Songtao Zhang

Bio: Songtao Zhang is an academic researcher from University of New Brunswick. The author has contributed to research in topics: Frame (networking) & Model predictive control. The author has an hindex of 1, co-authored 2 publications receiving 34 citations.

A principal component analysis model-based predictive controller for controlling part warpage in plastic injection molding

Abstract: Quality measurement on injection molded parts are related to process parameter data.The data experimentation data is analyzed using statistical methods (PCA), (ICA) and (ANOVA).Models are constructed to predict a quality index based on data analysis results.A MPC scheme that allows for on-line quality control was developed using the models.The presented scheme is extendable to a variety of manufacturing processes. Quality control is an important aspect of manufacturing processes. Product quality of injection molded parts is influenced by the injection molding process. In this study statistical tools were used to develop a model that relates injection molding process variables to part quality. A statistically based model predictive control algorithm was developed for controlling part quality with manipulated variables coolant flow rate and coolant temperature. This approach replaces the need of off-line quality measurement and provides real-time injection modeling quality control.

Applying expectation-maximization evaluation on approximate optimal control

Abstract: In this paper we proposed an approach of approximating optimal tracking via expectation-maximization (EM) evaluation. From the discussion of applying reinforcement learning (RL) for a system with unknown internal dynamics, we present the challenge of using a classical frame of Q-learning on a tracking task. Further we explained the idea of redefining the cost function (i.e. criterion) of Q-learning to satisfy the requirement for the system dynamic knowledge for the tracking task. We explained the advantages of dividing the original trajectory tracking task into two machine learning subtasks (i.e. learning the quadratic regulator and learning the baseline command generator) on-line. Details are given on the integration of the Q-learning frame and EM algorithm as well as the convergence to the optimum control via iterative estimation of an optimal regulator and a baseline generator. Initial simulation results of this approach using a second order system showed the ability of the Q-learning frame integrated with the EM algorithm approximates to the optimal tracking task.

Reinforcement Learning and Adaptive Dynamic Programming for Feedback Control

Abstract: This paper first offers a transversal view through microfluidics theory and applications, starting from a brief overview on microfluidic systems and related theoretical issues, covering different kinds of phenomena, from continuous to multiphase flow. Multidimensional models, from lumped parameters to numerical models and computational solutions, are then considered as preliminary tools for the characterization of spatio-temporal dynamics in microfluidic flows. Following these, experimental approaches through original monitoring opto-electronic interfaces and systems are discussed. Finally, a vision of two phase microfluidic phenomena is given through nonlinear analyses applied to experimental time series.

Quality Prediction for Injection Molding by Using a Multilayer Perceptron Neural Network.

Abstract: Injection molding has been widely used in the mass production of high-precision products. The finished products obtained through injection molding must have a high quality. Machine parameters do not accurately reflect the molding conditions of the polymer melt; thus, the use of machine parameters leads to erroneous quality judgments. Moreover, the cost of mass inspections of finished products has led to strict restrictions on comprehensive quality testing. Therefore, an automatic quality inspection that provides effective and accurate quality judgment for each injection-molded part is required. This study proposes a multilayer perceptron (MLP) neural network model combined with quality indices for performing fast and automatic prediction of the geometry of finished products. The pressure curves detected by the in-mold pressure sensor, which reflect the flow state of the melt, changes in various indicators and molding quality, were considered in this study. Furthermore, the quality indices extracted from pressure curves with a strong correlation with the part quality were input into the MLP model for learning and prediction. The results indicate that the training and testing of the first-stage holding pressure index, pressure integral index, residual pressure drop index and peak pressure index with respect to the geometric widths were accurate (accuracy rate exceeded 92%), which demonstrates the feasibility of the proposed method.

Intelligent Injection Molding on Sensing, Optimization, and Control

Abstract: Injection molding is one of the most significant material processing methods for mass production of plastic products. It is widely used in various industry sectors, and its products are ubiquitous in our daily life. The settings and optimization of the injection molding process dictate the geometric precision and mechanical properties of the final products. Therefore, sensing, optimization, and control of the injection molding process have a crucial influence on product quality and have become an active research field with abundant literature. This paper defines the concept of intelligent injection molding as the integral application of these three procedures—sensing, optimization, and control. This paper reviews recent studies on methods for the detection of relevant physical variables, optimization of process parameters, and control strategies of machine variables in the molding process. Finally, conclusions are drawn to discuss future research directions and technologies, as well as algorithms worthy of being explored and developed.