Search

Under the load conditions of clamping force and riveting force, the part deformation problems are induced in the assembly process. For the component deformation problems of the thin-wall part assembly, in this paper, from the viewpoint of micro-aspect, the fractal theory model is proposed to analyze the relationship between the part deformation, loads, and contact area for rivet-joint parts; the micro-deformation values for different stages are solved and obtained through theoretical analysis. According to the self-similarity of fractal theory model, the change of the micro-convex body for contact surface will be researched to reflect the macro-deformation properties. Through the measuring experiment for thin-wall parts, the actual three-dimensional coordinate value of each measuring point is obtained before and after riveting assembly, and then, the wavelet neural network method (WNNM) is adopted to predict experiment deformation value based on the obtained deformation value after processing the coordinate value through fitting surface method. Hereby, according to the deformation value for every thin-wall part after processing experiment data, the fractal dimension value will be determined based on the fractal statistic method to analyze the deformation status of different stages. Through analysis and comparison, the results show that the variation regularity of deformation value based on two or three fractal deformation stages is well conformed to and similar to the experimental value and prediction value, which also verify the correctness and effectiveness of the theoretical model, and provide a theoretical basis for further research on thin-wall part assembly deformation from a micro-perspective and improving assembly accuracy quality in future. [ABSTRACT FROM AUTHOR]

This paper proposes a novel modeling methodology for machine tool thermal error. This method combines the advantages of both grey model and artificial neural network (ANN) in terms of data processing. To enhance the robustness and the prediction accuracy, two kinds of grey neural network, namely serial grey neural network (SGNN) and parallel grey neural network (PGNN), are proposed to predict the thermal error. Experiments on the axial directional spindle deformation on a five-axis machining center are conducted to build and validate the proposed models. The results show that both SGNN and PGNN perform better than the traditional grey model and ANN in terms of prediction accuracy and robustness. So the new models are more suitable for complex working conditions in industrial applications. [ABSTRACT FROM AUTHOR]

Thermal deformation is one of the most significant causes of machining errors in machine tools. One effective method is to build a compensation system to offset the thermal errors. Therefore, an accurate model is the key part of the compensation system. This study proposed a modified Elman network (EN) to improve the prediction accuracy of the compensation model in machine tools. And the improved EN can be regarded as a feed-forward neural network with feedback from hidden layer and output layer as an additional set of inputs. The structure of this network determines its dynamic characteristic with memory function. On the other hand, thermal deformation of the spindle contributes the largest part of total thermal errors in precision machining. Then a precise finite element model of machine tool spindle was established. And a new method for calculating the heat transfer convection coefficient on the surface of the spindle was proposed in this paper. The improved EN was used to map the nonlinear relationship between temperature field and thermal errors of the spindle. At last, a verification experiment was implemented on a CNC center and some satisfying results were achieved. [ABSTRACT FROM AUTHOR]