Adaptive Soft Sensor based on moving Gaussian process window
Ali Abusnina, Daniel Kudenko⋆
Abstract
Ali Abusnina, Daniel Kudenko⋆
Abstract
Soft Sensors are used in different industrial applications for their relatively low cost, simple development, and ability to predict difficult-to-measure variables (e.g., process quality, production efficiency). As many industrial processes are time-variant, and they exhibit dynamic behaviours, the Soft Sensor should be adaptive so as to be able to capture process changes, and keep reflecting on real status of the process by giving accurate predictions. This paper proposes an adaptive method based on moving Gaussian process window to tackle the adaptability problem, and to enhance the prediction accuracy of the Soft Sensor. The moving window is updated by deleting input points that give rise to predictions with the highest predictive density error. We empirically show that this method results in a higher accuracy than a moving Partial Least Square (PLS) window. The contribution of this work is i) developing adaptive Soft Sensors based on Gaussian process, ii) updating the moving window based on the highest predictive density error.
OpenAlex reports 6 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
Soft Sensors are used in different industrial applications for their relatively low cost, simple development, and ability to predict difficult-to-measure variables (e.g., process quality, production efficiency). As many industrial processes are time-variant, and they exhibit dynamic behaviours, the Soft Sensor should be adaptive so as to be able to capture process changes, and keep reflecting on real status of the process by giving accurate predictions. This paper proposes an adaptive method based on moving Gaussian process window to tackle the adaptability problem, and to enhance the prediction accuracy of the Soft Sensor. The moving window is updated by deleting input points that give rise to predictions with the highest predictive density error. We empirically show that this method results in a higher accuracy than a moving Partial Least Square (PLS) window. The contribution of this work is i) developing adaptive Soft Sensors based on Gaussian process, ii) updating the moving window based on the highest predictive density error.
Key concepts: Soft sensor, Computer science, Gaussian process, Process (computing), Adaptability, Window (computing), Gaussian, Moving average