Research And Implementation Of Recommendation System Of Hoist Design Scheme

With the continuous development of computer,the rapid intelligent design of products has become the goal of modern enterprises,and therefore has become the research hotspot of scholars.However,the existing parameterized design system has some problems,such as low intelligence degree,poor universality and poor scalability.A product is composed of multiple components,the number of components,type,distribution,combination and its suppliers are in the process of product design plays an important role,so how to in a short period of time to get a to meet performance requirements,layout and reasonable cost as low as possible under the premise of solutions become the research direction in this paper.Papers on the analysis of the product rapid design,product layout optimization and the research status of combination of product suppliers,analyzed the importance and necessity of intelligent product design process,and the study direction of this a few related research results at home and abroad,combined with technologies such as deep learning and sort,in order to improve the efficiency of product design and make the rapidness intelligent product design process as the research direction,A product design proposal recommendation algorithm based on design scheme tree,deep learning technology and ranking learning technology was proposed.The main research work of this paper is as follows:This thesis proposes a design scheme tree model.Subject combining the actual application requirements of hoist products,using the principle of multi-way tree,combining the concept of product design and detail design,formed the principle and components design the mode of combining design scheme generates a tree structure,realize the systematization of design patterns,the intelligent design system,the diversification of design form.This design method can be applied not only to hoist,but also to other large and complex products.This thesis proposes a layout optimization method based on deep learning technology.Two dimensional layout diagrams of different layout forms are generated by summarizing the assembly rules of products,and then the typical classification model RESNET-50 of deep learning technology is used to classify the layout diagrams of different layout forms with the interference situation as the learning objective.Then according to the classification results and the priority of layout form,the optimal layout form of each scheme in the scheme set is determined and its area is calculated to generate a complete design scheme set with product attribute information and product layout information.This method avoids the traditional layout problems,such as complex model establishment and large amount of calculation,and simplifies the process of layout optimization from a new perspective.This thesis proposes an optimization method of component supplier combination based on learning-to-rank is proposed.The learning-to-rank algorithm commonly used in information retrieval is applied to the problem of product supplier combination optimization.According to users’ preferences for parts supplier information,the feedback behavior of users on the list of recommended product design schemes sorted by AHP-TOPSIS algorithm was collected to generate the corresponding training set data.Four kinds of ranking learning algorithms,Rank SVM,Rank Net,List Net and Coordinate Ascent method,were used to optimize the list of recommended schemes.Taking NDCG@10 as the evaluation index,it is determined that Rank SVM algorithm,compared with other algorithms,makes the reordered recommendation list more in line with user needs,reduces the page browsing time and improves the decision-making efficiency of the design scheme.Finally,taking the hoist as the experimental object,based on the above technologies and methods,combined with software development and database application technologies,a product design scheme prototype recommendation system based on design scheme tree,deep learning technology and ranking learning technology was developed based on Visual Studio and SQL Sever platforms.The feasibility verification and preliminary application of the above research results verify that the proposed method has higher efficiency and better recommendation effect than the traditional parametric design method.And the main content of the article is summarized and the prospect is forecasted.