Spectrum Of The Wear Pattern Recognition Method Based On Iron Studies

The ferrography is a wear diagnosis technology based on the analysis of wear particle. The practice has proved that the ferrography is the most effective method for wear condition monitoring and wear fault diagnosis. The recognition of the wear particle is the core step of the ferrography, the result of the recognition has direct relation to the correctness of the wear condition monitoring for the equipments. Because of the diversity and complicacy of the wear particle, the recognition procedure is carried out without guidance of mature theory. Currently, the recognition of the wear particle is carried out by the experts, the recognition process is very complicated, time-consumed and high cost. Those shortcomings limit the application of the ferrography. With the development of computer image processing and artificial intelligence, especially, with the development of neural network, it provides condition for the realization of automatic quantitative ferrographic analysis. The introduction of intelligence technologies will be helpful for the improvement of accuracy and automatization of ferrographic analysis. The recognition of wear particle is the core step to realize the automate ferrographic analysis. The thesis introduces a new method to ferrographic analysis based on the BP neural network. The main contents of the thesis are as follows:1. The monitoring technology for machines and equipments at home and abroad are evaluated synthetically. The significance aims of the project are presented.2. The thesis discusses and analyses the sorting of wear, the way the wear particle formed and the characteristics of the wear particles. The thesis proposes that the method of tribology system analysis should be used in the research on the wear particle, the characteristics of wear particle have relation to the condition and process of wear, the characteristics of wear particle include the condition characteristics and construction characteristics of the tribology system.3. The development of morphology characteristics extraction of wear particle is introduced. The advantages and disadvantages of some methods are pointed out. All those are available for further research of identification of ferrography wear particle in theory and practice.4. The principle of Artificial Neural Network (ANN) is discussed. The principle of BP neural network is emphasized. A new arithmetic adapt to the identification of wear particle is proposed based on the BP arithmetic, the possibility and validity of the improvement is examined by experiment.5. The BP neural network is applied to the recognition of wear particles, and a BP neural network sorting system expected to recognize severe wear particle, cutting wear particle, normal wear particle and fatigue wear particle is designed and trained.6. The function of the neural network's hidden layer is analyzed. A new way used to decide the neuron's number of the hidden layer is proposed based on the analysis and on the experiential way proposed by others. Construct a automate wear particlerecognition model.7. The Generalization Capability of the sorting system is tested. The thesis suggested that the standard for weighing the sorting system should be the generalization capability. Two ways to improve the generalization capability are proposed: (1) increasing the number of the training swatch, (2) increasing the hidden layer's neuron number of the neural network, and the second one is examined by experiment.