Detection of Unit of Measure Inconsistency in gas turbine sensors by means of Support Vector Machine classifier

Thermal error modeling of machine tool based on dimensional error of machined parts in automatic production line

Thermally induced error has proven to be the major source of machining error for the machine tool working in a non-temperature-controlled workshop. Current research on thermal error modeling and compensation is implemented based on the measured thermal deformation of machine tool, and the data is obtained referring to spindle idling rather than metal cutting condition, resulting in the disadvantages of the established model with low adaptability to varying conditions. In this paper, a modeling and compensation method based on the dimensional error of the machined parts is proposed to address the issue and verified through machine tools in an automatic production line. Compared with the modeling based on thermal deformation, the method formulates a more direct relations between temperature rise and machining error. The thermal error modeling is carried out by measuring the dimension deviation of the inner hole diameter of the motor end cover and the screened representative temperature variables. Meantime, considering the temperature coefficient in model is difficult to converge when modeling with the single-day data, the unified modeling with the multi-day data is realized by improving the conventional multiple linear regression model. Finally, the generalized model used for thermal error compensation in x direction of the machine tool is obtained. The real-time compensation of the thermal error based on the established model is realized with the machine tool machining in mass production. The verification results show that this error modeling and compensation method can reduce the machining error of the end cover by more than 52%, irrespective of experiencing various complicated working conditions. Stability and robustness of the modeling are also validated through application on the other machine tool with the same configuration and real machining over seven days.

Optimal Classifier to Detect Unit of Measure Inconsistency in Gas Turbine Sensors

Label noise is a harmful issue that arises when data are erroneously labeled. Several label noise issues can occur but, among them, unit of measure inconsistencies (UMIs) are inexplicably neglected in the literature. Despite its relevance, a general and automated approach for UMI detection suitable to gas turbines (GTs) has not been developed yet; as a result, GT diagnosis, prognosis, and control may be challenged since collected data may not reflect the actual operation. To fill this gap, this paper investigates the capability of three supervised machine learning classifiers, i.e., Support Vector Machine, Naïve Bayes, and K-Nearest Neighbors, that are tested by means of challenging analyses to infer general guidelines for UMI detection. Classification accuracy and posterior probability of each classifier is evaluated by means of an experimental dataset derived from a large fleet of Siemens gas turbines in operation. Results reveal that Naïve Bayes is the optimal classifier for UMI detection, since 88.5% of data are correctly labeled with 84% of posterior probability when experimental UMIs affect the dataset. In addition, Naïve Bayes proved to be the most robust classifier also if the rate of UMIs increases.