Simultaneous-Fault Diagnosis of Automotive Engine Ignition Systems Using Prior Domain Knowledge and Relevance Vector Machine

Abstract

Engine ignition patterns can be analyzed to identify the engine fault according to both the specific prior domain knowledge and the shape features of the patterns. One of the challenges in ignition system diagnosis is that more than one fault may appear at a time. This kind of problem refers to simultaneous-fault diagnosis. Another challenge is the acquisition of a large amount of costly simultaneous-fault ignition patterns for constructing the diagnostic system because the number of the training patterns depends on the combination of different single faults. The above problems could be resolved by the proposed framework combining feature extraction, probabilistic classification, and decision threshold optimization. With the proposed framework, the features of the single faults in a simultaneous-fault pattern are extracted and then detected using a new probabilistic classifier, namely, pairwise coupling relevance vector machine, which is trained with single-fault patterns only. Therefore, the training dataset of simultaneous-fault patterns is not necessary. Experimental results show that the proposed framework performs well for both single-fault and simultaneous-fault diagnoses and is superior to the existing approach.

Sparse Bayesian extreme learning committee machine for engine simultaneous fault diagnosis

“…Each time, 1 second of raw signal x k , including λ, ignition and sound signals, is simultaneously recorded with a sampling rate of 8092 Hz. Single-Faults d 1 Retarded ignition timing d 2 High resistance in secondary circuit d 3 Partially broken spark-plug cable d 4 Defective spark plug d 5 Narrow spark plug gap d 6 Misfire due to extremely lean mixture d 7 Carbon fouled in spark plug d 8 Engine knock d 9 Rich mixture d 10 Wide spark plug gap d 11 Clogged air filter element d 12 Defective fuel injector in one or more cylinder d 13 Defective spark plug or ignition circuit in one or more cylinders d 14 Leakage in air manifold d 15 Lean mixture Table 3. In other words, one case of each signal type has 8092 sampling data points.…”

“…Simultaneous-faults si 1 High resistance in secondary circuit & misfire due to extremely lean mixture si 2 Narrow spark plug gap & carbon fouled in spark plug si 3 Partially broken spark-plug cable & wide spark plug gap si 4 High resistance in secondary circuit & narrow spark plug gap & rich mixture si 5 Partially …”

“…In the engine ignition signal, firing voltage, burn time and average spark voltage are employed as domain knowledge (DK) features [1,2]. In the engine ignition signal, firing voltage, burn time and average spark voltage are employed as domain knowledge (DK) features [1,2].…”

“…One of the main challenges in the fault diagnosis of automotive engines is that, engine fault diagnosis is a simultaneous-fault problem, meaning the several independent single-faults may appear at the same time [1,2]. Due to the complex structure and running conditions of the engine, vehicle failure is usually caused by the occurrence of engine fault.…”

“…Another difficulty is a typical multi-signal fusion problem; it involves the use of multi-signal, such as vibration, sound, ignition and air ratio, to simultaneously detect and identify the engine faults [3][4][5]. The data-driven methods generally rely on signal-based analysis techniques [6], which are very suitable for analyzing the system whose domain information is insufficient or unknown such as automotive engine [2,11]. Model-based diagnostic methods rely on analytical models, while an exact engine analytical model is difficult to be determined because the engine is a complicated thermo-fluid and mechanical system [8,9].…”