Casting troubleshooting to be available through via Internet.

The AFS Molding Methods & Materials Div.'s Basic Concepts Committee (4-E) is developing an expert system for 21st century casting defect analysis.

The role of computer-mediated communication continues to grow within the foundry industry. Web sites and e-mail have become invaluable tools for organizations that need to communicate rapidly across the globe and receive information in real time. As the industry turns toward the 21st century, such new communication tools are arriving to try to share the spotlight.

At the forefront of advances for computer-mediated communication in the foundry industry is the expert system. Following in the footsteps of other industries such as medical diagnosis, the increasing demand for information in the foundry has spurred the development of these industry-wide resource systems for organizations in search of answers.

Expert Systems

Computer-based expert systems are a form of artificial intelligence that combine the knowledge and experience of industry experts into one source, able to perform diagnoses for a variety of problems that could previously be performed only by the industry experts themselves.

In practice, these systems work by gathering process and procedure information from the user to form a decision tree (Fig. 1)--where the action at any particular point on the tree is a result of the history of decisions above the point. Then, using the expert information it has already acquired, the system compares and analyzes the situation to pinpoint the user's breakdown in correct process or procedure.

[Figure 1 ILLUSTRATION OMITTED]

For manufacturing industries in general, this approach to problem solving has proven to be highly effective. For the foundry industry, this tool has remained at the back of the pack. The following is a brief description of some of the problem-solving expert systems that have been developed for foundries as reported by the AFS Engineering Div.'s Expert Systems Committee (1-G) in 1993:

* Casting Reject Elimination--A system was developed for the rapid elimination of casting defects and prevention of their recurrence. The methodology consisted of five steps: defect classification, defect cause enumeration, defect cause determination, defect correction and process monitoring of defect remedy for defect removal. The system implemented a defect coding and classification system, cause-effect diagrams to enumerate the various causes and to assist in the cause determination, and the use of process quality control systems to monitor the proposed remedies.

* Aluminum Alloy Design--This was developed as a knowledge-based system to aid aluminum alloy design for aerospace applications. It aids in reducing the number of iterations or shortening the average cycle time in the search for a suitable alloy design.

* Microstructural Analysis--A system was developed to interpret simulations of Ti-6Al-4V thermomechanical joining. The system includes a history scanner to reduce the data files and a metallurgical knowledge base to predict the part's microstructure and properties. The program also includes a library of micrographs in the knowledge base and an explanation module to assist the user in understanding its predictions.

* Casting Defects--A basic casting defects analysis system, called DEFCHAR, was developed for the Indian foundry industry by the Indian Institute of Technology. It is an expert system that involves rules to identify the defect by virtue of its appearance. It consists of three levels of defect characterization that leads to the exact identification of the defect.

Although these systems weren't accepted on a broad scale, their ability to provide expert information on demand was. This was the basis for the AFS Molding Methods & Materials Div.'s Basic Concepts Committee's (4-E) development of an industry-wide Expert System for Analyzing Casting Defects.

Analyzing Casting Defects

The development of the Expert System for Analyzing Casting Defects is based on two established systems--Mycin [an expert system developed by the American Medical Assn. (AMA) for the diagnosis of diseases] and the International Atlas of Casting Defects, published by AFS.

The system follows a rule-based, diagnosis tree design (as shown in the Analysis of a Casting Defect sidebar) that questions the user about its production procedure and problem casting. Each answer results in a different question from the system, until the system has enough information and is able to narrow down a diagnosis for the defect. At this point the system will present the diagnosis to the user with an image of the defect for comparison purposes. These images are part of an important difference between this system and its ancestors--it will be a full casting defects reference tool available via the Internet.

The foundry industry and, specifically, casting defect analysis are perfect subjects for this expert system, according to Dr. H. Md. Roshan, Maynard Steel Casting Co., Milwaukee, the principal developer of the defects expert system. Both are procedure-oriented, which allows the expert system to perform an extensive step-by-step analysis of processes. As with the disease diagnosis system from the AMA, the expert system begins with an ideal situation, whether it is a body or manufacturing line, and analyzes the user's situation to determine at what point the user's situation differs from ideal--the breakdown point.

