Materials And Electrical Engineering Interdisciplinary Integration Adaptation Mechanism For Power Supply Design Intelligent Assistant

When faced with the task of designing an inductor under high current conditions,switched-mode power supplies(SMPSs)engineers need to master the knowledge of saturation characteristics magnetic materials of inductor components and inductor current ripple.In the context of cross-disciplinary,this is an interdisciplinary research problem with the goal of interdisciplinary integration.In the background of cognitive psychology,it is a problem of cognitive skills adaptation to the changes imposed by the interdisciplinary environment with the approach of specialization of practical knowledge.On an abstract level,the inductor design and the analysis of the saturation characteristics of the inductance core material of SMPSs are typical interdisciplinary complex problems.The study of interdisciplinary integration and skills adaptation is of general significance for interdisciplinary research.However,the complexity of interdisciplinary integration and skills adaptation,as well as the limited human capacity for information processing can lead to cognitive overload for designers.Therefore,it is imperative to develop an interactive cognitive tool using artificial intelligence(AI)-assisted technologies to simulate,extend,and expand human cognition so as to improve designers' cognitive processes,reduce cognitive load,and achieve skills adaptation and interdisciplinary integration.In this paper,we take the SMPS as the research object to investigate the mechanism of cognitive skills adaptation to the changes imposed by the interdisciplinary environment,explore the methodology to achieve the goal of interdisciplinary integration,and provide cognitive skills representations and cognitive tools for AI-aided design.The complexity of problems,constraints,knowledge approaches,and systems in high-performance SMPSs design determines the need for interdisciplinary research.Compared with single disciplines,the cognitive tool is more difficult to assist interdisciplinary research,because single disciplines and interdisciplines have different learning ways.The single disciplinary approach focuses on knowledge kernel growth,whereas the interdisciplinary approach emphasizes more on skills adaptation and interdisciplinary integration.Traditional learning theories,such as rule discrimination theory,reveals unrealistic demands on the human's memory when solving the interdisciplinary complex problem as it is hard to know in advance what features of the situation are critical for future discrimination.Humans can draw conclusions from low amounts of data through processing the information varies with the environment,but are difficult to process large-scale information.On the contrary,AI systems can process large-scale and specialized information but are not good at making accurate decisions from a low amount of data.Therefore,it is necessary to develop a human-machine interaction system to establish a mutually reinforcing relationship between humans and AI,to break the limitations of human cognition,and thus greatly improve the efficiency of complex problem-solving.To establish a mutually reinforcing relationship between humans and intelligent systems,it is necessary to develop a friendly human-machine interface and trust association at the cognitive level.When solving the problems of high-performance SMPSs design,it is necessary to establish a dialogue-based interaction model to construct trustworthy interdisciplinary integration,which is helpful for highperformance SMPSs designers to effectively overcome cognitive overload and achieve materials and electrical engineering interdisciplinary integration.Therefore,the designer and the intelligent assistant need to achieve unification at two levels.On one hand,at the AI tutoring level,the unification of human-computer communication mode using the human-computer dialogue approach.On the other hand,at the cognitive level,the unification of the two types of learning approaches,skills acquisition and adaptation,for interdisciplinary integration using the designer-centered design method.To achieve the unification of these two aspects,we focus our analysis on the following questions:(1)the contradiction between the need for the designer's cognitive skills in the ideal case of interdisciplinary high-performance SMPSs design and the lack of relevant skills in the actual design;(2)the contradiction between the need to list the causes of power supply failure by learning from errors and the difficulty of listing all possible mistakes for a complex system;(3)the contradiction between the need to obtain the actual state of the designer's cognitive skills to achieve personalized AI tutoring and the lack of actual information;(4)the contradiction between the human tendency to represent knowledge and problems in a diversify and unstructured way and the computer preference to express knowledge in a unique and structured way.This paper introduces techniques,such as psychometric methods,fuzzy cognitive maps,knowledge graphs,and natural language processing to address these issues.The main contributions of this dissertation can be summarized as follows.(1)We propose a cognitive diagnostic assessment method for power management experiments based on the DINA model.On the basis of the comprehensive power management evaluation structure,the designers' power management experimental cognitive diagnostic assessment results are obtained using the response data of the designers' cognitive ability assessment of the buck regulator and the DINA model.(2)We propose a Bayesian network model-based cognitive diagnostic assessment method for SMPSs magnetic components design.We construct a Bayesian networkbased cognitive diagnostic assessment model for inductor components using an evidence-centered design framework.Then,we calculate the conditional probability between variables using the effective theta method and graded response model.Finally,we use the Markov Chain Monte Carlo estimation method to get the posterior probability distribution of proficiency variables to obtain the evaluation results of designers' inductor design knowledge or skills based on their design data from power magnetics volume and weight reduction design task.(3)We propose an SMPSs design skills evaluation method using exploratory factor analysis.We firstly use multiple imputation to handle missing data.Then we obtain a four-factor model using parallel analysis,factor extraction,and factor rotation.Then we identify the four latent subskills as efficiency design,passive device design,power magnetics reduction design,and power economy design.The average of the factor scores from each of the imputed datasets and the SMPSs design constraints are combined to obtain the cut-off scores.Finally,we use the cut-off scores to evaluate designers' achievements on these subskills.(4)We propose an SMPSs design skills assessment method based on fuzzy cognitive maps(FCMs)and construct an FCMs-based SMPSs design system.The system incorporates both technical requirements and human factors in the SMPS design domain,which contains 69 components and 714 causal connections.The proposed system provides useful guidances in terms of knowledge or skills improvements for SMPS designers and can help improve the power supply design process.(5)Based on the above cognitive diagnostic assessment results in the SMPSs design,we introduce knowledge graph technology to construct the knowledge graph in the power supply domain and designers' digital twins.Then,we develop an intelligent question answering system for power supply design to provide personalized answers for designers based on their knowledge structure and skill level.