0 introduction
Typically, the finite element analysis software is only displayed on the screen in a graphic or form in the post-processing, and does not perform system finishing and refining on the calculation result. During the repeated design process based on finite element calculations, designers tend to understand the design of defects and inappropriately, without properly accepting language characters by calculating the results maps or forms, without properly accepting defects or inappropriate and generated reasons When designing, design and analysis is often separate when designing complex project, designers may cause misunderstandings and deviations when understanding finite element calculation results, which will cause design modifications to the wrong direction. At the same time, in the design automation system based on finite element analysis, the calculation results of finite element must participate in knowledge-based reasoning process, and the calculation results are extracted as a knowledge source to participate in the design optimization process. Therefore, it is necessary to study how to organize and refine finite element analysis results to direct knowledge and text that can be given and reading. Currently, finite element system software (such as Marc, Nastran, ANSYS, MOLDFLOW, etc.) has developed to this extent today, and there is no finishing and refining of analysis results on the post-processing, mainly involving the following points: ( 1) There are too many specific applications facing. Each specific application requires a specific interpretation and processing, which is uncomfortable in general finite element system software, and a feasible method is targeted secondary development according to specific industries and applications. (2) System integration has not yet been developed to all relevant states, ie all relevant integration. The so-called full-correlation integration means that each sub-function module under large systems achieves a tight combination of functional and data management. The unsatisfactory integration of existing system software hinders the feedback path of finite element analysis results, making "analysis → design" to produce a degree of detachment. (3) The development and application of intelligent technology. To meet the full-correlation integration of the system, you must have support for intelligent technology. Intelligent technology does not only control the role of coordinated operation of each module, but also play a role in decision-making on the judgment problem, and the role of assistance in specific issues. However, in the current situation, "Design - Analysis" system integrated software has a very insufficient support of intelligent technology, leading to serious separation of the two-link in the reality. This article is a study on the knowledge of finite element results under the research topics integrating the characteristic modeling and finite element analysis system, and mainly combined with the complex field of plastic forming.
1 Implementation and method
During the knowledge of finite element calculation results, the following aspects are required: (1) First explicitly analyze the type. The analysis type can be divided into linear and nonlinear two categories from the whole. Secondly, it can continue to be subdivided into structural analysis (including vibration analysis, stable analysis, etc.), electromagnetic analysis, fluid analysis, thermal conductivity analysis, etc. Due to the different application and analysis purposes, the analysis type is different. The analysis type is different, and the method of finishing and refining the calculation results is different. For general linear elastic analysis, the key knowledge is the state and location of the unit. You know the state and location of the unit, you can evaluate the rationality of the design. As in the analysis of the beam structure, if the beam structure is invalid from the finite element analysis, the expression of the knowledge can be: the beam structure is invalid in a certain position. For nonlinear analysis, especially large deformation analysis in plastic forming, only the final analysis of unit state and location must be recorded, and the experience of analysis must be recorded, so that the source of failure and the history of evolution of the failure are found during failness. Therefore, different analysis types should adopt different sorting and refining methods, thereby different in terms of knowledge and treatment. The analysis type includes analyzing objects, analytical purposes and analysis methods, which indicate the environment of the analysis and the results that will reach, thus launching the finite elementary knowledge of finite element results. In the research development of system software, the interface of the analysis type is specified, and the analysis type can be visually specified by the graphical interface, and the data processing and storage hidden are automatically performed after the graphical interface. (2) Traverse all finite element units, understand the state of each unit, finish the overall calculation results. In general, the most concerned is to design the state of the unit in the critical area, and the inspection unit is in the failure state. If there is a failure of the unit, the results of the check can be sorted as the results of the analysis object, which can be invalidated at some point in the analysis object. The form of failure is different from the application. In the field of plastic forming, failure can be ruptured, wrinkled, thinned, dimensional deviation, poor shape. In terms of feature-based technologies, since the object-oriented technique is used, the result of the cell traversal analysis is transformed into the evaluation of the feature. For example, there is a restriction value of the adjacent 30 finite units in the feature 犃 exceeds the permissible