Signal Preprocessing and Validation

Abstract

The firsts step of the vibration signal processing is its validation. This step is often overlooked or taken for granted, but to obtain meaningful results it is crucial to secure good quality data. The following chapter presents a comprehensive study on data validation as a prerequisite for data storage followed by data analysis. Based on scientific approach, a path of data validation is presented which may be implemented by researches as well as by diagnostic engineers. First, for machinery working in Varying Operational Conditions it is hard to compare the data from different time periods. A proper data selection algorithm is a good way to approach this problem. Validation should start with process parameters, first of all wind speed, shaft speed and output power. Two steps, namely one-dimensional and multi-dimensional are proposed. For vibration data, several real signals are presented. They represent good and faulty signals. An important distinction between correct, incorrect and invalid signals is proposed. Next, a number of validation algorithms is proposed. Since the time series for slow rotating parts are long, it is necessary to check stationarity of signals. Each method is presented with mathematical formulas to implement it. The chapter also includes the case study presenting the results of the application of these methods for a set of real signals. The chapter is concluded with a complete data validation algorithm.