Dissertation > Excellent graduate degree dissertation topics show

Sparse Coding Based Machine Condition Recognition and Its Application in the Condition Monitoring of a Heavy Roller Grinder

Author: LiuHaiNing
Tutor: LiuChengLiang
School: Shanghai Jiaotong University
Course: Mechanical and Electronic Engineering
Keywords: State - based equipment maintenance Condition Monitoring Sparse Coding Feature extraction blind signal separation The heavy numerical control roll grinder
CLC: TG580.6
Type: PhD thesis
Year: 2011
Downloads: 153
Quote: 0
Read: Download Dissertation

Abstract

Equipment during service, security, reliability and maintainability is an important issue that should be solved in the current China's equipment manufacturing industry. By monitoring the device status and condition-based maintenance is an effective means to solve the problem, the device status can be monitored equipment manufacturing enterprises to improve product design and manufacturing quality, further important aspect of the high-tech value-added enhanced products. For heavy-duty CNC roll grinder is representative of some of the basic manufacturing equipment, the more obvious necessity of condition monitoring, but also faces more challenges: the interference of the monitoring signal conditions under grinding conditions; low speed sampling the signal in the device status information for a long period of time under the conditions of continuous extraction; based on the characteristic device status identification inherent disadvantages. In order to solve the many problems of heavy-duty CNC roll grinder special characteristics of the mechanical structure and working conditions brought about their condition monitoring from theoretical methods, experimental validation, technology integration, in-depth research and exploration, including the following aspects: effective extraction problem for device status information, drawing on the principles of information processing of biological perception system redundancy compression, vibration signal analysis based on sparse coding method, with its computational characteristics of adaptive feature extraction framework. Feature extraction in the framework of the two-step process: First, learning from each type of device state sample signal a sub-dictionary, the further integration of multiple sub-dictionary redundant dictionary constructed device status; then based on the redundancy dictionary solving sparse monitoring signal, and extract a sparse feature. Construct a device state redundancy dictionary purpose is to a priori knowledge of the accumulation of different status of the device, so that the sparse features extracted quantify the status of a device presence indicators. Based on the verification of the data set of the standard bearing vibration, sparse features of the extracted good separability of the vibration signal after the load change, and the constructed bearing condition redundant dictionary remains better adaptability. Status recognition method based on the characteristics of the device, there inevitably mistaken diagnosis \In this paper, the inevitable false diagnosis \Unsupervised clustering self-organizing mapping neural network analysis capabilities with data visualization capabilities of the model by means of the device status based on sparse feature space in a two-dimensional plane, so as to construct a device status maps; equipment maintenance personnel to determine the status of the device according to the mapping of the results of the monitoring signals on the maps themselves, and to take a cautious approach to ambiguous mapping results. In addition, by means of self-organization mapping neural network anomaly detection based on monitoring data in the device status monitoring process to continue to improve the state the ability to identify, through the establishment of the model update mechanism makes. 3 for grinding conditions monitoring signal interference problems, research for blind signal separation technique based on sparse component analysis; proposed based on a fundamental principle of the vibration signal separation, based on single-channel vibration signal blind separation method of matching pursuit: weakening of the blind source separation problem is more practical and beneficial device status to identify cyclical \projection on overcomplete dictionary of basis functions, in order to achieve the separation of the three signal components. Experiments show that the method is not only able to effectively capture the the weak transient vibration signal early equipment failure and further separation of non-stationary signal components can activate the distribution of the clustering structure basis function. The overall architecture of the condition monitoring system is designed for heavy-duty CNC the roll grinder state monitoring requirements, and application of the above theoretical methods. First of all, from the angle of the help system internal device status information flow and information processing capabilities to improve overall system architecture combining site diagnostic and remote monitoring; acquisition grinder at no load sensing and data acquisition program implementation under the conditions of vibration data, and track vibration data collected under the grinding conditions; by data analysis to determine the three vibration type: under the conditions of the grinding gear meshing vibration of the top relative sliding and grinding chatter and establish based on a sparse state recognition model; in technology integration applications, by regulating system data and application interface to ensure the scalability of the system, and its diagnostic capabilities through the establishment of state recognition model update mechanism as the device status the richness of the monitoring data and improved. Heavy-duty CNC roll grinding machine condition monitoring prototype system is realized through the integrated application of the theoretical methods and technical specifications. The practical application of the established based sparse state recognition model to efficiently detect out of the case because the grinder head and tail frame top misalignment caused an increase in the top slide, for grinding chatter identification grinder operator makes the scene timely adjustment of the processing parameters, the quality assurance process.

Related Dissertations

CLC: > Industrial Technology > Metallurgy and Metal Craft > Metal cutting and machine tools > Grinding and grinding machine > General issues > Grinding process
© 2012 www.DissertationTopic.Net Mobile