The Methodology of Detecting Dynamical Properties of Electrical Discharge Machining (EDM) Process towards Building a Timely Varied Linear Predictive Model
The electrical discharge machining (EDM) process recorded by a time series of gap states, which are scalars statistically quantified from discriminated discharging pulses, is partitioned into transient, efficient and deleterious subprocesses. Linear analysis reveals different machining powers in the three subsequent processes. Nonlinear analysis are mainly focused on the efficient subprocess because of its weak stationarity. The low correlation dimension of the process demonstrates the structural form of the time series, nonlinear analysis by surrogate method proves its deterministic nonlinearity. The results of correlation dimension and Lyapunov exponent demonstrate its chaotic behavior with a low structural dimension, indicating the predictability of gap states. Then a nonlinear deterministic model in a delayed embedding space is made and applied in detecting the signature of a latent change from efficient process to deleterious process in advance. The deterministic nonlinearity, correlation dimension, nonlinear prediction, and Lyapunov exponent of the process educes the building of a timely varied linear model, approximating the variations of gap states. Experimental verifications show that this predictive model can quickly and accurately provide one-step predictions with a high precision.
EDM Nonlinear analysis State space reconstruction Nonlinear prediction Timely varied linear model
Ming Zhou Fuzhu Han Boyan Tang
Department of Mechanical Engineering, Beijing University of Civil Engineering and Architecture, Beij Department of Precision Instruments & Mechanology, Tsinghua University, Beijing 100084, China
国际会议
The 16th International Symposium on Electromachining(第16届国际电加工会议)
上海
英文
73-78
2010-04-19（万方平台首次上网日期，不代表论文的发表时间）