Search

Operators from various industries have been pushing the adoption of wireless sensing nodes for industrial monitoring, and such efforts have produced sizeable condition monitoring datasets that can be used to build diagnosis algorithms capable of warning maintenance engineers of impending failure or identifying current system health conditions. However, single operators may not have sufficiently large fleets of systems or component units to collect sufficient data to develop data-driven algorithms. One potential solution to overcome the challenge of having limited representative datasets is to merge datasets from multiple operators with the same type of assets. However, directly sharing data across the company's borders yields privacy concerns. Federated learning (FL) has emerged as a promising solution to leverage datasets from multiple operators to train a decentralized asset fault diagnosis model while maintaining data confidentiality. However, the performance of traditional FL algorithms degrades when local clients' datasets are heterogeneous. The dataset heterogeneity is particularly prevalent in fault diagnosis applications due to the high diversity of operating conditions and system configurations. To address this challenge, this paper proposes a novel clustering-based FL algorithm where clients are clustered based on their dataset similarity. Estimating dataset similarity between clients without explicitly sharing data is achieved by training probabilistic deep learning models and having each client examine the predictive uncertainty of the other clients' models on its local dataset. Clients are then clustered for FL based on relative prediction accuracy and uncertainty. Experiments on three bearing fault datasets, two publicly available and one newly collected for this work, show that our algorithm significantly outperforms FedAvg and a cosine similarity-based algorithm by 5.1% and 30.7% on average over the three datasets. Further, using a probabilistic classification model has the additional advantage of accurately quantifying its predictive uncertainty, which we show it does exceptionally well. [ABSTRACT FROM AUTHOR]

This paper proposes a method for detecting insufficient lubrication in rolling bearings, by utilising a low cost MEMS ultrasonic microphone. We run a set of grease lubricated bearings under load, alternating between good and insufficient lubrication conditions by adjusting speed and temperature, over an 18 day experiment, followed by a short validation experiment on a different set of bearings on a secondary test rig. Vibration signals are recorded from an ultrasonic microphone, using a purpose built prototype sensor, then band-pass and high-pass filtered and subsequently analysed using statistical time-domain features. The viscosity ratio Kappa of the lubricant is used as a metric for quality of oil-film formation, and is calculated throughout the experiment from rotating speed and temperature. The results are compared with the predictions from a linear regression model and a simple Neural Network trained on ensembles of three and ten selected features. The three Hjorth's Parameters activity , mobility and complexity calculated on high-pass filtered microphone signals from the experiments show a good correlation with the viscosity ratio Kappa. When combining these, using a Linear Regression model or a simple two layer Neural Network model, we are able to predict the viscosity ratio with good precision. The proposed method may be used to assess the quality of oil film formation in bearings during operation under varying speed and temperature conditions. [ABSTRACT FROM AUTHOR]

• The paper focuses on the Hjorth's parameters application for condition monitoring. • The Hjorth's parameters are three: the Activity, the Mobility and the Complexity. • A new parameter called Detectivity is introduced to merge Hjorth's parameters. • Effectiveness of Detectivity is proved on both experimental and synthetic datasets. Hjorth's parameters are statistical time-domain parameters used in signal processing and introduced by Bo Hjorth in 1970. These parameters are Activity, Mobility and Complexity. They are related to the variance of the signal and of its subsequent derivatives. They are commonly used in the analysis of electroencephalography (EEG) signals for feature extraction, but also in the tactile signal analysis in robotic area. In this paper, Hjorth's parameters are applied to vibration signals for fault detection in ball bearings. In particular, two open-access datasets are used: the NASA bearing dataset of the University of Cincinnati and the Polytechnic of Turin rolling bearing dataset. In the first part of the paper the three parameters are used for monitoring the health of the bearings along their lifetime, proving their effectiveness in condition monitoring. In the second part, a new parameter is introduced, named Detectivity, that merges the information carried by Hjorth's parameters in a single value and is suitable for continuous monitoring of machines. [ABSTRACT FROM AUTHOR]

This paper presents a data-driven prognostic framework for rolling-element bearings (REBs). This framework infers a bearing's health index by defining a degree-of-defectiveness (DD) metric in the frequency domain of bearing raw signal, named DD-based health index (DDHI). Then, we systematically apply least-square support vector machines (LSSVMs) in the forms of Bayesian inference-aided one-class LSSVM (Bayesian-OCLSSVM) for anomaly detection in order to define the time-to-start (TTS) point of RUL prediction and the recurrent least-square support vector regression (Recurrent-LSSVR) model for predicting future values of DDHI for calculating the RUL. In addition, this paper addresses several pertinent challenges, such as failure threshold determination during anomaly detection and RUL estimation, by developing adaptive thresholds. We conduct extensive experiments on publicly available two benchmark datasets using a run-to-failure experiment. The results demonstrate the efficacy of the proposed framework compared to state-of-the-art methods in terms of the accuracy and convergence of the RUL estimation of bearings. [ABSTRACT FROM AUTHOR]