Juergen Hahn

Survey Automatic control in microelectronics manufacturing: Practices, challenges, and possibilities

Advances in modeling and control will be required to meet future technical challenges in microelectronics manufacturing. The implementation of closed-loop control on key unit operations has been limited due to a dearth of suitable in situ measurements, variations in process equipment and wafer properties, limited process understanding, non-automated operational practices, and lack of trained personnel. This paper reviews the state-of-the-art for process control in semiconductor processing, and covers the key unit operations of lithography, plasma etching, thin film deposition, rapid thermal processing, and chemical-mechanical planarization. The relationship of process (equipment) models to control strategies is elaborated because recently there has been a considerable level of activity in model development in industry and academia. A proposed control framework for integrating factory control and equipment scheduling, supervisory control, feedback control, statistical process control, and fault detection/diagnosis in microelectronics manufacturing is presented and discussed.

An improved method for nonlinear model reduction using balancing of empirical gramians

Abstract Nonlinear model predictive control has become increasingly popular in the chemical process industry. Highly accurate models can now be simulated with modern dynamic simulators combined with powerful optimization algorithms. However, computational requirements grow with the complexity of the models. Many rigorous dynamic models require too much computation time to be useful for real-time model based controllers. One possible solution to this is the application of model reduction techniques. The method introduced here reduces nonlinear systems while retaining most of the input–output properties of the original system. The technique is based on empirical gramians that capture the nonlinear behavior of the system near an operating point. The gramians are then balanced and the less important states reduced via a Galerkin projection which is performed onto the remaining states. This method has the advantage that it only requires linear matrix computations while being applicable to nonlinear systems.

Advances and selected recent developments in state and parameter estimation

Abstract This paper deals with two topics from state and parameter estimation. The first contribution of this work provides an overview of techniques used for determining which parameters of a model should be estimated. This is a question that commonly arises when fundamental models are used as these models often contain more parameters than can be reliably estimated from data. The decision of which parameters to estimate is independent of the observer/estimator design, however, it is directly affected by the structure of the model as well as the available data. The second contribution is an overview of recent developments regarding the design of nonlinear Luenberger observers, with special emphasis on exact error linearization techniques, but also discussing more general issues, including observer discretization, sampled data observers and the use of delayed measurements.

Controllability and observability covariance matrices for the analysis and order reduction of stable nonlinear systems

Abstract This paper presents a framework for nonlinear systems analysis that is based upon controllability and observability covariance matrices. These matrices are introduced in the paper and it is shown that gramians for linear systems form special cases of the covariance matrices. The covariance matrices can be transformed via a balancing-like transformation and nonlinearity measures are defined based upon these transformed covariance matrices. Subsequently, the covariance matrices are used for reduction of the nonlinear model. It is shown that the model reduction procedure reduces to balanced model truncation for linear systems for impulse inputs. Furthermore, it is also shown that several model reduction procedures that were developed by other researchers, and assumed to be independent from one another, are related. The findings are illustrated with an example.

Experimental and computational optimization of an Escherichia coli co-culture for the efficient production of flavonoids

Metabolic engineering and synthetic biology have enabled the use of microbial production platforms for the renewable production of many high-value natural products. Titers and yields, however, are often too low to result in commercially viable processes. Microbial co-cultures have the ability to distribute metabolic burden and allow for modular specific optimization in a way that is not possible through traditional monoculture fermentation methods. Here, we present an Escherichia coli co-culture for the efficient production of flavonoids in vivo, resulting in a 970-fold improvement in titer of flavan-3-ols over previously published monoculture production. To accomplish this improvement in titer, factors such as strain compatibility, carbon source, temperature, induction point, and inoculation ratio were initially optimized. The development of an empirical scaled-Gaussian model based on the initial optimization data was then implemented to predict the optimum point for the system. Experimental verification of the model predictions resulted in a 65% improvement in titer, to 40.7±0.1mg/L flavan-3-ols, over the previous optimum. Overall, this study demonstrates the first application of the co-culture production of flavonoids, the most in-depth co-culture optimization to date, and the first application of empirical systems modeling for improvement of titers from a co-culture system.

