Vibration-based diagnosis of engine valve faults

Owing to the complex structure and the presence of a multiplicity of sources, the vibratory signals collected from diesel engines are always complicated in composition. The features extracted from one kind of faulty signal often overlap with those collected from other kinds of faulty signals. This phenomenon confuses the machine operators and usually misleads them into drawing a wrong conclusion. The use of effective features and the adoption of a powerful diagnostic strategy are two critical issues leading to successful diagnosis of machine faults. Both a specially designed Genetic Algorithm (GA) and the Kernel Principal Component Analysis (KPCA) technique were adopted in the paper, and an effective method has been achieved for diagnosing the faults occurring on the engine valves. Five kinds of valve faults are simulated on the sixth exhaust valve of a 6135-type diesel engine. The experiments fully prove that the proposed method is effective on valve fault diagnosis and has a great potential to be more widely accepted in industry.