Fault Diagnosis by Mathematical Programming

This paper presents the problem of fault diagnosis for logically represented continuous systems that can be formulated through nonlinear mathematical programming. This problem is transformed to an integer-programming problem and solved. Possible modifications and extensions of the problem are given. Although failure tables must be prepared in ordinary fault diagnosis, they are not necessary with this mathematical programming approach. By modifying constraints in the mathematical programming problem, difficulties such as multiple faults, correlated faults, modifications of test conditions and cycles in the system, which are encountered in the ordinary failure table approach, are made tractable.