Browsing by Author "Zhang, Chang"
Results Per Page
Sort Options
Item Comparison of Run-to-Run Control Methods in Semiconductor Manufacturing Processes(2000) Zhang, Chang; Deng, Hao; Baras, John S.; Baras, John S.; ISRRun-to Run (RtR) control plays an important role in semiconductor manufacturing.In this paper, RtR control methods are generalized. The set-valued RtR controllers with ellipsoidapproximation are compared with other RtR controllers bysimulation according to the following criteria: A good RtR controller should be able to compensate for variousdisturbances, such as process drifts, process shifts (step disturbance)and model errors; moreover, it should beable to deal with limitations, bounds, cost requirement, multipletargets and time delays that are often encountered in realprocesses.
Preliminary results show the good performance of the set-valued RtRcontroller. Furthermore, this paper shows that it is insufficient to uselinear models to approximate nonlinear processes and it is necessary to developnonlinear model based RtR controllers.
Item A Hierarchical Structure For Finite Horizon Dynamic Programming Problems(2000) Zhang, Chang; Baras, John S.; Baras, John S.; ISR; CSHCNIn dynamic programming (Markov decision) problems, hierarchicalstructure (aggregation) is usually used to simplify computation. Most research on aggregation ofMarkov decision problems is limited to the infinite horizon case, which has good tracking ability. However, in reallife, finite horizon stochastic shortest path problems are oftenencountered.In this paper, we propose a hierarchical structure to solve finite horizon stochastic shortest pathproblems in parallel. In general, the approach reducesthe time complexity of the original problem to a logarithm level, which hassignificant practical meaning.
Item A New Adaptive Aggregation Algorithm for Infinite Horizon Dynamic Programming(2001) Zhang, Chang; Baras, John S.; ISR; CSHCNDynamic programming suffers the "curse of dimensionality" when it isemployed for complex control systems. State aggregation is used to solvethe problem and acceleratecomputation by looking for a sub-optimal policy. In this paper, a new method, which converges much faster thanconventional aggregated value iteration based on TD(0), is proposed for computing the valuefunctions of theaggregated system. Preliminary results show that the new method increases thespeed of convergence impressively. Aggregation introduces errorsinevitably. An adaptive aggregation scheme employing the newcomputation method isalso proposed to reduce the aggregation errors.Item Performance Evaluation of Run-to-Run Control Methods in Semiconductor Processes(2001) Zhang, Chang; Baras, John S.; Baras, John S.; ISR; CSHCNRun-to-Run (RtR) control plays an important role in semiconductormanufacturing processes. In this paper, RtR control methods are classified and evaluated. The set-valued RtR controllers with ellipsoid approximation are compared with two typical RtR controllers: the Exponentially Weighted Moving Average (EWMA) controller and the Optimizing Adaptive Quality Controller (OAQC) by simulations according to the following criteria: A good RtR controller should be able to compensate for various disturbances, such as small drifts, step disturbances and model errors; moreover, it should be able to deal with bounds, cost requirement and multipletargets that are often encountered in semiconductor processes. Based on our simulation results, suggestions on selection of a proper RtR controller for a semiconductor process are given as conclusions.Item Run-to-Run Control Methods Based on the DHOBE Algorithm(1999) Deng, Hao; Zhang, Chang; Baras, John S.; ISR; CSHCNMany run-to-run (RtR) control methods have been developed in recentyears. Two particular set-valued RtR control schemes based on the Dasgupta-Huang OptimalBounded Ellipsoid (DHOBE) algorithm are introduced. Compared to other RtR control schemes, the methods in this paper only need to know the bound of the noises, and are easyto implement.The DHOBE algorithm, for eachrecursion, returns an outer bounding ellipsoid of the estimated parameters. If the center of the ellipsoid each time istaken as the model coefficients, the explicit model update isimplemented which leads to a model-reference method. If we choose theworst-case point which maximizes the cost function in the set, then wecan apply the set-valued worst case approach. These two methods were compared with two other main RtRcontrol schemes: the Exponentially Weighted Moving Average (EWMA) methodand the Optimizing Adaptive Quality Controller (OAQC) method. Simulation results showed the superior performance of the RtRcontrollers based on the DHOBE algorithm. Furthermore this paper showedthat it is necessary to applynonlinear models to compensate for severe nonlinear processes.
Item The Set-Valued Run-to-Run Controller in Semiconductor Manufacturing Processes(1999) Zhang, Chang; Baras, John S.; ISRIn semiconductor manufacturing, run-to-run (RtR) control is paid moreand more attention. In this paper a set-valuedRtR control scheme is introduced. Different from conventional RtR controlmethods, the set-valued method first calculates the feasibleparameter set at the beginning of each run, then estimates the modelparameters within this set. Compared to other RtR control schemes, itdoes not assume any statistical property of the noises. In simulation itwas shown that it is robust tomodel and sensor errors, and it has the potential to be applied tohighly nonlinear processes. Furthermore the set-valued method can beapplied to other fields such as signal processing and chemical processes.Item The Set-Valued Run-to-Run Controller with Ellipsoid Approximation(2000) Zhang, Chang; Baras, John S.; Baras, John S.; ISRIn order to successfully apply Run-to-Run (RtR) control or real time control ina semiconductor process, it is very important to estimate the processmodel. Traditional semiconductor process control methods neglect theimportance of robustness due to the estimation methods they use.A new approach, namely the set-valued RtR controller with ellipsoidapproximation, is proposed to estimate the process model from acompletely different point of view. Because the set-valued RtRcontroller identifies the process model in the feasible parameter setwhich is insensitive to noises, the controller is robust to theenvironment noises.Ellipsoid approximation can significantly reduce the computation load for the set-valued method.
In this paper, the Modified Optimal Volume Ellipsoid (MOVE) algorithm is used toestimate the process model in each run. Designof the corresponding controller and parameter selection of the controller are introduced.Simulation results showed that the controller is robust toenvironment noises and model errors.