Font Size: a A A

Research On Methodology Of Analog Circuit Soft Fault Diagnosis With Tolerance

Posted on:2015-03-07Degree:DoctorType:Dissertation
Country:ChinaCandidate:Y C AoFull Text:PDF
GTID:1108330473456058Subject:Measuring and Testing Technology and Instruments
Abstract/Summary:PDF Full Text Request
Fault diagnosis for analog circuit is still a challenging subject in the circuit test research field. However, due to the inherent characteristics of analog circuits, such as the continuous variation of the component parameter, the tolerances of the component parameter and the nonlinearity of the response, etc, it is difficult for the conventional fault diagnosis theories and methods to achieve the expected diagnosed results in the engineering. Hence, it is important to explore some more efficient diagnosis theories and methods.Aiming at the unavoidable limitation of the existing diagnosis methods based on pattern classification, such as the inefficient on diagnosing the unknown faults, the incipient soft-fault and the multiple soft-faults, the diagnosis theories and methods on soft-fault under the tolerances were studied deeply in this dissertation on the ground of the probability distribution theory, the stochastic process theory and the particle swarm optimization. The main works and the contributions of the dissertation are as follows:(1)Combined with the classical slope fault model in soft-fault diagnosis of the analog circuit, a novel method of handling tolerances was proposed. Firstly, the slope fault characteristics of analog circuit under tolerances are treated as random variables(random variables of normal quotient distribution). Based on the definition of the slope fault characteristic, the approximating distribution function was deduced. Then, the monotonous and continuous mapping between normal quotient distribution and standard normal distribution was proved, and the estimation formulas of the range of the slope fault characteristic were deduced. After that, a new test-nodes selection algorithm based on the approximating condition was designed, and the fault dictionary was built. The given examples show that the proposed method of handling tolerances is effective and the diagnostic coverage is improved significantly.(2)For the poor diagnostic accuracy on the incipient soft-fault of the diagnosis methods based on the slope fault model and the normal quotient distribution, a new diagnosis approach based on the hidden markov model was presented. Firstly, inspired by the successful applications of the hidden markov model on the condition monitoring of the mechanical equipments, the variations of the components parameters were considered as the hidden state stochastic process while the measured output signals were considered as the observation stochastic process. Then, the soft-fault of the analog circuit was modeled dynamically by the hidden markov model. The fault simulation approaches, the selecting and traning approaches of the hidden markov model and the fault diagnosis approaches were given. The given examples show that the proposed hidden markov model based diagnostic method is effective and the diagnostic coverage on the incipient soft-fault under tolerances is improved significantly.(3)For the weak diagnostic capabilities on the unknown faults and the multiple soft-faults of the diagnosis methods based on pattern classification(including the hidden markov model based diagnostic methodes), a fault diagnosis method based on the particle swarm optimization was proposed. Firstly, on the basis of the voltage sensitivity matrix of the analog circuit, an effective test-nodes selection strategy was developed and the voltage increment equations at the selected test-nodes were constructed. Then, with the equations as constraint conditions, the problem of fault diagnosis is transformed into the problem of mathematical programming, and six fault diagnosis equations were built. After that, the particle swarm optimization was used to solve the six fault diagnosis equations uniformly. The given examples show that the proposed particle swarm optimization diagnostic method is effective, the incipient soft-fault and the double soft-faults under tolerances can be located and identified.(4)For the unsatisfactory diagnostic coverage of the particle swarm optimization based diagnostic method, an improved particle swarm optimization based diagnostic method was discussed. Due to the linear decrease of the inertia weight, the convergence ratio and the convergence speed of the basic particle swarm optimization drop drastically while solving the high-dimensional problems. Therefore, the diagnosis capability of the particle swarm optimization diagnostic method was affected severely. By monitoring and evaluating the movements of the particle swarm in real time, two improved particle swarm optimization algorithms with the inertia weight adjusted dynamically according to the current search status of the particle swarm were devoloped. Comared with some classical optimizations while solving several standard benchmark functions, the improved particle swarm optimization with adaptive inertia weight shows the obvious superiority on the convergence ratio, the convergence speed and the convergence stability. Finally, the improved algorithm is applied to diagnose the soft-fault in analog circuit. The given examples show that the improved particle swarm optimization diagnostic method is effective and the diagnostic coverage of the single soft-fault, the incipient soft-fault and the double soft-faults under tolerances are improved significantly.
Keywords/Search Tags:analog circuit, soft fault diagnosis, normal quotient distribution, hidden markov model, particle swarm optimization
PDF Full Text Request
Related items