Study On Self-propelled Gun Engines Wear Fault Diagnosis Based Ferrography Analysis

Wear is one of the main factors causing breakdown and fault of machine. The power of engine equipped on the self-propelled gun is large, its work environment is heavy, and its fault ratio is high. Ferrography technique for wear particle analysis can be an effective way for wear fault diagnosis of engine equipped on the self-propelled gun. The two difficult questions for ferrography technique popularized applications are wear particles image automatic processing and wear fault intelligent diagnosis, so it is important and difficult research contents for ferrography researcher.The research object is engine equipped on the self-propelled gun. Supported by the army scientific research project (310), a great deal of work encircling wear particles image numeric process and wear fault diagnosis were done in this paper, and some new ideas and processing methods were introduced.First, based on the structure and principle of engine, its main friction pairs were learned, and its wear pattern and wear regularity were studied, and wear particles generated during the wear process are classified by theory and experiment results. The theoretical base for farther study was settled.Secondly, based on the characteristics of engine wear particles, the principle and methods of wear particles image enhancement were learned, and the filter method for filter impulse yawp and gauss yawp in the wear particle image was put forward. The image can be in focus and processed conveniently. By research of image processing methods, put forward the auto-segmentation of ferrograph wear particle gray image based on fast two dimension entropy threshold algorithm, improved genetic maximum-variance algorithm and fuzzy genetic C-average clustering algorithm, and they optimize the segmentation algorithm, increase their calculating rate, and improve segmentation effect. According complex principle of wear particle color image, the auto-segmentation based on genetic maximum-variance algorithm and wavelet transform algorithm were applied to color wear particle image, and acquire better segmenting effect. Applying boundary detecting technique and boundary scouting technique, the boundary of wear particles was picked up. All wear particles in the image were marked, so it is convenient to distill their features and to identify them.Again, the extraction techniques of wear particle configuration features were studied. A characteristic system for describing wear particles including size, shape, color and texture features is established utilizing the computer image analysis technique, and their forty features were extracted. Because different sensitivity of variety wears to get the best description parameter and different point of view, their feature parameter was redundancy possibly. So, these particle characteristics were optimized with the help of the fuzzy rough sets and a simple、efficient particle description parameter system was obtained.Then, according to the fuzzy mathematics, nerve-net theory and grey theory, put forward to the wear particles identification based on fuzzy-NN, fuzzy-grey theory and fuzzy-grey clustering, improving wear particle identification precision.Finally, the wear particles analysis results were applied to engine wear fault diagnosis. Two engine wear style diagnosis models based on fuzzy-NN and fuzzy-gray associate theory were introduced; and the engine wear fault cause and fault position diagnosis model combined ant colony algorithm and SVM was advanced. According to the simulation results and application to experiment, the proposed method is proved effective and efficient for engine’s fault diagnosis.At the end of this paper, the further research work needed to be done in ferrography field is produced.