This paper focuses on the issue of outlier detection for time series in the process industry. Considering the characteristics of time series in process control systems, such as high non-linearity, strong noise and the special relationship between the input and output of the controlled object, a new outlier detection algorithm is proposed. The algorithm adopts an improved Radial Basis Function Network to construct the model of the controlled object and an Auto-Regression Hidden Markov Model to detect outliers. Unlike many conventional outlier detection methods, this algorithm does not need any prior data and can detect outliers accurately without preselecting the threshold. The proposed detection algorithm is validated by the application to the electrode regulator system of an arc furnace and comparison with Takeuchi's auto-regressive model detection approach.