A Proposed Bearing Load Identification Method to Uncertain Rotor Systems

Abstract

Bearings are considered as important mechanical components in rotating machines. Bearing load is used as an indication of monitoring rotor system health, but there are interval and probability uncertain parameters in the process of obtaining bearing load from the rotor system. A bearing load strip enclosed by two bounding distributions is then formed, rather than a single distribution that we usually obtain through the load identification method for a deterministic rotor system. In this paper, a computational inverse approach that combines the interval and perturbation analysis method with regularization is presented to stably identify bearing load strip. Using an interval analysis method, a calculated transient response of the rotor structure only subjecting to the bearing load can be approximated as a linear function of the interval parameters in the rotor system. The perturbation analysis method based on Taylor expansion is used to transform the problem of the bearing load identification involving in probability parameters into two kinds of certain inverse problem, namely, the bearing load identification combining the mean value of uncertain parameters with calculated transient response function and the sensitivity identification of bearing load to each probability parameter. Regularization is used to overcome ill-posedness of bearing load identification arising from the noise-contaminated observed response. A rotor system with two bearings is investigated to demonstrate the effectiveness and accuracy of the presented method.

A state-of-the-art review on uncertainty analysis of rotor systems

“…The most common research objectives in the literature concerning parameter identification with uncertainty in rotor systems are to identify imbalance [170], bearing properties [89,160,199,250,251] and damages [166,252]. Mao et al [253] proposed a microgenetic algorithm with the advance and retreat method to identify the eccentricity of the spindle-tool system subject to random inputs and outputs.…”

Non intrusive load identification method based on deep learning

“…According to the installation location of the acquisition device, the power monitoring system can be divided into invasive load monitoring (ILM) and non-invasive load monitoring (NILM). The principle of intrusive load monitoring is to install a signal acquisition device at the incoming end of each electrical appliance,Upload via data line or wireless WiFi to facilitate the detection and collection of power consumption data [1]. The principle of non-invasive load monitoring is to install acquisition devices such as voltage and current sensors on the user's incoming line side, transmit the collected information to the processor through the bus, and obtain the power change curve, current waveform or data track and other characteristics within the time period after analysis and processing.…”