This paper holds that there are two necessary conditions for innovative mechanical design: one is to fully acquire applicable knowledge; The second is to use a design system that conforms to and can stimulate innovative design thinking. The design process is full of contradictions, and the acquired knowledge should help to solve the contradictions quickly, which requires the close combination of knowledge acquisition tools and design process, so it is necessary to study the knowledge acquisition tools and design system in a unified way. In addition, human innovative design thinking mode is formed by summing up long-term successful design experience, so the design system must conform to the law of innovative design thinking. The thinking law of innovative design should be the theoretical basis of computer-aided innovative design system.
Based on the above considerations, this paper starts with the research of innovative design thinking, integrates knowledge acquisition methods, studies innovative design theory, and then develops innovative design system for mechanical products.
1 thinking law of mechanical innovative design
We often call the process of thinking "thinking" because we can use the path problem to explain the process of human thinking. This paper puts forward two thinking principles of mechanical innovative design:
The first is the shortest path principle. After getting the functional requirements of products, designers often retrieve the best design examples first, so as to get close to the goal most quickly. Then, by using the method of value engineering, they find out several parts with lower value as the research objects, and then analyze the contradictions of the obtained objects, trying to make the smallest changes to solve the contradictions. If the contradiction is not resolved, they plan to make greater changes or expand the scope of the research object, and finally get the best results. This method consumes the least energy and embodies the shortest path principle.
The second is similar association. According to Hideki Yukawa's identity theory, associative ability is the creativity to find out the similarities between things, and similarity refers to the internal relations between things.
It is very difficult to use computer system to assist designers to discover the similarity of different things from nature, so this paper only studies similarity mining from mechanical product examples to promote mechanical innovative design.
The mechanical design process is a mapping process from functional requirements to action principles to physical structures [1]. In CBR system, functional requirements, action principle and physical structure can all be used as instance indexes, so they can be collectively referred to as index items. The association between different indicators of the same indicator can be called vertical association, and the association between different indicators of the same indicator can be called horizontal association.
The basis for judging whether Lenovo is reasonable is similarity, which is determined by the existing product examples. For example, the product example of ultrasonic grinder makes the action principle of ultrasonic vibration and the functional requirements of grinding have longitudinal internal relations; For example, many product examples can meet the same functional requirements, so their functional principles and physical structures are similar.
Functional requirements are the starting point of association. Experienced designers usually remember a large number of design examples, so they can grasp the similarity between the vertical and horizontal directions, so that they can quickly associate the horizontal and vertical directions, draw examples with similar action principles and physical structures (referred to as similar examples for short), make combinatorial optimization, and finally get the optimal solution.
These two principles have been unconsciously adopted by many design methods. Case-based reasoning can not only approach the optimal solution quickly, but also embody the shortest path principle. The material field analysis method (TRIZ) analyzes millions of design examples, and determines the internal relationship among functional requirements, action principles and physical carriers, as well as the substitution relationship between different action principles or physical carriers, so that designers can find suitable action principles and physical carriers according to functional requirements and embody the principle of similar association.
2 Computer Aided Innovative Design System
The design of computer-aided innovative design system fully embodies two innovative design thinking principles, and the system also adopts a variety of innovative design methods and artificial intelligence technology. The flow chart of computer-aided innovative design system is shown in figure 1, including the following key technologies:
2. 1 instance retrieval
When using case-based reasoning (CBR) technology, we should first study its advantages and disadvantages. CBR is an example-based knowledge providing method. At present, there are still the following shortcomings: first, in order to achieve practicality, the system usually establishes a huge case base, which leads to difficulties in management and low efficiency of the system; Secondly, only one or a few examples are obtained through retrieval, and other examples that do not meet the retrieval requirements but contain applicable knowledge are not utilized, which is not enough to support innovation; Finally, case adjustment relies heavily on domain knowledge and is difficult, so many CBR systems are simplified as case retrieval systems [2]. The deep reason of these three shortcomings is that the examples are independent of each other, and the knowledge contained in different examples is difficult to be combined and utilized. In order to overcome this contradiction, this paper proposes to find similar examples through similarity association, and optimize the combination by genetic algorithm to realize the reuse of case knowledge.
The case retrieval function of the system is realized by the product structure and configuration management function and search function in IMAN, a commercial PDM system. The visual representation and management of cases depend on IMAN's product structure tree function.
