Remaining useful life prediction is one of the essential processes for machine system prognostics and health management. Although there are many new approaches based on deep learning for remaining useful life prediction emerging in recent years, these methods still have the following weaknesses: (1) The correlation between the information collected by each sensor and the remaining useful life of the machinery is not sufficiently considered. (2) The accuracy of deep learning algorithms for remaining useful life prediction is low due to the high noise, over-dimensionality, and non-linear signals generated during the operation of complex systems. To overcome the above weaknesses, a general deep long short memory network-based approach for mechanical remaining useful life prediction is proposed in this paper. Firstly, a two-step maximum information coefficient method was built to calculate the correlation between the sensor data and the remaining useful life. Secondly, the kernel principal component analysis with a simple moving average method was designed to eliminate noise, reduce dimensionality, and extract nonlinear features. Finally, a deep long short memory network-based deep learning method is presented to predict remaining useful life. The efficiency of the proposed method for remaining useful life prediction of a nonlinear degradation process is demonstrated by a test case of NASA’s commercial modular aero-propulsion system simulation data. The experimental results also show that the proposed method has better prediction accuracy than other state-of-the-art methods.

Lubricating oil monitoring technology is a commonly used method in aeroengine condition monitoring, which includes particle counting technology, as well as spectral and ferrography technology in offline monitoring. However, these technologies only analyze the characteristics of wear particles and rely on physical and chemical analysis techniques to monitor the oil quality. In order to further advance offline monitoring technology, this paper explores the potential role of differences in wear particle kinematic characteristics in recognizing changes in wear particle diameter and oil viscosity. Firstly, a kinematic force analysis of the wear particles in the microfluid was carried out. Accordingly, a microfluidic channel conducive to observing the movement characteristics of particles was designed. Then, the wear particle kinematic analysis system (WKAS) was designed and fabricated. Secondly, a real-time tracking velocity measurement algorithm was developed by using the Gaussian mixture model (GMM) and the blob-tracking algorithm. Lastly, the WKAS was applied to a pin–disc tester, and the experimental results show that there is a corresponding relationship between the velocity of the particles and their diameter and the oil viscosity. Therefore, WKAS provides a new research idea for intelligent aeroengine lubricating oil monitoring technology. Future work is needed to establish a quantitative relationship between wear particle velocity and particle diameter, density, and oil viscosity.

A high-precision wear particle monitoring instrument based on numerical simulation, machine vision, multi-target tracking algorithm, and microfluidic technology is designed and developed in this paper. The instrument not only performs the traditional particle number and size detection, but also proposes a particle density detection model based on motion velocity. Firstly, an on-line visual microfluidic chip for oil particles motion was specially designed. Then a mathematical model relationship between particles velocity and particles diameter, particles density, oil viscosity, oil density was established by force analysis and solving the equation, which was used to explain the working principle of the instrument. Meanwhile, a multi-target tracking algorithm incorporating the characteristics of wear particles based on YOLOv5-Deepsort was custom-developed to detect the number, size, and motion characteristics of particles under the optical field of view of the image sensor. Secondly, the oil flow field distribution and the four factors affecting the particle motion characteristics in the self-developed chamber, particle size, particle material, particle radial position, and oil velocity were analyzed by using the finite element analysis in COMSOL. Finally, a pin-disk wear experimental validation results demonstrated this approach has excellent suitability to monitor the number, size, and density of wear particles with diameters above 20μm and densities above 2000 kg/m3 simultaneously. This portable oil particle monitoring instrument provides technical support for condition monitoring, fault diagnosis, and intelligent maintenance of rotating mechanical equipment.

Since inline monitoring method has the advantages of no sampling, being real-time, no human intervention, and low error, this paper innovatively proposes to study the inline monitoring of wear particles in an oil pipeline, from the perspective of the different motion characteristics of the particles. In this paper, an inline optical sensor was designed and developed by studying the velocity characteristics of different particles through theoretical calculations, numerical simulations, and experimental analysis. First, an equation for particle motion was statistically established, based on the forces acting on wear particles in an oil-filled vertical tube. Then a finite element model of particle motion in a full-flow oil pipeline was created, to simulate particle motion with various diameters, densities, locations, and shapes. Finally, the results of the theoretical study were effectively applied to design an inline optical monitoring sensor, and the experimental validation results demonstrated that the inline sensor has excellent suitability for monitoring wear particles. This study has significance for the safe operation of large rotating machinery.

A novel low-pass filtering self-adaptive (LPFA) denoising method combining complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and a wavelet threshold (WT) strategy is proposed to solve the problem of the aero-engine gas-path electrostatic signal noise, which challenges the gas-path component condition monitoring and feature extraction techniques. Firstly, the integration of CEEMDAN addresses modal aliasing and intermittent signal challenges, while the proposed low-pass filtering method autonomously selects valuable signal components. Additionally, the application of the WT in the unselected components enhances the extraction of useful information, presenting a unique and advanced approach to electrostatic signal denoising. Moreover, the proposed method is applied to simulated signals with different input signal-to-noise ratios and experimental fault electrostatic signals of a micro-turbojet engine. The comparison with several traditional approaches in a denoising test for the simulated signals and experimental signals reveals that the proposed method performs better in extracting the effective components of the signal and eliminating noise.

The prediction of the remaining useful life (RUL) of equipment is a pivotal function in realizing industrial intelligence within the Industrial Internet of Things (IIoT). It enables the formulation of corresponding predictive maintenance strategies, enhancing the reliability, availability, and safety of IIoT equipment. Given the limitations in existing literature, where the sojourn time distribution in each hidden state of the hidden semi-Markov model (HSMM) lacks universality, a hyper-Erlang distribution is employed to describe the distribution of sojourn times, which more accurately reflects the actual degradation process. The degradation process of the system is modeled as a three-state HSMM. States 0 and 1 are unobservable, representing healthy and warning states, respectively. Only state 2 is observable, representing the failure state. The expectation-maximization (EM) algorithm is utilized to estimate the unknown state parameters and observation parameters of the model. The posterior probability that the system is in the warning state is updated in real-time using Bayesian theorem. An explicit expression for the distribution of the RUL is derived. Predictions are made using an accelerated life test dataset for the gear shaft under varying load. A comparison is performed with the prediction methods, including the hidden Markov model (HMM), the mixture of Gaussians hidden Markov model (MoG-HMM), and the HSMM based on the Erlang distribution. The results demonstrate that the hidden semi-Markov degradation modeling approach, when considering the hyper-Erlang distribution, not only effectively reflects the degradation state of the gear shaft, but also facilitates high-precision prediction of the RUL of IIoT equipment.