Determining the fault status of a component and its readiness, with a distributed automotive application

https://doi.org/10.1016/j.engappai.2008.10.004Get rights and content

Abstract

In systems using only single-component tests, the fault status of a component is ready if a test only supervising the component has been evaluated. However, if plausibility tests that supervise multiple components are used, then a component can be ready before all tests supervising the component have been evaluated. Based on test results, this paper contributes with conditions on when a component is ready. The conditions on readiness are given for both centralized and distributed systems and are here applied to the distributed diagnostic system in an automotive vehicle.

Introduction

Fault diagnosis is important in many applications and has become even more so with the increase of on-board computing power in technical processes, see for example Isermann (2005). This work is motivated by the diagnostic systems, i.e. the system responsible for fault diagnosis and supervision, used in automotive vehicles (Gertler, 1998, Hristu-Varsakelis and Levine, 2005) and in particular that used in a Scania heavy-duty vehicle. These systems typically store a diagnostic trouble code (DTC) when a component is found to be faulty (SAE, 2003, ISO, 1999). In the first generations of diagnostic systems, each diagnostic test checked exactly one component for abnormal behavior. Therefore, the DTCs could be used to exactly state the fault status of a component, which is faulty if a test supervising the component has responded and normal otherwise.

Due to higher demands on fault diagnosis, such as reduced emission levels (European Union, 2005), more components must today be supervised for abnormal behavior compared to the first generations of diagnostic systems. Since the number of sensors in the system is limited, the industry has been forced to introduce diagnostic tests that check the correct behavior of several components at the same time, often denoted plausibility tests. The plausibility tests are for example based on analytical redundancy relations (ARR) (Gertler, 1998, Isermann, 2001, Nyberg, 1999). See Cordier et al. (2004) for a thorough discussion on how analytic redundancy relations relate to consistency based diagnosis. The plausibility tests come into conflict with the framework based on single-component tests that has previously been used in automotive vehicles. With the addition of plausibility tests, a component can now be only suspected to behave abnormally and its fault status is therefore suspected.

In addition to the fault status, if fault diagnosis is performed using precompiled diagnostic tests then there is an advantage if the repair technician or the control system can get an indication on when the evaluation of additional diagnostic tests cannot change the fault status of a component, and the fault status is in this case said to be ready. The evaluation of all tests gives readiness for all components, however, due to for example limited processing power and active diagnostic tests, it is in automotive applications not always possible to evaluate all tests. Therefore, the cost in evaluating a test has to be weighted against the improvement in fault diagnosis. If for example the evaluation of additional diagnostic tests will facilitate the repair by improving the fault statuses, it is an advantage to evaluate these tests before the repair is started. In automotive applications, readiness codes are used to state if all tests supervising a component or a subsystem have been evaluated or not (ISO, 1999). The difference between readiness in ISO (1999) and this paper is that a component might here be ready even though not all diagnostic tests have been evaluated.

This paper contributes with an algorithm, denoted the FSR (fault status and readiness) algorithm, which computes the fault status and readiness for all components. The FSR algorithm is here applied to the diagnostic system used in a heavy-duty vehicle from Scania.

The notation of fault status and readiness is defined with respect to a centralized system. However, many applications, and especially those used in the automotive industry, include multiple electronic control units (ECUs), generally denoted agents, which includes diagnostic systems (Leen and Heffernan, 2002, Navet et al., 2005, Hristu-Varsakelis and Levine, 2005, Biteus, 2007). In these distributed systems, diagnostic tests might exist in one agent that supervise components that belong to another agent. Therefore, an additional contribution of this paper is that the notations of fault status and readiness are extended to distributed systems. This extension render it possible use the fault status computed locally, to state the fault status applicable for the complete distributed system. The application on the heavy-duty vehicle is extended to the distributed case.