limit strain value, and the 30 finite units on the feature are invalid. Attributed the failure of the characteristic A, That is, "Feature A is invalid". In order to alleviate the workload of all units, "Key Region Scanning Method" can be used, i.e., only the state of the key area or characteristic, the state of the characteristic is analyzed. How to determine critical areas or features, one method is designated by designers or analysts prior to analysis, and another method is to determine intelligent technology or some reasonable computational judgment. The first method is preferred in this system. In the application interface, two unit state scanning methods are provided: (a) a comprehensive scan; (b) key area scanning. In "Key Area Scan", it is divided into two ways: a) Manually specify critical areas; b) Automatically determine critical areas. (3) Understand the history of failure or defect development, and recognize the cause of failure. In many non-wire finite element analysis, when the failure or defect occurs, it is often traced back to the history of failure or defects, especially when the failure or defect is generated, and the state of the various parts of the object is analyzed, and the reason for the failure or defect And make targeted modification design. In nonlinear deformation analysis, you want to understand the history of analytical object deformation, it is necessary to traverse the calculation results of each loading step size of the deformation analysis, that is, the intermediate result of the calculation.
By organizing and analyzing the intermediate calculation results, you can understand the history and causes of failure. In this step, the processing mode based on feature or critical areas is used, that is, on the basis of the second step unit state scan, since the area where the failure or defect occurs, it is traced back or these fail or defects. When the roots and changes in the history, only the changes of these features or regions are concerned, without having to care about those status or areas. Thereby, according to the "material point status tracking method" or "deformed mesh method" in finite element, "feature status tracking method" based on feature or critical areas is proposed. In the "Feature Status Tracking Law", track the change history of "problematic" feature, thereby achieving an understanding of the status of the feature state. However, in some analysis, since the characteristics are interrelated, the state change of a certain feature involves the influence of other features, so sometimes the state of the feature is not analyzed in the history of the characteristics. Learn. Therefore, in this analysis, "state graph scanning by key feature" is adopted, key feature is the feature of analysis, or characteristics of failure. First, the state of the key feature is scanned, and then the state of the adjacent feature is scanned. In plastic forming, changes in materials are severe nonlinear, namely, nonlinear, nonlinear materials, and nonlinearity, which are mainly manifested as rupture and wrinkles. For these two failure forms, they are more concerned about the status of materials when they just occur. At this time, the system should be able to explain the flow trend of the material in this state and the state of each portion of the material. (4) Knowledge formal expression of finite element calculation results. The formal expression of knowledge is a key step in the conversion of finite element analysis to knowledge. In the above explanation, the method of processing the data is specifically described. However, there is data and its processing methods, the system does not directly expose it to the form of knowledge, this requires the conversion of "data → knowledge". "Data → Knowledge" Transformation requires consideration of the method of conversion, the provisions of the conversion knowledge, and the problem of knowledge and reliability of knowledge. According to the analysis, "Data → Knowledge" transformation can have two methods: (a) "Data → Knowledge" transformation method based on the connectionism method. The principle of the couplingism is mainly the connection mechanism and learning algorithm between neural networks and neural networks. The transformation of "data → knowledge" is used to use the neural network, and the reasoning process is implied in a simple weight calculation, which is a direct jumping transformation form. The entrance to the neural network is a digital representation, which is prior to "data" inlet portion in "data → knowledge"; the exit of the neural network is also a digital representation, which must be interpreted as a corresponding form by a predefined explanation. Knowledge, because the predefined explanation is human set, so explaining the conversion process is just a simple map operation. Therefore, "data → knowledge" in "data → knowledge" can be expressed in the "Data → Knowledge" transformation based on neural networks. There have been a large number of research and work to achieve the conversion of "data → knowledge" in neural network technology, but have not been specifically proposed or noted. During the conversion of "data → knowledge", "data" is accurate, but the corresponding "knowledge" does not necessarily expand it, so the fuzzy principle introduces the "data → knowledge" transformation process. In this study, "data → knowledge" conversion based on fuzzy neural network is to obtain some qualitative descriptions, which play a modified design of the auxiliary designer in practical applications, and often do not Requires a precise digital description. (B) "Data → Knowledge" conversion method based on symbolism (Symbolism). The principle of symbolism is mainly the physical symbol system (ie symbolic operating system) hypothesis and limited reasonable principles. In the system, develop rules-based "data → knowledge" transformation methods.
