Thesis Title: Automatic Feedback Control in High Precision Manufacturing 

Advisors:  

Dr. Jianjun Shi, School of Industrial and Systems Engineering, Georgia Tech 

Dr. Kamran Paynabar, School of Industrial and Systems Engineering, Georgia Tech 

Committee members:  

Dr. Jing Li, School of Industrial and Systems Engineering, Georgia Tech 

Dr. Chun Zhang, School of Industrial and Systems Engineering, Georgia Tech 

Dr. Lance Waller, Rollins School of Public Health, Emory

Date and Time: 9:00 am (EST), Tuesday, April 20th, 2021  

Meeting URL:   https://bluejeans.com/254150254  

Meeting ID:  254150254 (BlueJeans)  

Abstract:  

Automatic feedback control has been widely used in manufacturing systems to reduce process variability and ensure on-target product quality.  In this thesis, the methodologies of automatic control are further investigated to address model uncertainties, high-dimensional sensing feedback control, and their applications in challenging engineering problems.  Motivated by real needs from current industrial production systems, three control methods are studied in this thesis in Chapters 1, 2, and 3.

In Chapter 1, an adaptive cautious regularized run-to-run control scheme is developed for overlay control in photolithography processes. Photolithography is the bottleneck for quality improvement in semiconductor manufacturing. The decreasing critical dimensions of the semiconductor product require more effective run-to-run control technology. Currently, Exponential Weighted Moving Average (EWMA) control scheme is widely used in the overlay control of lithography processes. In this chapter, three shortcomings of the current EWMA run-to-run control scheme are investigated: (i) the existing EWMA control scheme has its weight parameter λ set as a fixed value, which does not perform well when the process changes; (ii) the existing EWMA control scheme does not consider the model and parameter uncertainties in practice; and (iii) the adjustable range of the control variables is not considered in the existing EWMA control scheme. To address these limitations, we propose a new adaptive, cautious, and optimal run-to-run control scheme. The effectiveness of the new controller is validated through surrogated simulation studies.

In Chapter 2, an image-based feedback control strategy is developed by using tensor representation and analysis. The problem is motivated by the photolithography process, where the system output is image signals measuring the overlay error, and the control inputs are tuning vectors.  To develop a control strategy, one first needs to off-line estimate the process model by finding the relationship between the image output and vector inputs, and then to obtain the control law by online minimizing the control objective function. The main challenges in achieving such a control strategy include (i) the high dimensionality of the output in building a regression model, (ii) the spatial structure of image outputs and the temporal structure of the image sequence, and (iii) non-i.i.d noises. To address these challenges, we propose a novel tensor-based process control approach by incorporating the tensor time series and regression techniques. Based on the process model, we obtain the control law by minimizing a control objective function. Although our proposed approach is demonstrated with the 2D images as the system output, it can have the potential to be extended to the higher-order tensors such as video signals or point cloud data. Simulation and case studies show that our proposed method is more effective than benchmark methods in terms of relative mean square error.

Chapter 3 will investigate how to achieve half-fuselage assembly via active control. In a half fuselage assembly process, shape control is vital for achieving ultra-high precision assembly. To achieve better shape adjustment, we need to determine the optimal location and force of each actuator to push and pull a fuselage to compensate for its initial shape distortion. The current practice achieves this goal by solving a surrogate model-based optimization problem. However, there are two limitations in this surrogate model-based method: (1) Low efficiency: Collecting training data for surrogate modeling from many FEA replications is time-consuming. (2) Non-optimality: The required number of FEA replications for building an accurate surrogate model will increase as the potential number of actuator locations increases. Therefore, the surrogate model can only be built on a limited number of prespecified potential actuator locations, which will lead to sub-optimal control results. To address these issues, this chapter proposes an FEA model-based automatic optimal shape control (AOSC) framework. This method directly loads the system equation from the FEA simulation platform to determine the optimal location and force of each actuator. Moreover, the proposed method further integrates the cautious control concept into the AOSC system to address model uncertainties in practice. The case study with industrial settings shows that the proposed Cautious AOSC method achieves higher control accuracy compared to current industrial practice.