Forecast Of Production Safety Accident Based On Wavelet Analysis Theory

In recent years,although the security management work has been improved, but the total amount of accident is still high and economic losse is very serious. The foreeast can contribute to grasp the tendency of the accidents, and can effectively reduce accidents. It’s an important content of the safety production management,and grate significant to improve the prediction accuracy. Accident are typical non-stationary time series with characteristics of nonlinear and random, make it difficult to forecast accident. Wavelet analysis theory is very suitable for non-stationary signals. The paper applies wavelet analysis theory to predict accidents. The main work and conclusions are as follows:(1) The traditional gray GM(1,1) model is established to prdict road accident of Shaanxi Province from 2002 to 2012, using the test method for evaluation of the posterior poor prediction accuracy ratin,using posterior deviation test to test the model accuracy. Since the volatility of the accident sequence data is relatively large,The accuracy of the model is the third level(barely qualified),the result is not ideal. On this basis, the introduction of an improved metabolism model prediction accuracy is increased from three to two(passes), we can see that if we continue to add new information and at the same time to weed out the old information is valid and re-modeling.(2) The number of 132 data about the road accidents of a Province characterized in month time from 2002 to 2012 which is denoised with wavelet multi-resolution analysis. The Mallat algorithm is used for the original sequence decomposition and reconstruction,the incident signal is decomposed into low-frequency approach partand high-frequency oscillatory part, the trend of low-frequency is close to the original signal, the high-frequency part contained most of the non-steady-state components called noise signal. Therefore,the accident signal denoising by wavelet analysis, so as to effectively extract useful information accident signal while the noise signal is separated, to make a more reasonable explanation for the characteristic of Occurrence of the accident.(3)The number of more than larger accident deaths of the province from 2005 to2012 in road traffic accidents, the trade and industry and mining industry is established forecast model which is based on wavelet analysis of the gray GM(1,1).First the wavelet function db N is used for the accident sequence wavelet decomposition and reconstruction,and then GM(1,1) model is used to forecast the wavelet transform series, and the prediction results are compared with the single gray GM(1,1) model. Matlab simulation results show that the paper choose the gray GM(1,1) model to predict the death toll accidents of industry and trade sectors of a province and the accuracy of predicted result does not meet the requirements.But the prediction accuracy of the various sectors of the province safety prediction have met the requirements based on wavelet-basis gray GM(1,1) model, showing that wavelet analysis based on GM(1,1) model is applicable to all sectors of production safety accident forecast. For more gradual accident data sequence,which is based on wavelet analysis of GM(1,1) model is a little better than single gray GM(1,1) model, but for volatile accident sequence, the prediction accuracy based on wavelet analysis of GM(1,1) model is significantly higher than the direct use of gray GM(1,1) model, then wavelet analysis of grey GM(1,1) model is proved the superiority in the accident prediction, has some practical value.