Working-Hour Calculation of Construction Machinery Based on Outlier Analysis

This paper focuses on a class of vibration signals collected near the engine of construction machinery, which is used to propose a working-hour statistical method that conform to labor rights and interests. According to the quartile method to analyze the short-term amplitude distribution of the samples, a method of counting outliers is proposed to classify the idle state and working state of the mechanical equipment to achieve fine-grained state classification of the machinery. Based on the actual scene requirements, time smoothing is adopted to revise the idle state with short duration in the collected samples to the working state, while the continuous idle state with long duration remained unchanged. The experimental analysis shows that the method based on counting outliers is hopefully suitable for the state classification of various construction machinery (manufacturer, date of manufacture, size). The method proposed in this paper is simple in implementation and avoids the work of collecting data on the different running states for different devices and labeling in methods based on supervised learning, which is greatly convenient for practical engineering applications.