Security early-warning model based on genetic wavelet neural-network

Abstract

The invention discloses a security early-warning model based on a genetic wavelet neural-network. The model includes the following steps: 1) carrying out fuzzification of input data; 2) determining the layer number of the genetic wavelet neural-network and the node numbers of an input layer, an output layer and hidden layers; 3) carrying out information encoding; 4) carrying out population initialization; 5) carrying out fitness function calculation; 6) carrying out a selection operation; 7) carrying out a genetic operation; 8) selecting an optimal individual for decoding; 9) carrying out training according to parameters of the BP network set in the first step; 10) optimizing the wavelet neural-network; and 11) optimizing a genetic function. According to the security early-warning model ofthe invention, a structure of the wavelet neural-network is simpler, a convergence speed is higher, the neural network is enabled to have higher learning ability through using wavelet basis functionsas neurons of the neural network, precision is higher, immediate, highly-efficient, scientific, accurate and early warning on a chemical production system is realized, and chemical production accident appearing is avoided.

Description

technical field [0001] The invention relates to the field of chemical production early warning, in particular to a safety early warning model based on a genetic wavelet neural network. Background technique [0002] The chemical production process involves a large number of hazardous chemicals, harsh production process requirements, and large-scale, continuous and automated production equipment. Once an accident occurs, the consequences will be extremely serious. The chemical production process is a typical giant system. Using general It is difficult to give a scientific and accurate early warning to the security system by the mathematical model construction method. As a typical feed-forward neural network, BP neural network has been widely used in many fields. It has nonlinear, self-learning, self-organizing However, in practical applications, neural networks still have certain limitations: the convergence speed is slow, and other factors of the network, such as various para...

AI Technical Summary

Problems solved by technology

[0002] The chemical production process involves a large number of hazardous chemicals, harsh production process requirements, and large-scale, continuous and automated production equipment. Once an accident occurs, the consequences will be extremely serious. The chemical production process is a typical giant system. Using general It is difficult to give a scientific and accurate early warning to the security system by the mathematical model construction method. As a typical feed-forward neural network, BP neural network has been widely used in many fields. It has nonlinear, self-learning, self-organizing However, in practical applications, neural networks still have certain limitations: the convergence speed is slow, and other factors of the network, such as various parameter settings, also affect the convergence speed, which obviously has nothing to do with security warnings. Immediate efficiency does not match the system requirements

[0003] Existing technologies such as Chinese Invention Authorized Patent Documents, Authorized Announcement Number: CN 103077408B, the invention provides a method for converting seabed sonar images based on wavelet neural networks into acoustic substrate categories, using the algorithm of genetic wavelet neural networks, which can perform Local analysis, optimize the initial parameters of the network through the genetic algorithm, avoid falling into a small local area, effectively avoid noise and local extremum, and make the conversion from seabed sonar images to acoustic bottom types more accurate and reliable. It has important practical value, but for the same task, the structure of wavelet neural network is complex, and the convergence speed is not ideal

Images