Applications of Artificial Intelligence in CAD Technology

A new feature to be found in modern CAD software releases is KBE (Knowledge Based Engineering) to support diagnosis, selection, and monitoring of tasks. KBE relies on capturing and storing experiential knowledge which includes proprietary design and manufacturing practices exercised during a product development cycle. KBE helps engineering companies to retain and preserve in-house knowledge and intellectual information.

A related technology which could significantly augment problem solving capabilities in CAD software is AI (Artificial Intelligence), which was introduced in the mid-1980s. The purpose of AI is to learn and replicate human problem solving capabilities. AI software, unlike procedural programming software (such as Fortran, C, C++ or Java), models non-numerical cognitive processes of pattern matching and decision-making. AI captures the expertise of experts who have the knowledge and experience to perform tasks which involve synthesis, diagnosis, planning, interpretation, and execution of design tasks. The language of AI software is therefore non procedural, but it is based on rules which determine how decisions are made.

Currently, the primary function of CAD software is to automate the analytical steps of a design. CAD software is used to create computer models of parts, to fit them together, and to model the performance of parts and assemblies so that they meet design specifications. The analytical steps of a design process are iterative, because design reviews, performed by experts, determine whether changes should be made (design synthesis). With AI-based tools, design synthesis can be performed directly without going through a separate design review and synthesis, because the knowledge and experience of experts is available in AI tools.

This article answers these questions:

• How could AI be merged with CAD?
• How does AI work as a rules-based programming language?

How Could AI Be Merged with CAD?

AI could function as a design automation tool for well-defined design tasks. The method by which AI is merged into CAD relies on MBR (Model Based Reasoning). MBR uses qualitative and quantitative simulation to predict interactions between connected components within a design assembly. Quantitative simulation produces unique results from analytical software such as Finite Element Analysis and from conventional CAD design software.

On the other hand, qualitative modeling cannot produce unique results, because the effectiveness of AI modeling depends on the quality of rules-based reasoning which resides in AI software. The rules and decision making are obtained from product design experts. Thus, AI functions as a replacement for a product design review team.

If knowledge-based reasoning and decision-making procedures obtained from product design experts are properly implemented in AI software, the product development cycle could be shortened considerably.

The mechanics of merging AI and CAD involve these essential features:

• Components for a product should be stored in a structured hierarchical form in which relationships between components are implemented in an objected oriented format
• Product structure is also stored in a structured hierarchical form. Links are provided between components or parts within the product structure, or links are determined by rules-based reasoning and methodology.
• Product behavior is deduced by a combination of quantitative and qualitative simulation.
• It should be easy to add new components or parts into the database, and it should not be difficult to add new knowledge-based rules and decision making procedures into the AI logical framework.

How Does AI Work as a Rules-Based Programming Language?

It is not within the scope of this article to explain the details of rules-based programming, but a summarized description is provided. The language of AI is designed to enable a computing machine to simulate human intelligence and reasoning. When all rules and constraints are satisfied, the rules “fire” and design decisions are made.

AI relies on “expert systems”, which simulates the knowledge and analytical skills of human experts. For example, to create AI software for a production process, several interviews are conducted to “milk” experts for their knowledge, reasoning and decision making when they are involved in building a successful product. The AI software is written to capture the knowledge, experience, logical reasoning, and decision-making procedures obtained from experts into a rules-based simulation software, which provides the qualitative simulation portion of a CAD-AI system.

Rather than explain the details or mechanics of AI programming, a few examples are given to show that useful results have been obtained from AI software implementations.

• AI software has been used in smartphones, in GPS navigational systems, and for 3D body-motion simulations in the Xbox 360.
• AI software was used by the IBM question answering system to defeat the two greatest champions of Jeopardy (a TV question-answer game show).
• Wolfram Alpha uses AI to provide an online service which answers questions from structured data.
• The Homeland Security and The United States Department of Defense use AI software called SEAS (Synthetic Environment for Analysis and Simulation) to predict and evaluate future events and to determine courses of action to take in response to probable events.
• The company Tampella Power Industries in Finland successfully used AI in a high-level design language called Design++ for the design of piping systems. The design of piping components was performed by the quantitative part of Design++, while the qualitative portion was performed by AI software. The AI simulation used design rules to make decisions concerning layout, connections between components, and for meeting specifications and design standards. Because each boiler system is different, AI software is written separately for each design implementation.

There are many other areas where AI is utilized. For example, projects involving neurology, machine learning, linguistics, gaming technology, driverless cars, and robotics use AI programming. It is reasonable to expect technologies which rely on human knowledge, intelligence and reasoning to rely heavily on AI programming.

Conclusions

Integration of AI and CAD can significantly speed up product development by incorporating the reasoning and decision making expertise of experienced product design engineers.

Because not all design projects are exactly the same, AI implementations cannot be generalized, but should be tailored to each project. For design or production processes which can use an integrated CAD-AI system, the investment involved in creating a CAD-AI system will provide rich rewards.

– IndiaCADworks