Sliding mode observer based multi-layer metal plates core temperature on-line estimation for semiconductor intelligence manufacturing

This paper focuses on the technological development of core temperature estimation algorithm of multi-layer metal plates. It applies new control technology in the field of real time monitoring of semiconductor producing processes. To achieve real time core temperature estimation in semiconductor equipment processes, this paper will follow the order of: system description, model derivation, parameter identification, and design of a robust core temperature observer of multi-layer metal plates, which will then be cross validated with experimental data. In the metal heating system model proposed in this paper, external cooling is considered as an unknown interference. Since the system contains an unknown interference, the sliding mode observer (SMO) will use the equivalent conversion technique to tackle with the uncertainty. In order to improve the degree of freedom and flexibility in the design of the gain matrix, and considering the effects of parameter identification uncertainty, this paper introduces the multi-objective linear matrix inequality (LMI) in the design of the sliding mode observer to suppress the impacts of the non-matching uncertainty on the system and to reduce the gain matrix which also satisfies the convergence requirement of the designer. In terms of algorithm implementation, the parameters of the thermally processed multilayer plate model are first identified through the minimum difference filter (MDF), the data is then filtered offline, and lastly, the model parameter is derived using the least square (LS) method. The filtering technology can greatly improve the accuracy of parameter identification and improve the SMO estimation precision. Finally, the experimental data of the actual semiconductor machine and the estimation from the proposed method verifies the usefulness of the SMO in core temperature estimation of multi-layer metal plate heating systems.