipati, C. Sankavaram, K.R. Pattipati, Y. Zhang, M. Howell and M. Salman, “Multiple Model Moving Horizon Estimation Approach to Prognostics in Coupled Systems,” IEEE Aerospace and Electronic Systems Magazine, Vol. 28, No. 3, pp. 4-12, March 2013.

Conference Publications

1. C. Sankavaram, A. Kodali, and K. Pattipati, “An Integrated Health Management Process for Automotive Cyber-Physical Systems”, IEEE ICNC 2013 International Workshop on Cyber-Physical Systems (CPS), San Diego, CA, January 2013.
2. B. Pattipati, B. Balasingam, C. Sankavaram, K.R. Pattipati and Y. Bar-Shalom, “An EM Approach for Dynamic Battery Management Systems,” 15th International Conference on Information Fusion, Singapore, July 9-12, 2012.
3. C. Sankavaram, B. Pattipati, K. Pattipati, Y. Zhang, M. Howell, and M. Salman, "Data-driven Fault Diagnosis in a Hybrid Electric Vehicle Regenerative Braking System", in Proceedings of IEEE Aerospace Conference, Bigsky, Montana, March 2012
4. C. Sankavaram, A. Kodali, K. Pattipati, B. Wang, M. Azam, and S. Singh, "A Prognostic Framework for Health Management of Coupled Systems", IEEE International Conference on Prognostics and Health Management, Denver, CO, June 2011.
5. S. Deb, S. Ghoshal, M. Azam, V. Malepati and K.R. Pattipati, “Cost of Not Having a Sensor,” Proceedings of the Infotech@Aerospace 2011,St. Louis, Missouri, March 2011.
6. B. Pattipati, C. Sankavaram, K. Pattipati, Y. Zhang, M. Howell, and M. Salman, "Multiple Model Moving Horizon Estimation Approach to Prognostics in Coupled Systems", IEEE Autotestcon, Baltimore, MD, September 2011 (winner of Walter E. Petersen Best Technical Paper Award).
7. A. Kodali, K. Pattipati and S. Singh, “A Coupled Factorial Hidden Markov Model (CFHMM) for Diagnosing Coupled Faults”, IEEE Aerospace Conference, Big Sky, Montana, March 2010.