2.2 Visual example model expression and contradiction analysis
The development direction of conceptual design technology is to study the unified expression method of design scheme [3]. Literature [4] extends the FBS diagram proposed by Japanese scholar Yoshikawa Hiroyuki, and describes the functional layer and structural layer of a design scheme with two frames respectively, and stores the corresponding relationship between functional units and structural units, so that computers can understand the structure and function of products. The disadvantage of this method is that the relationship between structure and function is not intuitive enough. Therefore, this system adds a functional relationship diagram to the functional hierarchy diagram and the structural hierarchy diagram, and describes the structure and its functional relationship in the way of semantic network, so that the structure and function are in the same diagram, and the designer can intuitively understand the product principle, and according to the functional relationship diagram, use the value engineering method to analyze the contradictions existing in the example.
The key to innovation is to correctly analyze the contradictions in products [5]. The basic contradiction of product design is that the ratio of product function to cost can not meet the requirements of users. This contradiction has two manifestations: one is that some product function quality goals have not been achieved; Second, the quality of some functions has been improved, while the quality of some functions has deteriorated.
The results of contradiction analysis are used to guide the association of new action principle and new physical structure, and then similar examples are found.
2.3 Web-based knowledge base of innovative design
The innovative design knowledge base of the system includes action principle base, physical structure base and case base. When the system searches for a new action principle or physical structure according to the similarity, it automatically calls up the corresponding example.
The development of action principle library and physical structure library draws lessons from the achievements of TRIZ, and then supplements and sorts out more than 240 action principles (including more than 50 basic measures) in the mechanical field. Under each action principle, various physical structures are stored separately to form a physical structure library. The case base is mainly developed for several common household appliances.
Innovative design knowledge base is the core component of innovative design system, and it is a WEB text knowledge base. Mark the text with the mechanical knowledge XML developed by the author, so that the knowledge base can be established on the international standard XML text, so as to realize the sharing of knowledge resources in different places, and establish a computer-aided innovative design system of mechanical products based on WEB on this knowledge base to meet the needs of collaborative design in different places.
2.4 Quantitative method of similarity and improved genetic algorithm
Each product has a different structure and needs different genetic algorithm coding. In order to improve the operation efficiency, this system adopts floating-point coding.
In the traditional genetic algorithm, the initial population is generated by random method [6], which has certain blindness. Therefore, this paper proposes to use the principle of action of examples or the similarity of physical structure as the basis for screening examples to produce initial population.
The key to realize this method is the quantification of similarity, that is, the calculation method of similarity. The essence of similarity is the relevant knowledge of examples, and it must be mined in the example set through a certain algorithm. The similarity of vertical correlation is essentially the degree of correlation between functional goals and means of realization, while the similarity of horizontal correlation is essentially the degree of substitutability of means of realization. Higher similarity means more support from existing product examples. Screening the initial population according to similarity is equivalent to using the previous design experience, which gives the initial population a reasonable basis, so it can speed up the convergence of genetic algorithm. Based on the principle of similarity correlation, this paper puts forward the following similarity calculation methods for vertical and horizontal correlation.
Let the product instance set be C, the functional element set be F, and the action principle or physical structure element set be G, respectively marked as: C={Ci|i= 1, 2, …, n }；; F={Fj|j= 1，2，…，m }； G={Gk|k= 1, 2, …, q}. The instances Ci in the instance set belong to Fj and Gk respectively, and have different membership degrees uij and uik. Let the longitudinal correlation similarity from element Gk to element Fj be rkj, then:
rkj =
Let there be elements Gk and Gm in G space. Example Cji belongs to elements Gk and Gm, the membership degrees are uik and uim respectively, and the horizontal correlation similarity from Gk to Gm is rkm, then:
rkm =
Membership is stored as a property of the instance object. According to the above algorithm, the system mines similar knowledge from case sets, assists designers to associate from the direction of high similarity, and is used to guide the generation of initial population of genetic algorithm, thus promoting design innovation.
3 Conclusion
This paper studies the thinking law of innovative design and uses it to guide the development of innovative design system for mechanical products. The successful application of the system proves the correctness of judging the thinking law of innovative design and the feasibility of various new technologies. Through the analysis and correlation of contradictions, the system finds the applicable principle of action, measures, physical structure and examples to solve contradictions, and completes the function optimization and principle optimization in the conceptual design stage, which is a new achievement to realize the generalized optimization design method of machinery.