Study On Fault Diagnosis System Of I.C. Engine Based On Angular Vibration Of Crankshaft

As an important power machine, I.C. engine is usually applied in power generation, locomotive vehicle, engineering machine, automotive, marine power, and so on. Therefore monitoring I.C. engine's working condition, diagnosing its fault, ensuring it to work normally, and increasing its maintaining quality and efficiency is very necessary. Because the angular vibration of crankshaft contains a lot of information about the working condition of I.C. engine, many scholars are studying on I.C. engine's fault diagnosis through it, and they have made rapid progress. Neural Networks (NN) has many advantages, such as high parallel, association and generalization capability, pattern recognition ability and fault tolerant ability, so it has been widely applied to fault diagnosis of I.C. engine, and pushed its development largely.Therefore, a system of crankshaft torsional vibration measurement and fault diagnosis system (CTDS) is developed in this paper, based on the crankshaft torsional vibration measurement and analysis system (TVM), which adopts the strategy of Two Diagnosis Classes. Comparing with TVM, there are two main improvements in CTDS. First, CTDS has the ability of fault diagnosis, second, CTDS is portable.The portable character of CTDS system is realized by serial communication technique, component technique and multi-computer-communication technique, and it is more convenient and complete than TVM. The methods of TVRS method and Single Order Diagnosis method are used in The First Diagnosis Class by CTDS. In this paper, the author has programmed them, with simulation proceeded to verify them. BP NN is used in The Second Diagnosis Class, Which is also verified by experiment.Presently, most researches on working process fault diagnosis of I.C. engine, based on angular vibration, are about The First Diagnosis Class, such as cylinder misfire, while seldom of them cover The Second Diagnosis Class. So In this paper, it is mainly studied how to diagnose the fault in working process by means of torsional vibration of crankshaft combining with BP NN, Which belongs to The Second Diagnosis Class. A new method is put forward that different diagnosticparameters are picked up for different faults, or called "multi-pattern-rule", and two new principles are given on picking up different parameters for different faults. Two measures have been adopted to set up BP NN. First, a training function fitting for fault diagnosis is selected, and the proper number of hidden layer neuron is determined. Second, the diagnostic information is strengthened by a new method of samples subtraction, and the noise is weakened by lessening the dimensionality of sample space. It is demonstrated by experiment that the BP NN shows good performance and can pick out the category of fault from the sample assembly.Through the research in this paper, the author has enriched the fault diagnosis technique for I. C. engine, and drawn some valuable conclusions. The BP NN is provided for diagnosing the fault in working condition of I. C. engine, and the system of CTDS, developed in this paper, has the value of application.