Fault Tolerant Structure Design Of Linear Digital Filter

A digital filter is an important part in modern communication systems. If it breaks down in extreme environments, such as high temperature, outer space, underground, underwater and complicated electromagnetic radiation environments, it will cause unnormal operations of the system and economic losses. So the filter performance and the reliability of continuously running are of great significance and the study of fault tolerance is very urgently needed. Filter fault tolerance refers to: when the filter is interfered or some failures of these structures occurs, the normal use of filters will not be too much affected and the filter can work in the scope that its performance index is allowed.The research of filter fault tolerance at present is through backing up redundant modules and the fault-tolerant mechanism of partial reconstituted system in hardware. These methods need to add extra redundancy and detect the failures, but the ways to deal with the failures are different. Increasing the redundant components can help to reduce the uncertainty of the equipment, but due to the increase of integration density in digital signal system and the increasingly complex manufacturing technology, the extra redundant modules will lead to big cost of hardware resources and the limited fault-tolerant ability. What's more, partial reconstituted system in hardware needs to be controlled by external controller. It will cost a large amount of calculation and long fault-tolerant time. Designing a digital filter that can tolerate general fault and has excellent performance is the topic of the article. In this paper, the specific job mainly includes two parts: the design of fault-tolerant structures for digital filters, and simulation analysis of fault tolerance for digital filters.It is important to design a digital filter. This paper first introduces performance indicators of digital filters, and then introduces two kinds of design methods of digital filters: the identification methods of transfer function coefficients, and the identification methods of structure parameters. Through the detailed analysis of these two kinds of methods, we can see for the digital filter with a complex structure, the transfer function design method can not meet the use requirements and the applicability is not strong. This paper attempts to explore a design method for the design of optimal digital filter structure based on the requirements of the target. According to the performance requirements of the complex environments, this paper introduces the concept and the design method of fault-tolerant filters, which provides a useful reference for the research ideas and methods of this paper. Finally, a digital filter design method which can satisfy the general fault tolerance is proposed.The research works in several aspects:(1) This paper presents a method of digital filter design based on genetic algorithm. This method can automatically generate the signal flow graph and construct the legal structure space. In this paper, the genetic operators are modified, and a reasonable coding strategy and the fitness function are designed, so that the filter structure evolve in the legal structure space.(2) An improved optimization algorithm is proposed in this paper. The algorithm doesn't need that the objective function is continuously differentiable, and can be directly optimized to the target. In this algorithm, the search step lengths and the waiting generations are dynamically adjusted, the convergence speed is accelerated, and the optimization ability of the multidimensional space is improved. It further optimizes the structural coefficients of the digital filter.(3) The digital filter structure and parameters can be optimized at the same time. First, using genetic algorithm optimizes the filter structure, and then using differential evolution algorithm and the improved optimization algorithm optimize the coefficients of the multipliers in the digital filter. The proposed method can be used to design the filter without the transfer function or with a complex transfer function, it can meet the complex requirements. As long as the design objectives and constraints are given, the method can determine the optimal filter structure and parameters without any additional prior knowledge, which can improve the level of the automation design of large-scale systems.(4) Analyzing the tolerant ability of the designed filter. The proposed structure and the classic structures based on the traditional filter prototype are compared.This paper proposes a design method of linear fault-tolerant digital filter structure. Without considering transfer function, it can directly generate the optimal filter structure according to the target features. Differential evolution algorithm and the improved optimization algorithm have made the further optimization of digital filter coefficients, bringing us the digital filter with general fault tolerance. Simulation experiment shows that compared with the traditional design method of digital filter structures, this method in the paper of the filter structure is of good robustness when multipliers in the digital filter appear faults of the open or short circuit. With the increase of failure rates, this method shows stronger adaptability and better fault-tolerant performance.