Optimal sensor location for improved multifixture assembly system fault diagnosis: A methodology.

This thesis presents a multi-sensor placement methodology developed and implemented for dimensional fault diagnosis, and the process control of assembly lines such as those in sheet metal assembly. An assembly line is defined as consisting of a sequence of fixtures. Sensors are used to measure dimensional characteristics of the subassemblies as they move from fixture to fixture in the sequence. Fault diagnosis involves identifying the specific fixture and the functional part of the fixture at fault, if the assembly does not meet dimensional acceptance criteria. Fixture fault diagnosis is a critical component of techniques aimed at assembly variation reduction. The effectiveness of the diagnosis is contingent on the effectiveness of the sensor measurements of assembled parts. A novel means of enhancing diagnosability is proposed for multi-fixture sheet-metal assembly, using in-line measurement sensors, through a methodology which develops an optimized configuration of sensor locales for measurement. The methodology incorporates specifications (representations) of fixtures, assemblies, and sensor locations. Sensor placement optimization techniques are provided for three levels of system complexity in the development of the multi-sensor methodology: (i) for a single fixture assembly system, (ii) an enhancement for a multi-fixture assembly system with sensing at a measurement station in a single end-of-line location, and (iii) an enhancement to deal with multi-fixture assembly systems with sensing distributed at multiple locations in the assembly sequence. For a single fixture assembly system, the methodology involves optimization carried out on sensor location and number to achieve optimal diagnosability for that fixture. For a multi-fixture assembly system, a hierarchical group description of the assembly is used to build a state-transition representation which, with CAD fixture information, is used in multi-level hierarchical optimization to arrive at the optimal for the overall assembly. In the end-of-line sensing system, this can then optimally diagnose faults caused by failures anywhere in the multi-fixture assembly line upstream of the measurement station. The distributed sensing enhancement allows for enhanced fault diagnosis in an assembly line by finding an optimal locale of sensor positions at multiple, distributed, measurement stations. In each case, a configured optimal locale provides an optimally distinctive signature for each fault in the assembly, and optimally discriminates among all possible faults. The developed approach has significant utility in automotive body assembly, where system complexity makes the choice of sensor location vital to fault isolation. Examples using simulated and industrial automotive body assembly sequences are provided to help illustrate and validate the proposed methodology.Ph.D.Applied SciencesAutomotive engineeringIndustrial engineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/131687/2/9929861.pd