The rules use "if ... then ...". This knowledge conversion method is an indirect process conversion method relative to a neural network method, and the final result is derived step by step according to logic reference. The IF portion of the rule is the description of the data, which can be used to compare the data of the data to compare the comparison of data, and some algorithms can be used to calculate the population changes trend. The THEN part of the rules is the knowledge part, which is partially described to describe the knowledge effect of data generated. In the specific application, if the actual situation meets the data description of the IF portion, it will be the language interpretation of the THEN. IF (Data Analysis), THEN (Knowledge Representation), with "Data → Knowledge" conversion method, which will be refined in data. It is well known that the finite element analysis result data is very large, so all data cannot be used as data inlet or IF of the neural network, and these data must be refined to obtain critical data for design guiding significance. Infining data, in the system, "Question-Answer" is used in the system. "Questions" is not directly emitted to the system, but is a targeted problem in a targeted problem based on summary practices and multiple finite element simulation results. These "Questions" is stored in the system's knowledge base, and the system performs finite element calculation results based on these "Questions" one by run, thereby obtaining targeted data analysis. For example, "Question 1" --- Whether the workpiece is broken (or whether there is a certain unit of equivalent to the strength of the material)? Thereby, the system starts the unit scanner, and the state of each unit is determined. According to the "question", the system is handled, so that "answer" is obtained. There are only two ways to "answer": Yes, no (no). According to "Question - Answer", the system refines the huge finite element calculation results, finds the key, meaningful results. Rule-based knowledge conversion is closely related to "Question - Answer". "Answer" answer the conditional judgment in the IF, if "Yes" is (YES), execute the knowledge specified by then; if "answer" is "No (no), no THEN section. In the THEN section, the specified knowledge represents "Question" in "Question - Answer", but the1n part is the affirmation of knowledge, and "Question" is a question of knowledge. As in the following example: Provisions: If the equivalent stress of the (IF) unit is greater than the stress strength of the material, the material is broken. Question: Is there any equivalent pressure of the unit to be greater than the stress strength of the material? Answer: Yes (YES) / No (NO). Finally, the key to the problem is focused on the source and reliability of knowledge. The content that knows rules and "Questions" is pre-prescribed, so how to specify these rules and "Questions" in advance is important, it directly related to the quality of the system for financing of finite element calculation results. First, a large number of practices work, and summarize general and targeted knowledge on this basis, put it into the system, and test and modify it in the repetite operation of the system; secondly in the use of the system, constantly add new knowledge. In the plastic forming of the sheet, there is a large number of literatures that have been analyzed, and the knowledge specified from which the knowledge will be met to meet the general description and explanation of the system. Since the system is based on feature, the representation of the characteristics is also introduced in the knowledge specification. The characteristic knowledge regulations can be connected to the characteristic design, thereby guiding the design and modification of the feature.
2 Connect the language According to the needs, the knowledge of finite element calculation results is studied. In practical applications, the program is used to design the above principles. On the basis of plastic forming finite element simulation, the series of consistent functions are achieved on the basis of defects → defects → defects from defects → defective prevention measures for defects. "Analysis → Design" information feedback to make up for the phenomenon of existing CAD and CAE system, while providing better and more accurate knowledge of CAPP process design.
---------------------------------------------- 公,, honesty, Yong, Yi Xuexuan, Arctic brilliant; Ze was sent by Wanfang, Xiao Xiangxiang; 哉 府 府, 文; 生 教, 国 强; 仞 仞 仞;;;;; 强 强 强Director of the work;