Prediction of measuring instrument calibration interval based on risk priority number and reliability using machine learning

The present study develops and introduces a new technique for predicting the calibration interval of Measuring instruments using machine learning (ML) with the features of risk priority number (RPN) and reliability (R). The proposed method predicts the calibration interval by considering risk, R based on instrument life cycle distribution and ML techniques. To check this prediction method, the data related to 220 measuring instruments of the steel company were used, and for each measuring instrument in this section, according to the opinion of the company’s experts, RPN was determined, and then based on the life cycle distribution of each, the Reliability index was calculated. Two hundred twenty measuring instruments were placed in three clusters of 12-month, 18-month, and 36-month calibration intervals using the K-Means clustering technique to label the data. Then, to predict the calibration interval of new measuring instruments, three conventional classifiers in the application of ML in maintenance, namely K-NN, RF, and SVM, were employed and tested for the data of the new measuring instruments. Finally, evaluating the performance accuracy of these three methods for prediction according to the data class, K-NN, and RF methods provided better performance.