Optimization Technology Of Design For Diagnostic Strategy Based On Dependency Model

The performance of weapon equipments has been greatly improved with the widely application of high technology. On the other hand, the complexity of technology and structure of equipments increasing significantly raise challenges in equipment testing, diagnosis and maintenance which include poor testing and diagnosis, low accuracy, low efficiency and high consumption on time and cost. Extensive researches and practices make us recognize that Design for Testability (DFT) must be carried out concurrently with functional design of weapon equipments in order to fundamentally solve these problems.Diagnostic strategy design, as an important aspect in DFT, is crucial for reducing test cost and improving diagnostic accuracy and efficiency. A model-based method of diagnostic strategy design is widely used recently, in which the strategy design and testability evaluation are made following the construction of equipments dependency model.Although there have been numbers of theoretical findings, many complex problems still exist in this area.Supported by the advanced project "Equipments Testability/BIT Analysis, Design and Evaluation Technology", this dissertation makes the further studies on diagnostic strategy design methods mainly aiming at the currently existing key problems in the area, such as design in some typical complex situations (unreliable test, multiple faults) and design for systems with typical complex structure (multiple modes, multiple loops, multiple levels). The main contents and productions of the dissertation are as follows.1. Study on the theories of dependency modeling and optimization designs of diagnostic strategy.In order to accomplish the quickly optimal searching for diagnostic strategy under complex situations, a new searching algorithm, referred to as Quasi-Depth First Search (QDFS) algorithm, is presented by combining existing one-step look-ahead search algorithms with multi-step one. With information gain heuristic strategy as its core, the new algorithm and can be applied in complex situations. Experimental results show that the algorithm can reach an ideal trade-off between accuracy and computational complexity as well as produce optimal/near-optimal diagnostic strategies being of low cost and high efficiency.2. Study on the optimization design technologies for diagnostic strategy in typical complex situations.(1) In the situation of unreliable tests (two-valued and multi-valued respectively). Firstly a fault-test uncertainty dependency matrix is built using statistical method. Based on it, a new heuristic function used to guide the generation of diagnostic strategy is then presented. With the use of QDFS, an optimization algorithm is presented, as well as the construction of several prediction functions which are applied to evaluate diagnostic accuracy. Experimental results show the diagnostic strategies that follows from the proposed algorithm are lower costly and more accurate compared to the existing ones.(2) In the situation of multiple faults (in systems with and without redundancy respectively). First, based on the symptoms analysis of concurrent multiple faults with dependency model, three typical multiple faults mathematical models are constructed. Then a multiple fault inference engine based on Boolean logic is presented following the mechanism analysis of misdiagnosis caused by existing single fault inference engine. Based on the inference engine, an optimization algorithm is presented. Experimental results show the diagnosis of multiple faults is quick and missed or false diagnosis can be avoided.3. Study on the optimization design technologies for diagnostic strategy for the systems with complex structure.(1) In multi-mode systems. First, a series of fault-test dependency matrices for various system modes are constructed by analyzing fault propagation features of each mode. An optimization method of generating optimal mode sequences based on a heuristic function follows and then an optimization algorithm is presented. Experimental results show that the proposed algorithm carries out quick isolation of those faults that cannot be detected or isolated with any single mode.(2) In multi-loop systems. A loop-searching algorithm is presented firstly by the analysis of dependency model of multi-loop systems. There follows an optimization method of generating optimal feedback loop-breaking points based on a heuristic function. Then an optimization algorithm is drawn up. Experimental results show the diagnosis of multiple faults in the feedback loops is quick and with fewer breakpoints applied.(3) In multi-level systems. Firstly fault-test dependency matrixes of each replaceable unit at each level of system are constructed respectively according to the system hierarchy analysis. Then, existing heuristic function is improved in order to generate optimal diagnostic strategies which satisfying various fault isolation requirements. Based on QDFS and the new heuristic function, an optimization algorithm is presented finally. Experimental results show that the proposed algorithm can isolate the faults to any desired level and satisfy different maintainability requirements.4. Study on the software platform developing and the practical application of the methods mentioned above. Based on the researches mentioned above, a CAD software tool is developed for dependency modeling and diagnostic strategy optimizing.The application in a missile system validates the efficiency of the presented methods and software. Following the construction of dependency mode of the system, an efficient diagnostic strategy is derived with the aid of the tool. Then, the results of fault injection experiments and their evaluation show that the proposed strategies make the diagnosis of single fault or multiple faults accurate, quick and low costly, therefore satisfy the testability requirements.