Decision Preference Mining Method Base On Multimodal Fusion And Its Applications In Process Control

Traditional industrial process optimization control always has difficulties in establishing an accurate mathematical model in the complex environments;as well,such as controller parameter tuning issues,etc.often rely on expert decision-making experience when the process model changes.However,towards different decision-making goals,because there is lack of a consistent model for describing expert decision-making preferences,expert decision making experiences are often difficult to learn.Therefore,it is necessary to develop an effective way to gradually obtain the decision preference model along with the expert tuning process.Decision preference performs comprehensively based on the expert's personal cognition,emotion,theoretical knowledge and experience accumulation,which is difficult to be fully quantified.Thereafter,decision preference also contains the direction and goal that expected by the expert.At present,there is no unified model to explain and reflect personal preferences in expert decision-making.With the in-depth study of the human psychology and the advancement of emotional computing technology,compared with traditional decision-making methods,cognitive affective can not only better quantify human wishes in decision-making,but also quantify expert decisions preferences.The expert decision preference mentioned in this article specifically refers to the preference under the affective computing model.The description of expert decision preference is a multi-modal process.Different modalities describe different aspects of affective including complementary information.Affective recognition is generally carried out through modal information such as facial expressions,voice,EEG,etc.,but it is generally difficult to meet these conditions in actual engineering operations.Therefore,in regard with the industrial control process,this paper considers multi-modal fusion of expert operation data and control curves.Combining this information,this work build a more robust emotion recognition model,which can identify expert decision-making decision preferences.Therefore,it is necessary to explore the affective interaction mechanism that generates expert decision preferences and study how computers mine expert decision preferences,which can improve the efficiency of the operation optimization process and effectively reduce the workload of operators.Undoubtedly this work has academic research significance and engineering application value.The main research contents of this paper are as follows:1.A new cognitive affective computing model based on personality and PAD(Pleasure-Arousal-Dominance)affective state model is proposed.In addition,a new affective preference is defined to explain the affective computing model updating in affective interaction.Hence,an affective parameter mining strategy based on genetic algorithm was established.2.A multi-modal fusion-based expert decision preference mining method is developed.A hybrid neural network based on CNN-LSTM is used for the multi-modal feature fusion process for affective classification and recognition.Then,the expert decision preference is mined through the affective computing process.3.The proposed method is applied in the PID controller tuning of the coupled loop system and the optimization problem in the batch beer fermentation process,verifying the feasibility and effectiveness of our proposed multi-modal affective computing expert decision preference mining method.