Genetic/quadratic search algorithm for plant economic optimizations using a process simulator

Abstract The genetic/quadratic search algorithm (GQSA) is a hybrid genetic algorithms (GA) for optimizing plant economics when a process simulator models the plant. By coupling a regular GA with an algorithm based upon a quadratic search, the required number of objective function evaluations for obtaining an acceptable solution decreases significantly in most cases. The GQSA combines advantages of GA and quadratic search techniques, e.g. determining a global optimum for a problem with a high probability for discontinuous as well as non-convex optimization problems while at the same time providing faster convergence than conventional GA. The performance of both the GQSA and the GA was compared using four different test functions and an economic optimization problem for a turbo-expander process. Numerical test results indicate that the convergence of the GQSA is either better than or at least comparable to those of GA for all tests employing the same genetic parameters.

Introduction of a nonlinearity measure for principal component models

Abstract Although principal component analysis (PCA) is an important tool in standard multivariate data analysis, little interest has been devoted to assessing whether the underlying relationship within a given variable set can be described by a linear PCA model or whether nonlinear PCA must be utilized. This paper addresses this deficiency by introducing a nonlinearity measure for principal component models. The measure is based on the following two principles: (i) the range of recorded process operation is divided into smaller regions; and (ii) accuracy bounds are determined for the sum of the discarded eigenvalues. If this sum is within the accuracy bounds for each region, the process is assumed to be linear and vice versa. This procedure is automated through the use of cross-validation. Finally, the paper shows the utility of the new nonlinearity measure using two simulation studies and with data from an industrial melter process.

Fault detection and classification in chemical processes based on neural networks with feature extraction.

Feed forward neural networks are investigated here for fault diagnosis in chemical processes, especially batch processes. The use of the neural model prediction error as the residual for fault diagnosis of sensor and component is analyzed. To reduce the training time required for the neural process model, an input feature extraction process for the neural model is implemented. An additional radial basis function neural classifier is developed to isolate faults from the residual generated, and results are presented to demonstrate the satisfactory detection and isolation of faults using this approach.

Reduction of nonlinear models using balancing of empirical gramians and Galerkin projections

Nonlinear model predictive control has become increasingly popular in the chemical process industry. However, computational requirements grow with the complexity of the models. Many rigorous dynamic models require too much computation time to be useful for real-time model based controllers. This presents a need for model reduction techniques. The method introduced here reduces nonlinear systems, while retaining most of the input-output properties of the original system. The reduction itself is based on empirical gramians which capture the nonlinear behavior of the system in a region around an operating point. The gramians are then balanced and the less important states reduced. A Galerkin projection is performed onto the remaining states. This method has the advantage that it only requires linear matrix computations while being applicable to nonlinear systems.

Investigation of IL-6 and IL-10 signalling via mathematical modelling

Steatosis, i.e., the accumulation of fat in hepatocytes, plays an important role in the progression of non-alcoholic fatty liver disease (NAFLD). It has been shown that STAT3 activation is involved in a decrease of lipid accumulation while C∕EBP is correlated with an increase of fat content and steatosis. It is known that STAT3 and C∕EBP are activated by IL-6 and that IL-6 signalling is also affected by IL-10, even though the exact mechanism is unclear. This paper develops a model for IL-6 and IL-10 signal transduction and then investigates the effect that stimulation with these cytokines has on the transcription factor dynamics. In an initial step, some parameters of a previously developed IL-6 signalling model are re-estimated based upon newly developed experimental data for the Jak-STAT pathway. Furthermore, the Erk-C∕EBP pathway model is extended to also include the activated transcription factor C∕EBP in the nucleus. Since IL-10 signals through the Jak-STAT but not the Erk-C∕EBP pathway, a model was developed which includes interaction between IL-6 and IL-10 signalling as both mechanisms share signal transduction through the Jak-STAT pathway. Based upon the model, the activity ratio of Jak-STAT and Erk-C∕EBP was investigated for different stimulation levels of IL-6 and IL-10.