Research On The Key Technologies Of Numerical Control Machine Servo System Fault Diagnosis Based On Heterogeneous Sensor Fusion

With the high level of machining precision, quality stability, and production efficiency, Numerical Control Machine is the cornerstone of modern manufacturing technique and equipment. Servo system is a key component of CNC machine tools and many complex numerical control equipments, which directly affects the accuracy, efficiency and reliability of the device. With the development of CNC machine tools towards high degree of structural complexity and automation of servo system, the internal parts are more closely associated with each other, and slight fault scan often result in a chain reaction, reducing the machining accuracy and reliability. Severe consequences include malfunction of performance, shortened service period and even the risk of being scrapped. The key of predictive maintenance philosophy is to guarantee "profits from maintenance and prediction", which is, based on monitoring the working conditions of machinery equipments, to predict the failure and its evolution, detect the causes of the failure, and take timely measures of maintenance. CNC machine tool servo system has the same characteristics and demands the same research methods with servo system of heavy equipments in national defense industry and aviation and aerospace industry. Therefore, it is necessary to carry out studies on the theory and methodology of failure diagnosis of CNC machine tool servo system.The dissertation focuses on three key technical issues for heterogeneous sensor fusion, that is, "what information to gather", "how to collect information", "how to use information". It carries out intensive study step by step on three existent questions about fault diagnosis of present CNC machine servo system, including a single parts of servo system is the object of many research, the lacking in excavation value about the built-in sensor information, the question of fault diagnosis of CNC machine servo system regarded as pattern recognition and fault classification.The dissertation starts with the mathematical model and stability analysis of modeling of servo system. Then it analyzes the typical fault mechanism of servo system, the mapping relations of the internal parameter representations theoretically, and conducts tests through simulation. It provides theoretical explanations for the built-in sensor machine which obtains ontological information of the machine, and proposes an approach to the research of heterogeneous sensor fusion by taking advantage of the complementary relationship of the internal and external sensors, and enriches the channels.The dissertation sets up a testing system for heterogeneous sensor fusion, studies the key data alignment technique of obtaining servo system information through the internal and external sensor system. It proposes a data alignment scheme with adaption of the present experimental conditions of Siemens 802 Dsl numerical control system and NI data collection system combined with the existing experimental basic condition. It broadens the channel and source of information, in contrast with the methods of collecting typical failure information only through external sensor or internal sensor.The dissertation discoveries error of the theory value of fault characteristic frequency in the experiment of detecting rolling bearing by external sensors, and studies the accumulated and transitive process between the fault characteristic frequency error and harmonics frequency doubling error systematically and comprehensively. The rolling bearing fault diagnosis method has its own uneliminatable fuzziness by the research of all kinds of technical means of improving error and enhancing frequency resolution. And it proposes a new method of rolling bearing fault diagnosis based on data-driven diagnostic method with no need for calculation. Then, the definition of intuitionistic fuzzy sets is advocated in the field of fault diagnosis and a new idea of intuitionistic fuzzy evidence acquisition under the framework of random sets and multivariate decision fusion is proposed, by further study of theoretical analysis and physical meaning of probability parameter uncertainty obtained by the new method of fuzzy evidence. The multi-source information fusion fault diagnosis studied by means of classification and pattern recognition can also be regarded as multivariate decision fusion by the result of experiments.The dissertation sets up hierarchical fault diagnosis model of CNC machine tool servo system based on intuitionistic fuzzy weighted fusion decision. First of all, it studies the multi-domain characteristic parameter extraction method by the combination of the time domain, frequency domain and wavelet packet de-noising with EMD decomposition, and the data dimension reduction based on the characteristics of the extreme spacing selection and correlation analysis. And then, it builds hierarchical fault recognition model of the multiple classifier based on the genetic BP network and RBF network and SVM and further analyzes the diagnostic ability of all the above-mentioned models. And it proposes CNC machine tool servo system intelligent hierarchical diagnosis model based on weighted aggregation operators of intuitionistic fuzzy decision fusion by regarding diagnostic accuracy of single classifier model as weight coefficient. The method proved by experiments has the advantage of high recognition ability and accuracy in samples of differences between different classifiers and embodies its own ability of fault tolerance and self correcting.