| The ball screw pair, as one of the key driving parts of the CNC machine, whose life is important to the CNC machine processing performance. It is always the focus of concern and research for the machine tool industry, exports and scholars at home and abroad, to ensure CNC machine operating safely and reliably, decrease the downtime caused by ball screw pair defaults as far as possible, predict and prolong the maintenance cycle of the CNC machine.Sponsored by the National Science and Technology Major Project of CNC Machine Faults Predict and Diagnosis (Project Number:2009ZX04014-102) and Domestic High-grade CNC Machine Demonstration Application in Typical Plane Structure Processing (2010ZX-04015-011), the Y direction ball screw pair (THK type) of the Changzheng 718 CNC machine's feed driving systems is taken as reseach object in this thesis. The Changzheng 718 CNC machine is one of machine tools used in a mechanical manufacturing enterprise. The factors that influencing life of the ball screw pair, have been analyzed from the CNC machine structure characteristics and dynamics, and the conclusion is that the key factors affecting life of ball screw pair are its working load and running time. It has discovered that the ball screw pair's performance degradation, which mainly reflected in the changes of driving torque and vibration while it runs, is the key to determine its life. Current sensors on the drive motor of ball screw pair and vibration sensors on ball screw pair, therefore, are used to monitor the ball screw pair's state, warn its defaults, track and predict its life by means of the hardware and software systems for signals acquisition, processing, operation and compaeison.The signals collected are filtered by the Empirical Mode Decomposition algorithm, then the eigenvalues of the signals time and frequency domain are extracted out, and the Principle Component Analysis method is adopted to acquire the time-frequency domain edgenvalues which are sensitive to the vibarated state changes of the ball screw pair. The eigenvalues form eigenvalue matrix in the time sequence at last.By cutting tests, the cutting force signals and drive motor current signals from the maichine tool's 20 kinds of commonly processing conditions are measured and extracted, and the mapping relationship between them is analysed. The Radial Basic Function network is used to construct the cutting force prediction model, which is used to prcdict cutting force only through current signals. At the same time, the current signals are used to track the ball screw pair's running time.After considering the parameters of the machine tool itself, processing conditions, running time, and some other factors, the load on ball screw pair is analyzed when machine tool is operating, and the expected residual life which consist of expected residual life vector in time series of the ball screw pair, is calculated.Dynamic Fuzzy Neural Network (DFNN) is used to fit the mapping relationship between vibrated state eigenvalues and expected residual life, and the performance degradation model, which is used to research the performance degradation law and predict the ball screw pair residual lifetime, is established through the DFNN.Finally, the life prediction method is complied into Matlab program, and called by the main program of CNC machine fault warning and performance evaluation system in a certain time intervals. It's realized to predict the residual life and assess the health condition live and online for the ball screw pair. It can provide CNC machine users with evidence for production or maintenance plan. |
PDF Full Text Request