In addition, an expert system for defect analysis concentrates, from a revenue standpoint, on a major concern for foundries.

"For the foundry industry, defective castings can cost a foundry from 3-20% of its revenue," said Dr. Roshan. "Individual foundries invest a large amount of money in maintaining quality. The focus of the expert system is to provide these foundries with the ability to define these revenue-decreasing defects using the industry's expertise as a resource."

But the casting defects expert system holds other advantages over its predecessors, according to Roshan. The lack of support for the previous expert systems may be due to three factors, he said. The first is a lack of focus by the industry on computer-based communications. The second is the industry's general lack of skills in knowledge engineering. And the third reason may be the overall lack of appropriate shell programs to build a usable and understandable expert system.

Roshan envisions the expert system as an integral part of daily production.

"The Casting Defects Expert System is to serve as a tool to the quality assurance departments of the foundries. The defect characterization module is a tool to the inspectors, and the diagnosis module is to provide a resource to identify the cause and encourage foundrymen to appreciate the importance of collecting relevant data during the production of the casting."

The committee's first goal with the expert system was to develop the information resource. Twenty-five different experts were solicited from across the globe for all types of metals and casting processes to provide the information and images to be used as the basis for the decision trees. The resource developed through this solicitation will be able to provide an unparalleled diagnosis drawn from the experiences of foundrymen worldwide. The expert system preserves the knowledge of the foundrymen of today and yesterday, which would be lost if not documented, for the foundrymen of tomorrow.

The second goal of the committee is the creation of a standard casting defect classification system through the use of the International Atlas. Although the International Atlas is available, its ability to characterize the defects in the individual foundries is not wide-spread.

According to Roshan, there is a need to popularize the characterization of defects in the foundries so that when experts provide information specific to the defect codes in the International Atlas, individual foundries can take advantage of the information. In addition, when a foundry needs to discuss a defect outside its organization, a universal characterization code will deter an identification problem from arising. By using the classification of the International Atlas, the expert system can begin to establish a universal classification system and save the man-hours spent on scrap meetings looking at defective castings.

Lastly, with the development of this technological tool, the foundry industry can enhance its metalcasting community further with industry-wide resources available to share the wealth of knowledge and experiences of foundrymen from across the globe.

"The expert system represents a community of resources for the industry," said Roshan. "The Internet is a powerful information tool that will provide this resource for all of the foundry industry."

But this powerful tool also will be the greatest obstacle to the success of foundry expert systems, said Roshan. Their reliance on computer-based communication as its medium will alienate organizations not accustomed to the expanded needs of the 21st century. But with the possibility of other expert systems developing for gating and risering, and material selection, the Internet will only become more significant.

"The people have to fully embrace the Internet and computer-based communication technology," said Roshan. "Hopefully, the expert system will become the universal and powerful tool we envision because people will become more aware of it and in turn will integrate it into their daily production."

Plans are for the completed expert system to be available to users by the end of this year. Currently, the details of this AFS project with examples of the system are in place on the web at http:// www.4ebasic.com. Foundrymen can also exchange information related to the casting defects at the following web site: http://www.4ebasic.com/messager.

RELATED ARTICLE: Analysis of a Casting Defect

Sometimes even the best foundrymen are stumped by casting defects. The following scenario may be familiar to your foundry, but, with a new solution:

The gray iron housings your foundry has been running on its green and sand line have developed sand inclusions. You computer modeled this casting from the start and can't understand where the process is breaking down. Since the project is your baby, the job of troubleshooting the casting has fallen into your lap.

After an initial examination of procedures, which turns up nothing, you decide to trust technology and use the new ExpertSystem to Analyze Casting Defects. You load up the program and click on the defect classification for your castings--G-Inclusions or Structural Anomalies--and the defect code--G-131 Sand Inclusions. The expert system begins to ask you questions:

yes

Is there any evidence of poor ramming or soft spots in the molds?

Is there any evidence of leakage during pouring?

no

Is there any evidence of improper gating practice such as metal impinging at high velocities or directly on a core?

Is there any evidence of improper mold closing leading to crush?

Is there any evidence to improper mold handling after the mold is closed?

Diagnosis: Sand inclusions are caused by a lack of care during molding.