Monitoring, Diagnosis & Control of Manufacturing Processes
Manufacturing process monitoring is the activity of identifying characteristic changes of a process without interrupting the normal operation by evaluating the process signatures. Process monitoring usually precedes diagnosis and control activities. Diagnosis refers to the identification of causes for the change or failure of the process. On the other hand, control refers to the predetermined or corrective action necessary to ensure the desired process behavior to maintain workpiece quality, high production rate and close dimensional tolerances.
This research is concerned with the development of generic methodologies for sensor-assisted intelligent integrated monitoring and diagnosis of manufacturing processes and machine tools. These methodologies emphasis on investigating sensor data representation schemes, sensor data fusion techniques, and neural network models, artificial reasoning methods for a variety of industrial applications.
We have developed a methodology for on-line flank wear estimation in turning processes. The results indicate that the methods investigated provide accurate on-line flank wear estimation. Our current plan is to extend the methodology for on-line tool wear estimation in milling and drilling, and in-process workpiece surface roughness assessment in turning, milling, and drilling. Eventually we plan to develop control strategies for assuring workpiece dimensional tolerances and surface finish requirements.
Design and Manufacturing in Mass Customization
As markets reach their saturation limits for many products, and customers grow more demanding, manufacturing industries are forced to enter into the production paradigm of mass customization in which products are designed, manufactured, tested and delivered as per the customers' exclusive requirements. This kind of concentration on the individual customer is taking place in consumer products, automobiles, telecommunications, computer hardware, and a wide range of other products.
We have made good progress into identifying the principles of mass customization and the different factors that influence this customer-oriented business strategy. We have developed a typology of mass customization systems and identified different technologies useful for supporting them. We have also been developing functional and information models of mass customization systems to address the issues related to cost implications, operations planning, systems integration, and information management in mass customization environment.
Our work in mass customization helps industrial practitioners in several ways. Our classification of mass customization systems assist a company in determining what type of mass customization system it should adapt taking into account its existing production capabilities. Using the set of influential factors identified by us, the company can plan, design and implement an appropriate mass customization system for the company. We also offer to build intelligent computer tools for operating mass customization systems.
Product Realization for Recyclability and Reusability
This research investigates the product realization for recyclability and reusability. The idea is to look beyond the conventional product life cycle namely: design, manufacture, use, and dispose. This research is motivated by increasing environmental awareness among consumers and producers, government recycling regulations, and resource conservation needs. This research is conducted in collaboration with the Director of the Laboratory for Responsible Manufacturing, Professor Surendra Gupta who is a leading international authority in this area of research.
Process Monitoring for Product Quality Control
A major goal of process monitoring for product quality control is to detect any special disturbances in the process as early as possible, so that investigation of the process and corrective actions can be taken before many nonconforming products reach the final stage of production. The direct benefits of having a properly designed monitoring system include improved product quality, better plans for maintenance, more effective control of operating machines, and better manufacturing decisions with process and workload plans. In process monitoring for product quality control, any observed variation in a process variable such as the diameter, length, or surface finish of a part is attributable to either random causes or assignable causes. From the process control viewpoint, the tendency of any arbitrary pattern to repeat itself should cause concern for investigation into potential assignable causes.
We have developed a general-purpose scheme for process improvement that detects all systematic repetitive patterns from a measured data. This new method automatically identifies all repetitive patterns of any structure, even if their length is relatively small. Once these symptomatic patterns are correctly identified, the results can be used to identify the underlying reasons and to plan preventive control actions. Some of the numerous possible reasons can be found in voltage fluctuation of a power source, shift changes of operators, unfavorable humidity or temperature changes, and machine tool behavior under different operating conditions. Our pattern recognition-based method is very effective and could lead to an effective process monitoring and diagnosis for quality control.
This research focuses on developing a multiagent-based system for synthesis and analysis of design solutions that satisfy a given set of functional requirements to generate creative design solutions. These artificial intelligence-based computer tools will be useful for faster product development.
Neural Networks and Knowledge-Based Systems
Neural networks and knowledge-based systems are the intelligent computer tools useful for building powerful decision making systems which can emulate the capabilities of a human expert in a given domain of expertise. In industry, they can be used for several purposes which include interpretation, selection, prediction, monitoring, diagnosis, fault detection, design, planning, control, scheduling, maintenance, etc.
We have developed neural networks and knowledge-based systems for several applications such as formwork selection in construction, undergraduate course selection, computer-aided process planning and cost estimation for a sheet metal company, cost reduction in purchasing for the Naval aviation supplies office, and tool wear estimation.
Database Systems Production Environment
We can develop high quality customized database systems
that can support group technology, design data retrieval, computer-aided
process planning, cost optimization and estimation, just-in-time inventory