Diagnosis of the Accuracy of the Vehicle Scale Using Neural Network

Abstract

Abstract The article describes a method for diagnosing the accuracy of the vehicle scale without using standard weights. The novel method defines the possibility to estimate whether the scale would pass the test for error of indication in the next verification or not, only by using the results from simple tests with load of estimated weight and appropriate classifier. The method is primarily developed for users of these scales. Created classifier is based on the neural network algorithm. The neural network was trained with data from verifications, which are provided by Slovak Legal Metrology. Well trained classifier can provide not only information whether the scale will potentially pass the mentioned test or not, but reliability which is associated with this result as well. In this way, the user has valuable information about the scale in the period between the verifications.