| As an important power source and safety support equipment in coal mine production,mine air compressor plays a decisive role in the safe and efficient production of coal.The intelligent level of mine compressor restricts the reliability of mine ventilation system,and the monitoring and accurate and fast identification of mine compressor working status is the premise of its intelligent control.At present,the mine compressor mainly relies on manual inspection to monitor its working status,and relying on the experience of operators to diagnose faults,so it can not realize automatic monitoring and self-adaptive adjustment of working status.Therefore,it is necessary to study the key technology of working status identification of mine compressor in order to improve the intelligent level of compressor.In this paper,the working status recognition of mine compressor is taken as the research object.The infrared thermal imaging technology is used to obtain the working status images of key parts of compressor.The methods and technologies of online removal of complex background noise from original infrared images,rapid dimensionality reduction of ultra-high dimensional image features,and accurate recognition of working status of compressor based on machine learning are deeply studied.The main research contents are as follows:(1)Analyzing the basic structure of mine blower and combining with the actual production,the working principle of suction,sealing,compressing and exhausting of the screw type fan is studied in detail.Combined with the functional requirements of the state recognition system of the mine fan,the general framework of the state identification of the mine's blower is built,and its main components and identification process are analyzed.(2)An infrared image denoising algorithm based on the optimized wavelet threshold is designed to eliminate the noise and impact noise contained in the original infrared image.The improved thresholding algorithm is used to get the thresholding threshold of each order,and the dynamic step size distribution operator is introduced to enhance the global and local optimization ability for the standard fruit fly algorithm.(3)The low dimensional representation method of ultra high dimensional data is studied.A nonlinear dimensionality reduction method based on manifold learning is proposed to reduce the dimension of infrared image.The state recognition and evaluation system of mine blower is designed.The factors of blocking rate ? and blocking degree ?,the temperature deviation factor ? and the health-status evaluation factor H are introduced,and the working state of mine blower is divided in detail.(4)An algorithm for identifying the running state of mine compressor based on the optimized support vector machine is proposed,and the key parameters of the non-linear support vector machine are optimized by using the improved bat optimization algorithm.On the basis of the basic bat algorithm,elite population and exploratory population are introduced to enhance the global optimization ability of the bat algorithm to solve the premature problem of the bat optimization algorithm.The working condition recognition system of mine blower was designed and built.Experiments were carried out in the laboratory and in the Dragon King Zhuang coal industry.The experimental results show that the system can effectively process the infrared image of the mine compressor,accurately identify the working status of the compressor,and lay the foundation for further improving the intelligent level of the compressor.In this dissertation,there are 50 figures,18 tables and 146 references. |
PDF Full Text Request