In the AI field (Reiter, 1987, Dressler and Struss, 1996), the dominant methodology for fault diagnosis has been so-called consistency based diagnosis, on which this paper is based. The methodology of consistency based diagnosis has strong relationships with the methods for fault diagnosis used in engineering disciplines (Cordier et al., 2004), such as control theory and statistical decision making (Gertler, 1998, Gerler et al., 1995, Basseville and Nikiforov, 1993). Within this methodology, a diagnosis points at a set of components whose abnormal behavior could explain why a system does not function as intended and are primarily used for repair and fault tolerant control (Shin and Belcastro, 2006). The fault status differs from diagnoses in that the fault statuses give the information whether a component behaves abnormal or not component-wise. That is, for each component, the fault status tells if the behavior is: abnormal for certain, might be abnormal, or not abnormal for certain. Further, it is in some applications intractable to compute the diagnoses since the complexity increases exponentially with the number of tests, this is in contrast to the fault status that has only a linearly increasing complexity and is therefore in practice always tractable.

Section snippets

Background to consistency based diagnosis

In this section the framework in consistency based diagnosis will be introduced. This framework will be used in the rest of this paper.

A system consists of a set of components C that should be supervised by the diagnostic system. A component is something that can be diagnosed, such as sensors, actuators, cables, and pipes. Here, only the abnormal and the not abnormal mode is considered, where the abnormal mode does not have a model. Further, the set notation used in for example GDE is employed (

Fault status and readiness

This section will focus on systems with one agent, while Section 4 will extend the results to distributed systems consisting of multiple agents.

Distributed systems

The fault status and the readiness are here extended to distributed systems such that it is possible to use the fault status and readiness computed locally in each agent to state the fault status and readiness for the complete system. First, distributed systems will be exemplified and a framework for distributed systems will be designed. The fault status and its readiness will then be extended to distributed systems.

Computing the fault status and its readiness

The propositions in Section 3 can be used to design an algorithm that computes the fault status and the readiness for all components. A direct design of the algorithm loops, for each component, over all diagnostic tests and checks if any of the conditions in the propositions are fulfilled. Even though such a direct design will give the desired results, the algorithm require quite some processing power since the algorithm must loop over all tests for all components. This section will therefore

Automotive vehicle application

The FSR algorithm has been evaluated on a part of the diagnostic system used in the heavy-duty truck from Scania that is described in Section 4.1. The part consists of the selective catalytic reduction (SCR) system, which will be studied in detail, and the engine management system (EMS), which will be included to create a distributed system. The SCR lowers nitrous oxides from diesel engines and is used in new trucks from all European heavy-duty vehicle manufacturers, see Fig. 3 for a schematic

Conclusions

Motivated by applications used in automotive vehicles, the fault status of a component is defined as faulty, suspected, or normal. Also defined is the readiness of the fault status that states if the evaluation of additional diagnostic tests could change the fault status or not. An important aspect of the readiness defined in this paper is that a component could be ready even though all tests that supervise the component have not been evaluated. The relations between fault status, readiness and

References (21)

There are more references available in the full text version of this article.

Cited by (3)

  • A new strategy for automotive off-board diagnosis based on a meta-heuristic engine

    2011, Engineering Applications of Artificial Intelligence
    Citation Excerpt :

    The CBR feasibility for decision help in industrial supervision was shown by Mille et al. (1999) and a practical study by Rasovska et al. (2008) for maintenance and diagnosis of industrial equipment shows a precision of 95% in problem solving. A statistical analysis of the sales of BMW (Biteus et al., 2009) reveals that less than 1% cars are identical, that implies a real challenge in the editing and updating of the vehicle model (knowledge representation and acquisition) in automotive diagnosis tools. Hence the advantage of ontology techniques for the creation of case representation shows several advantages in such cases as reported in a study by Bergmann and Schaaf (2003).

  • FlexDx: A reconfigurable diagnosis framework

    2010, Engineering Applications of Artificial Intelligence
    Citation Excerpt :

    This is done by dynamically ranking the possible measurement points according to the expected information obtained from the measurement. A related approach is found in Biteus et al. (2009) and Biteus (2007) where tests are ranked and selected according to some information criterion. The approach adopted here for test selection is fundamentally different and the objective is to find a subset of tests, out of a predefined set of possible tests, such that running these tests provides all information available in the model and observations given the current set of diagnoses.

  • A survey on diagnostic methods for automotive engines

    2012, International Journal of Engine Research
View full text