Research On Quality Control Intelligent Method Of Complexity Production Process

Quality, as one of the essential factors of the socio-economic development, has been got more and more people’s attention. With the development of science and technology, the quality management of increasingly complex production processes has become a major subject of the researchers. Quality management implements all activities of the products or services of quality planning, quality control, quality assurance and quality improvement, which are based on the formulation of quality policy. Process quality control, as a key part of the quality control, is the main stage of the product creation and an effective way to achieve quality management to ensure process quality. With increasingly complex production process, the process has the characteristics of high-dimensional and non-linear data, the uncertainty process model, sub-process interfere with each other and strong coupling, its quality control is more difficult than the traditional production process. In this paper, the complex system theory combined with the diversified, intelligent quality control methods are used to realize the quality control of the complexity production process.Support vector machine (SVM) is an intelligent method of machine learning based on VC-dimensional theory of statistical learning theory and structural risk minimization. SVM uses the kernel function to make the nonlinear problem transform to linear problem, which makes it better to solve the small samples, nonlinearity, high dimension and local minima problems, and has greater generalization ability. Particle swarm optimization (PSO) algorithm simulates birds’predatory behavior to search the optimal solution of complex space by the collaboration and competition between individuals, which has the characteristics of simple concept, few control parameters, easy to implement, evolutionary computation and swarm intelligence optimization. PSO is an effective optimization tool to solve integer nonlinear optimization, nonlinear continuous optimization problems and combinatorial optimization problems. In this paper, support vector machine as a tool to create complex production process quality model, the particle swarm algorithm and its improved algorithm are adopted to optimize the parameters of support vector machine and select model optimal solution, so that implement the quality control of complex production process.The fault detection of complex production process is a significant problem of quality control. In this paper, wavelet packet transform is chosen as the denoising tools of the process sample data, which uses a multi-hierarchical division method to eliminate the noise and interference of sample data to get the approximation of the original signal, so that extracts the effective data samples. And then, multivariate statistical process monitoring model, which use kernel principal component analysis to make the nonlinear problem transform to linear problem, is proposed to realize the fault detection of complex process. A numerical example and a case study of the Tennessee Eastman (TE) process are used to verify the proposed model, PCA and KPCA fault detection methods are chosen to comparative study, the results show that the proposed method is feasible and efficient; and evidently improves the effect of fault detection.Control charts as an effective tool of the statistical process control, the patterns recognition plays an important role in dealing with the complex production process quality problems. For the identification difficulty of control chart mixture pattern, statistical and sharp features of observation data are used as the eigenvalue of quality information, and then principal component analysis as the second feature extraction to get the effective data for the classifier. Multi-class support vector machines apply for recognizing the control chart patterns. And adaptive mutation particle swam optimization is used to optimize the SVM classifier by searching the best values of the parameters of SVM. The six basic patterns and four mixed patterns are used to analyze the proposed model, and the other three methods for comparative study, respectively, neural networks, support vector machine and support vector machines based on principal component analysis. The simulation results show the proposed method has better recognition accuracy compared with other methods. It provides reference value for the control chart pattern recognition research.According to the complex production process characteristics of multi-input, multi-output, nonlinear, a multi-model quality prediction approach based on local models is proposed. Firstly, classify the operation condition using the K-means clustering algorithm; then establish the local quality prediction models using support vector machine based on the multiple loading conditions; finally, get the local model weights using the adaptive mutation particle swam optimization, so that obtain the global model to realize the production process quality prediction. A complex processe, the normal mode of the TE process, is used to simulate and verify the proposed multi-model, and then uses local model, BP and SVM model as the comparative study. The results show that the prediction of the model evidently improves the prediction accuracy, which compared to other forecasting methods, and the relative error is controlled about1%. At the same time, the results also show that the proposed method is feasible and efficient.The complexity production process is difficult to describe by a purely abstract mathematical model, and manufacturers hope to simulate the production process by virtual simulation, which can analyze the quality factors of the production process and find the factors of its quality problems. Hot mix asphalt (HMA) aggregate gradation control, which based on the frame work of mathematical model, simulates HMA mix aggregates gradation under different production utilizing discrete event simulation software Arena. Multi-factors control logic is established by the embedded compiler module of Arena software, and then image output is used to observe the production status changes. The simulations results show that system optimize control results are related with control strategies, the suitable control strategy can make a significant reduction in the quality problems. Simulation control method provides a good way for HMA quality control. It greatly shortens the test time, timely controls gradation deviation, and improves the overall quality of the product. It can be better to reduce the deviation for improving the production quality and provides an efficient way for HMA gradation control in the real world.In this paper, the intelligent methods, like support vector machines, particle swam optimization, kernel principal component analysis, wavelet packet analysis and simulation technology, are used for quality control of complex production process, Control chart pattern recognition, fault detection, quality prediction and virtual intelligent control are adopted to implement quality control of complex production process. The proposed methods have high theoretical value and production practical significance, and laid a solid foundation for the future study.