let you have underling model as y_t=meu+epsilon and you want to monitor meu, another problem arises i.e. mean and variance of error term may not be constant, assuming they are constant ((or after transforming)), then by modifying the underling model as y_t=meu+y_(t-1) +epsilon. monitoring the meu and y_(t-1) is lag which help to tackle the autocorelation, besides this more lag can be used.

Thanks dear Raja

I found this useful.

http://cdn.intechopen.com/pdfs/38136/InTech-Applications_of_control_charts_arima_for_autocorrelated_data.pdf

Thanks dear Mohad. I really appreciate.