Cheol W. Lee

Construction of fuzzy systems using least-squares method and genetic algorithm

The fuzzy basis function network which was proposed in Wang and Mendel (IEEE Trans. Neural Networks 3(5) (1992b) 807) provides a way of representing fuzzy inference systems in a simple structure similar to those of radial basis function networks. In this paper, two new algorithms based on the least-squares method and genetic algorithm are proposed for autonomous learning and construction of fuzzy basis function networks when training data are available. The proposed algorithms add a significant fuzzy basis function node at each iteration during training, based on error reduction measures. The first, a least-squares algorithm, provides a way of sequentially constructing meaningful fuzzy systems which are not possible to achieve with the orthogonal least-squares algorithm, while the second, an adaptive least-squares algorithm based on the combined least-squares and genetic algorithm, realizes hybrid structure-parameter learning without human intervention. Simulation studies are performed with numerical examples for comparison of its performance against the orthogonal least-squares algorithm, backpropagation algorithm, and conventional genetic algorithm. The adaptive least-squares algorithm is also applied to a real world problem to construct a fuzzy basis function network model for surface roughness in a grinding process using experimental data.

Intelligent Model-based Optimization of the Surface Grinding Process for Heat-Treated 4140 Steel Alloys With Aluminum Oxide Grinding Wheels

This paper presents implementation results of surface grinding processes based on the model-based optimization scheme proposed by Lee and Shin (Lee, C. W., and Shin, Y. C., 2000 Evolutionary Modeling and Optimization of Grinding Processes, Int. J. Prod. Res. 38(12), pp. 2787-2813). In order to accomplish this goal, process models for grinding force, power, surface roughness, and residual stress are developed based on the generalized grinding model structures using experimental data. The time-varying characteristics due to wheel wear are also investigated in order to determine the optimal dressing interval. Grinding optimization is considered as constrained nonlinear optimization problems with mixed-integer variables and time-varying characteristics in this study. Case studies are performed with various optimization objectives including minimization of grinding cost, minimization of cycle time, and process control. The optimal process conditions determined by the optimization scheme are validated by experimental results.

Evolutionary modelling and optimization of grinding processes

The objective of this study is to develop a framework of modelling the complex grinding processes and finding optimal process conditions to meet the general class of process requirements. In order to achieve the above goal, novel modelling schemes and optimization methods based on evolutionary algorithms (EA) are developed. The optimization problem of grinding processes can be formulated as a constrained non-linear programming problem with mixed-discrete variables. The adaptive least-squares (ALS) algorithm proposed by Lee and Shins 1998 study is extended for modelling multi-input-multi-output (MIMO) complex grinding processes using fuzzy basis function networks (FBFN), while the modified evolution strategies (ES) is proposed for successful optimization of grinding processes. Two grinding optimization problems demonstrate the superior performance of the proposed scheme.

Dynamic Optimization of the Grinding Process in Batch Production

This paper presents a novel dynamic optimization framework for the grinding process in batch production. The grinding process exhibits time-varying characteristics due to the progressive wear of the grinding wheel. Nevertheless, many existing frameworks for the grinding process can optimize only 1 cycle at a time, thereby generating suboptimal solutions. Moreover, dynamic scheduling of dressing operations in response to process feedback would require significant human intervention with existing methods. We propose a unique dynamic programming-evolution strategy framework to optimize a series of grinding cycles depending on the wheel condition and batch size. In the proposed framework, a dynamic programming module dynamically determines the frequency and parameter of wheel dressing while the evolution strategy locates the optimal operating parameters of each cycle subject to the constraints on the operating ranges and part quality. Case studies based on experimental data are conducted to demonstrate the advantages of the proposed method over conventional approaches.

Construction of fuzzy basis function networks using adaptive least squares method

A novel algorithm based on the least squares (LS) method and genetic algorithm (GA) is proposed for autonomous learning and construction of FBFNs when training data are available. The proposed algorithms add significant fuzzy basis functions (FBF) at each iteration during training, based on error reduction measures. The adaptive least squares (ALS) algorithm based on the combined LS and GA, realizes hybrid structure-parameter learning without any human intervention. Simulation studies are performed with numerical examples for comparison with conventional algorithms. The ALS algorithm is applied to the construction of a fuzzy basis function network model for surface roughness in a grinding process using experimental data.

