Soot blowers are employed in a predefined sequence and schedule to clean the ash particles on heat exchanger surface. However, considering the detrimental effects of frequent soot blow operation a criteria based system has to be adopted that involves soot blow activation based on a predefined cleanliness level. This insists the need for an online monitoring model to predict the Cleanliness Factor (CF). This paper analyzes the deposition of ash in power plant reheater with the objective of developing accurate prediction model for CFusing Autoregressive Integrated Moving Average with eXplanatory variables (ARIMAX) model. In practice the covariates that should be included in the model is not known a priori and often with more number of candidate variables. The findings in this work reveal that the ARIMAX model including flue gas input temperature, 1 lag value of flue gas output temperatures, air flow rate, coal flow rate and attemperator flow rate in time series analysis of CF dataset produces a more robust predictive model. The model performance is quantified by twoindicators namely and Root Mean Square Error (RMSE) and Akakie Information Criteria (AIC). To ensure the adequacy of the model residual diagnostics was performed which revealed that the Auto Correlation Function (ACF) plot of the residuals are uncorrelated and there is no considerable departure from white noise as the test statistic clearly shows that the p-values exceeds the 5% significance level for all lag orders. The recommendation suggested in this study can be applied to criteria based soot-blow operation where the soot blowers are operated when a predefined CF is reached.