Study Of Multi-stage Processing Error Separation And Diagnose Based On Error Spectrum

The main quality problem of mechanical products lie that complex parts manufacturing levels are low and bad quality. The products do not have a stable quality and causing quality fluctuating. The reasons for bad quality is that lack of clearly comprehension of multistage variation propagation. And also the variation sources and variation structure are not well understood. In view of this, this article proposed a variation deducing method based on variation spectrum. This article has a series of content as following:(1) Modeling for variation sources. The parts variations, including dimensional variations and surface topography variations, are formed by many variation sources. This article created the models for machine thermal variation, fixture variation and spindle rotation variation by analyzing their characteristics. The models provided references for further variation deduction and also gave a guideline for other variation sources modeling.(2) Multistage variation propagation study based on multistage variation spectrum. This section introduced a method of stream of variation based on state space for multistage variation propagation. Then a new method called multistage variation spectrum was proposed to describe the multistage variation propagation. Multistage variation spectrum made the multistage variation propagation visualized, However, the state space method is very difficult for visualization.(3) Surface topography variation separation and deduction based on surface variation spectrum. Surface topography variation is constituted by many variation components; these components are called surface variation spectrum in this article. This article applied three distinctive methods to decompose the surface topography variation into different surface variation spectrums and the analyzed each spectrum to find the variation sources. 1) In terms of the independence of each variation sources, the Independent Components Analysis(ICA) was proposed to decompose the surface topography into spectrums which caused by their corresponding variation sources. 2) In terms of the variation spectrums with distinctive scales, a hybrid method combining Wavelet Transform and Empirical Mode Decomposition was applied to Surface topography variation, a genuine variation spectrums were obtained as a result. 3) In consideration of the non-continuous surface, a triangular interpolation method was proposed to decomposed the discontinuous surface into system error and random error respectively. Then the subsurface of system error is further decomposed by Bi-dimensional Empirical Mode Decomposition(BEMD) to get the variation spectrums varying in scales: surface roughness, waviness and profile. By using all the proposed, realizing the variation deduction from stages to inter-stage, forming a systematical and complete variation deduction theory.