A 3D Intelligent Modelling, Learning and Control for Complex Tempo-spatial Processes in the Production Industry

Description

Many industrial processes are complex distributed parameter systems that have a strong tempo-spatial nature. It is very difficult to model and control this kind of process due to the infinite-dimensional nature, limited sensing and actuating means available in practice. The researchers aim to develop a novel three-domain (3D) intelligent modelling and control approach for a broad range of tempo-spatial processes in production industry. In view of difficulties inherent in the complex tempo-spatial process, it is essential to resort to a carefully considered hybrid modelling approach to obtain a 3D model. First, a 3D kernel is proposed to decompose the process under the traditional Volterra series framework, upon which the distributed time-space coupling is separated by the Karhunen-Loeve (K-L) method. Various intelligent learning methods can then be utilized for model enhancement to compensate time-domain and spatial uncertainties, respectively. Afterwards, a model-based hybrid intelligent control/supervision framework will be innovatively developed to maintain a good 3D performance. The low-level fuzzy control system will play a dominant role for nominal performance with the guaranteed stability region; while the high-level supervision will fine-tune the setpoints for a better global performance based on the developed 3D model.