Intelligent Direct Torque Control of Brushless DC motors for hybrid electric vehicles

This paper investigates the application of neural networks for Direct Torque Control (DTC) of a Brushless DC (BLDC) motor with non-sinusoidal back EMF. Conventional DTC technique controls the torque directly by providing appropriate switching signals from a predefined switching table based on torque error, stator flux linkage error and the stator flux angle. Applying this method for hybrid electric vehicles, results in serious torque ripple and power loss due to several system limitations. An intelligent neural network based direct torque control of BLDC motors for hybrid electric vehicle applications is proposed in this paper. The proposed method decreases the torque ripple and the number of switching and hence the switching power loss. Both the conventional DTC method and neural network based DTC of BLDC motor are simulated in MATLAB/SIMULINK and the results are compared and discussed to verify the proposed control.

Estimation Strategy for a Series of Grinding Cycles in Batch Production

Typical grinding operations in batch production are characterized by multiple data streams sampled at distinct intervals. A unique estimation strategy is proposed for integrating rapidly sampled sensor signals with postprocess inspection data from a series of grinding cycles. After a nonlinear state-space model is derived from existing analytical models, system observability is tested for various combinations of sensors and measurement settings. A multirate simultaneous state and parameter estimation scheme is developed based on extended Kalman filters for real-time estimation of the model parameters and part quality. Results from case studies demonstrate that the proposed scheme enables challenging estimation tasks to be undertaken that cannot be performed using traditional approaches.

Modeling of complex manufacturing processes by hierarchical fuzzy basis function networks with application to grinding processes

A framework for modeling complex manufacturing processes using fuzzy neural networks is presented with a novel training algorithm. In this study, a hierarchical structure that consists of fuzzy basis function networks (FBFN) is proposed to construct comprehensive models of the complex processes. A new adaptive least-squares (ALS) algorithm, based on the least-squares method and genetic algorithm (GA), is proposed for autonomous learning and construction of FBFNs without any human intervention. Simulation studies are performed to demonstrate advantages of the proposed modeling framework with the training algorithm in modeling complex manufacturing processes. The proposed method is implemented for the surface grinding processes based on the hierarchical structure of FBFNs. Process models for surface roughness and residual stress are developed based on the available grinding model structures with a small number of experimental data to demonstrate the concept. The accuracy of developed models is validated through independent sets of grinding experiments.

Development of the process model for plunge grinding and optimization of grinding process

This article presents improved grinding roughness model established for cylindrical plunge grinding though analysing the existing roughness models. The proposed roughness model is to consider the uncertain effects of grinding, such as changes of grinding conditions and circumstance, as grinding procedure is progressed. In order to consider the uncertain effect of grinding, the time delay between programmed and actual infeed rate of grinding table is selected as weighting factor of the proposed roughness model. The developed roughness model is also used for the optimization algorithm of grinding procedure. Optimization algorithm in this study is constructed to minimize the grinding cost and to obtain the optimized dressing and grinding conditions under grinding constraints such as no-burn condition, limitation of roughness of workpiece, grinding power, etc. The optimized results also give an optimized dressing interval in batch production. The used optimization algorithm is an Evolutionary Strategy algorithm, and performance of the proposed algorithm was evaluated with experiments.

A New Dynamic State Space Model for the Cylindrical Grinding Process

A new dynamic state space model is proposed for the in-process estimation and prediction of part qualities in the plunge cylindrical grinding process. A through review on various grinding models in literature reveals a hidden dynamic relationship among the grinding conditions, the grinding power, the surface roughness, and the part size due to the machine dynamics and the wheel wear, based on which a nonlinear state space equation is derived. After the model parameters are determined according to the reported values in literature, several simulations are run to verify that the model makes good physical sense. Since some of the output variables, such as the actual part size, may or may not be measured in industry applications, the observability is tested for different sets of outputs in order to see how each set of on-line sensors affects the observability of the model. The proposed model opens a new way of estimating the part qualities such as the surface roughness and the actual part size based on application of the state estimation algorithm to the measured outputs such as the grinding power. In addition, a long term prediction of the part qualities in batch grinding processes would be realized by simulation of the proposed model. Possible applications to monitoring and control of grinding processes are discussed along with several technical challenges lying ahead.Copyright