Intelligent fusion of sensor data for product quality assessment in an industrial machine

• Main Author: Jain, Abhinandan
• Language: English
• Published: 2009
• Online Access: http://hdl.handle.net/2429/13736

Sensor fusion (or data fusion) refers to the combined and synergistic use of information from multiple sensors in order to increase the reliability, accuracy, and overall effectiveness of the sensor-based operation. This thesis investigates knowledge-based fuzzy sensor fusion methods that reliably fuse both redundant and complementary data. Fuzzy sensor fusion methods use fuzzy logic in arriving at a "fusion decision", and they can fuse data that are imprecise, incomplete, or even conflicting, to provide meaningful information required for decision-making related to a process. Such decisions are useful, for example, in assigning quality grades to products, assessing the performance of a process, and in feedback control, both direct and supervisory. 'Fuzzy techniques are particularly useful when the errors and uncertainties associated with the sensors are subjectively assessed rather than based on precise statistics. These methods are applicable when the plant/process is complex, incompletely known, and difficult to model either analytically or experimentally. In this thesis, the uncertainty and error in sensor data are considered to be both subjective and qualitative and hence related to the concept of fuzziness. The thesis specifically considers the implementation of fuzzy sensor fusion methods for the quality assessment in industrial production machines. Primary attention is given in this regard to an industrial fishcutting machine developed in the Industrial Automation Laboratory, University of British Columbia. Also some attention is given to CNC router machines used in material cutting. Three appropriate methods of multiple-sensor fusion using fuzzy logic are adopted, implemented, and evaluated in the present work. The first sensor fusion method is based on Mamdani's max-prod (or, sup-prod) composition, and it places equal weights on all the data sources, without considering their merit or importance. The second sensor fusion method is based on degree of certainty. It assigns weights proportional to the degree of certainty of sensor data, and in addition to the fused output, it provides information about the certainty of the output. The third sensor fusion method is based on the concept of compatibility of data. It provides a fused output and additional knowledge about the degree of confidence in that output. This method is specifically effective, when sensors provide conflicting information. Thus, fusion is expected to improve the reliability of the sensor information and hence the knowledge of the system. The three knowledge-based sensor fusion techniques are implemented in a prototype fishcutting machine. The results are critically examined to determine the effectiveness and relative merits of the techniques.