【作者】 刘彬；

【导师】 潘复生；

【作者基本信息】 重庆大学 ， 材料科学与工程， 2011， 博士

【摘要】 镁及镁合金由于具有密度低、比强度高、电磁屏蔽性好、弹性模量高等一系列优点,因此受到极大关注,被誉为“21世纪的绿色工程金属结构材料”。但目前镁合金的研究和生产中一直存在不少困难。而由于材料研究及其制备加工过程的影响因素非常复杂且具有很大的不确定性,材料研发采用的是依赖大量实验,大面积筛选的方法,消耗巨大。为提高传统的实验研究效率,开拓新的研究方法,加快镁合金研究进展,利用多学科交叉融合、计算机辅助设计、引入人工智能技术等方法已成为镁合金研究的新趋势。作为人工智能应用研究最活跃和最广泛的专家系统,其善于处理强非线性关系,非常适用于研究及制备加工过程复杂且需要丰富经验的材料领域。本文基于对目前镁合金研究和生产的需求分析,设计并开发了一个镁合金专家系统,其主要包括核心知识库、数据库、人机交互界面及数据交互机制。鉴于镁合金发展至今已有较为丰富的数据资源,本课题采用基于数据的方法,应用人工神经网络技术,并采用新的优选人工神经网络建模参数的方法——参数全排列组合训练,建立了精度较高的变形镁合金抗拉强度、屈服强度和延伸率预测的网络模型、铸造镁合金抗拉强度、屈服强度和延伸率预测的网络模型,以及铸造镁合金晶粒尺寸预测的网络模型。这些人工神经网络模型组成了镁合金专家系统的核心知识库。基于这些网络模型,可以充分挖掘现有镁合金数据潜力,实现对镁合金性能预测以及对铸造镁合金晶粒尺寸预测功能。根据对镁合金数据需求的分析及系统设计,采用数据库管理技术,在完成数据库概念结构、逻辑结构基础上,建立了包括镁合金选材数据、知识库信息数据、用户信息数据在内的系统数据库,实现了对系统数据的管理和对镁合金的初步选材功能。在完成系统知识库和数据库构架的基础上,根据系统整体设计目标和模块化的功能设计,采用面向对象的程序设计语言Visual C++开发了镁合金专家系统的人机交互界面。通过采用Matlab计算引擎,以Visual C++和Matlab的混合编程,实现了镁合金专家系统的人机交互界面和知识库模型的耦合;采用Visual C++提供的数据库访问方式ADO作为系统数据库与人机交互界面的连接方式。由这两种方式实现了镁合金专家系统的数据交互。已实现的镁合金专家系统运行稳定,人机交互友好,使用灵活方便,已达到设计预期。系统应用实例表明其对性能和晶粒尺寸的预测值与实验值符合较好,误差在可接受范围内,可以初步为镁合金相关研究和生产提供参考和支持。更多还原

【Abstract】 Magnesium and magnesium alloys have been of tremendous concern and been praised as the "metal materials of the green engineering in the 21st century" due to they have a number of advantages, such as low density, high strength to weight ratio, good electromagnetic shielding characteristics, high elastic modulus and so on. But there still exist many difficulties in the research and production of magnesium alloys. Because of the great complexity and uncertainty of factors in the research and the process of materials, large quantities of experiments and large scale screening have been used for the research of materials, which result in very huge consumption. To improve the efficiency of traditional experimental research, develop new methods and speed up the research progress of magnesium alloys, it has become a new trend that the use of interdisciplinary integration, computer aided design and artificial intelligence in the research of magnesium alloys. As the most active and widest field in artificial intelligence research for applications, expert system is good at dealing with strongly nonlinear relationship. Thus, it is very suitable for the material research areas, where study and process are complex and require extensive experience.In this paper, based on the needs analysis of current research and production of magnesium alloys, a magnesium alloy expert system has been designed and developed, which mainly includes the core knowledge base, database, human-computer interface and data exchange mechanism.In view of the rich data resources with the development of magnesium alloys, the data-based method has been used in this subject. And by using artificial neural network and a new method for optimizing modeling parameters of artificial neural network—all permutations and combinations training of parameters, the network models with relatively high precision have been built, which are the predicted ultimate tensile strength/yield strength/elongation model of wrought magnesium alloys, the predicted ultimate tensile strength/yield strength/elongation model of cast magnesium alloys and the predicted grain size model of cast magnesium alloys. Then, the core knowledge base of magnesium alloy expert system has been formed by all of these artificial neural network models. Based on these network models, the mechanical properties of magnesium alloys and the grain size of cast magnesium alloys can be predicted through exploiting the potential of existing data of magnesium alloys fully. According to the needs analysis of magnesium alloy data and system design, after completion of the conceptual and logical structure of database, the system database has been established by using database management technology. The system database includes data for selecting magnesium alloys, knowledge base information data and user information data. So, the management of system data and the initial selection of magnesium alloys can be carried out.Based on the completion of system knowledge base and database, then according to the goal of overall system design and the modular functional design, a human-computer interface of magnesium alloy expert system has been developed by using the object-oriented programming language Visual C++.Through Matlab engine, the coupling has been achieved by using mixed programming of Visual C++ and Matlab, which is between the human-computer interface of magnesium alloy expert system and the models in knowledge base. The ADO, a method of accessing database in Visual C++, has been used as the connection between the system database and the human-computer interface of magnesium alloy expert system. Therefore, the data exchange of magnesium alloy expert system has been accomplished by the two methods.The established magnesium alloy expert system has met the design expectations, which is stable during operation, has friend man-computer exchange, and can be used flexibly and conveniently. System applications indicate that the predictive values of mechanical properties or grain sizes agree with the experimental ones well and the errors were in the acceptable range as well. Therefore, the magnesium alloy expert system can provide reference and support for the research and production of magnesium alloy preliminary.更多还原