Part Dimension Deviation Control In Batch Manufacturing Process Based On Monitoring Error Sources

A machining system of complicated product is usually a serial-parallel multi-variation-source multi-stage manufacturing system, in which the final product variation is an accumulation or stack-up of variation from all machining stages. Product dimensional variation is one of most important factors that affect product quality, productivity and response time to market directly. Those dimensional variations will be introduced, increased, decreased, stack-up, propagated, and become the final product variation when one workpiece goes through the whole machining system. A lot of research work about dimensional variation analysis, diagnosis, and control in a single stage has been done by domestic and international researchers. However, these work does not fully characterize dimensional variation accumulation process when a workpiece goes through multiple stages in a systemic level, and does not provide the integration of propagate analysis of dimensional variation, predict, diagnosis, control.The multi-stage machining process is studied in this research, which provides some general technologies and methodologies for predicting and reduction of product quality variation in a machining process so that the machining system can be well controlled.This research work can be summarized as follows:1) Development of methodology for dimensional variation analysis and modeling in a multi-stage machining system. Through mapping key Characteristics variations into changeable state variables when a workpiece goes through the whole machining process, this work studies how to map the relationship of variations from all stages into a mathematical model, and describe form and propagation of variations, dynamical changes, as well as analyze propagation process of stream of variations in a machining process.2) Development of a EWMA control system for part dimension features. This work studies the stastical control of part dimension features in batch production of parts, and also presents a study of quantized evaluation system for partial diagnosable manufacturing system, as well as provides some measures that can improve the diagnosable capability of variation sources for partial diagnosable manufacturing system. This work also studies the effective diagnosis methods for variations sources detection in a multi-variation-source multi-stage machining system.3) Design of multivariate stastical control system. To detect part dimension errors in batch production, spatial relationship between part dimensional features and error sources was established based on analysis of part dimension error and its propagation model. Multivariate Exponentially Weighted Moving Average (MEWMA) was adopted to control error sources in batch production of parts,and method to control part dimension error in batch production through statistical control of error resources was proposed. Efficiency and reliability of this model were verified by simulation analysis.4) In order to achieve sufficient as well as effective measuring sensing data, optimal sensor distribution system should be carried out so that maximal error source detection capability is enabled. Therefore, spatial relationship between three types of error sources and measuring sensing data in multistation manufacturing system is established based on state space model, and a variance-detecting sensitivity index is proposed for characterizing the detection ability of process variance components and the optimization problem for sensor distribution is formulated for a multistation manufacturing system. Then a data-mining-guided evolutionary method is devised to solve this optimization problem. An illustrative example is performed to validate the importance and effectiveness of the proposed analytical procedure, which lays foundation for the establishment of sensor distribution design theories, methods and tools.In this paper, spatial relationship between error sources and part dimension deviation has been presented based on part dimension error propagation model which shows that part dimension deviation can be estimated by measuring error sources. Therefore, in batch manufacturing, part dimension deviation can be statically controlled by directly control error sources or by indirectly control part dimension features calculated from error source data. The proposed method can be used for variation control in nondestructive testing for products when part dimension features cannot be measured directly. Also, since the error sources are directly controlled, the proposed method provides basis to apply variation prediction, design evaluation, tolerance analysis and synthesis, and faults diagnosis of machining process.