Control charts are widely used as a tool in process quality monitoring to detect anomalies and to improve the quality of a process and product. Nevertheless, their limitations have increased in the face of increasingly complex manufacturing processes. They do not have capability of handling large streams of non-normal and autocorrelated multivariate data, which is in most real applications. This may lead to an increase in false alarm signals and/or missed detection of anomalies. They are not designed to automatically identify the root causes of an anomaly when the process is out-of-control. Several machine-learning techniques were integrated with control charts to improve the sensitivity and specificity of anomaly detection. Nevertheless, some existing techniques still produce a high false alarm rate and/or missed detection. The root cause analysis is seldom performed. In this paper, we propose a new integration that combines the logical analysis of data regression technique (LADR) and the exponential weighted moving average (EWMA) as a new model-based control chart. LADR is based on the traditional LAD methodology, which is a supervised data mining technique for pattern generation. LADR transforms the original independent variables into pattern variables by using cbmLAD software to develop a regression model. The LADR–EWMA increases the sensitivity of anomaly detection in the process and uses the patterns to perform root cause analysis of that anomaly. We applied LADR–EWMA to a real application: a concrete manufacturing process. We compared its performance with Linear regression, Support vector regression, Partial Least Square regression, and Multivariate adaptive regression Spline. The results demonstrate that the LADR–EWMA, which is based on pattern recognition, performs better compared to the other techniques in terms of a reduction of false alarms and missed detection. In addition, LADR–EWMA facilitates interpretation and identification of the root cause of the detected anomaly.