SBIR-STTR Award

We propose to study and construct an artificial intelligence program which can identify electronic malfunctions, then learn from its successes and failures to be more efficient, accurate, and expert. The program will draw on a database of general troubleshooting techniques, specific device knowledge, and experience with the system under test to direct a technician's debugging activities. The program will ask the technician focused questions about the system's status. The questions will require probing for voltages, logic levels, etc. The information generated will lead to more questions, utilimately resulting in recommendations for repair. Correct analytic conclusions will be incorporated into the program's memory, reducing the time to find the same problem thereafter. Technicians will be able to question wrong conclusions and to uncover the logic leading to the errors. Malfunctions located by program recommendations or by technicians hence increase the program's knowledge. Mistakes also lead to new knowledge and the formulation of new program rules within the knowledge base. The program amplifies and extends the technician's expertise. Overall, the project will model and elaborate AI programs now running successfully in other fields, and will utilize ADA for maximum systemic utility.

Keywords:
THIN FILMS ION BEAMS OPTICAL COATING DEPOSITION CLUSTERING STRUCTURAL PROP CLUSTER BEAMS IONIZED CLU

Phase II of Microexpert will create a full consulting type helicopter troubleshooting program. Phase I Microexpert is a prototype. It can act as informational backup for a practice technician. Or it can create the problem solving strategies, reasoning traces, and even hunches that improve the performance of more novice users. Microexpert trains its users and learns from its history. Through use, it becomes more time and cost effective. Phase I Microexpert proved the concept for a subsystem. It had limited reasoning, rule checking, and fault finding abilities. Phase ii will automate the reasoning process, add multiple symptom recognition/multiple causing handling capabilities, improve rule formation and manipulation, and generally progress in sophistication and breadth of application. It will create a maintenance expert shell into which specific system data can be inserted. The orientation of phase ii will remain highly practical. Microexpert will stay portable, fast, and friendly. Microexpert will train technicians while they are being productive, grow more efficient with age, retain troubleshooting expertise after the expert has gone, reduce fault isolation and reporting time, and generally improve the quality of field maintenance.