To discuss Pm variation law with t in large scale coal dust explosion equipment, shock wave pipe is used and a 60 m linear explosion test system is built. P m of dust explosion is measured by flame collection system. ARIMA (p,q,d) time series analysis model of t and P m is built, differential methods are used to eliminate upward trend of the maximum flame length data. By calculating the autocorrelation, the partial correlation coefficient, and the value of AIC and SBC, autoregression and moving average orders are got. Model parameters are estimated by least squares method. 20 groups of flame length prediction show that: ARIMA model have high precision, prediction value of optimum ignition delay time is consistent with measured value, which realizes reasonable analysis on fluctuation characteristics of P m. This model can reasonably predict the up and down fluctuations of the local flame length data, reflecting the superiority of the model, which is not achievable by general prediction models. The results of this study can greatly reduce the number of coal dust explosion tests, save material costs and provide a new idea for studying the relationship between ignition delay time and coal dust explosion flame length.