Due to the harsh working environment of gearboxes and strong noise interference, the accuracy of fault identification is seriously affected. In addition, the traditional ResNet network has a large number of network parameters, resulting in slow training speed. Experiments show that the accuracy of the proposed Fast-SeResNet network is maintained at 99% compared to the traditional ResNet network, while the number of training parameters has been reduced by an order of magnitude, and the training time for each Batch is reduced from 66 seconds to 10 seconds with the same hardware support. The results show that the Fast-SeResNet network structure can improve the diagnostic speed to a large extent in a noisy environment with a small improvement in network accuracy, and has some practical value. © 2023 IEEE.

Inverters having high voltage levels, high power density, and high integration are widely used. However, many high-frequency switch units also increase the probability of failure. Therefore, developing an accurate and stable fault diagnosis method is necessary. This paper proposes a fault diagnosis algorithm based on deep learning and the evidence reasoning (ER) rule. It not only ensures high diagnostic accuracy, but also enhances the stability of the diagnostic results. The algorithm takes the three-phase voltage source inverter as the research object and extracts the three-phase current signals with different types of faults as features. First, Convolutional and Deep Neural Network methods were utilized independently to determine a preliminary diagnosis. Second, the softmax functions of the Convolutional and Deep Neural Network outputs provided the probability distribution of the fault category, which was used as the evidence body for the ER rule to construct the fusion diagnosis. In addition, a new method of determining the reliability and the importance factors of the evidence was proposed in which the evaluation index of the deep-learning diagnosis result was applied. Finally, the final classification result was obtained using the ER rule. The proposed method can effectively enhance the accuracy and robustness compared with a single classifier. © Published under licence by IOP Publishing Ltd.

In tribology, bio-inspired surface textures are a potentially significant area of investigation and have achieved excellent success in past practice. In this study, a texture was designed that can be applied to cylinder liner-piston rings (CL-PR) as a solution for enhancing tribological properties, with reference to the microstructure of ventral snake scales. Three different sizes of bionic textures were processed on CL surfaces by laser surface texturing. Then, a reciprocating friction test machine was used to examine the influence of these textures on the tribological behavior of CL-PRs. A single speed (100 rpm) and three loads (200, 400, and 600 N) were applied. The experimental results showed that the combination of dentate structures and dimples in the artificial texture, mimicking snake ventral scales, could achieve enhanced lubrication. The adoption of these textures reduced the friction coefficient up to 35%. Interestingly, there was a violent interaction between the PR and the CL texture, which was sufficient to change the surface structure of the PR. This effect produced additional friction and reduced the degree of influence of oil film thickness on the friction coefficient. This study provided a reference for the application of friction reduction through texture for CL-PRs in diesel engines. It was simultaneously informative for future in-depth studies of bionic textures. © 2023, Harbin Engineering University.

As one important supporting component of the ship propulsion system, the water-lubricated stern tube bearing has profound effects on navigation safety. One the other hand, it is difficult to ensure the adequate lubrication for water-lubricated stern tube bearing under low speed working conditions. In order to mitigate this problem, contemporary bearing systems have been built of Ultra-High Molecular Weight Polyethylene (UHMWPE). However, it is observed that material surface texture features have a significant impact on the lubrication effectiveness. In this research, three types of micro convex textures such as cuboid structure were designed and surfaces of UHMWPE samples were moulded. A series of experimental tests were then carried out in a specially designed tester to investigate the tribological characteristics of different convex texture designs. Comparative analyses were conducted on friction coefficients, wear mass losses and worn surfaces for different convex textures and running conditions. Analysis results showed that there are significant differences in tribological properties of rubbing pairs with different convex textures. The variations of the friction coefficient of convex textured samples are less than the original sample. The wear performance of the convex textured design is superior to that of other convex textures. The convex textured is the most effective design in improving wear properties at a low sliding velocity of 0.063 m/s. This work has established a practical basis for optimal texture design, of water-lubricated material, for reduced wear and improved lubrication performance.

The performance of diesel cylinder liner-piston ring (CLPR) friction pairs with different surface textures are very important and affect the service life, reliability, and economy of diesel engines. The aim of this study was to gain insights into interactions between thread groove surface texture (TGT) and the friction and wear behavior of a marine diesel CLPR. Four kinds of TGT with different widths, including 1, 2, 3, and 4 mm, were designed and machined on cylinder liners and then tested using a four-stroke CLPR friction and wear tester. The cylinder liner pressure, contact resistance between cylinder liner-piston ring, and worn surface morphologies of cylinder liners were obtained to examine cylinder liner performance with different width TGT. Compared with untextured cylinder liners, the experimental results showed that TGT significantly affected tribological behavior and consequently affected sealing performance of CLPR systems. Specifically, 3 mm TGT had the clearest effect on CLPR system performance, as the CLPR antifriction performance showed an average friction reduction of 30.9%, oil film lubrication performance, reflected by contact resistance, increased by 33.3%, and sealing performance improved by 14.4%. These results aided in the understanding of specific applications of surface texture on wear performance in CLPR friction pairs which could be applied in slower sliding long stroke marine engines.

Rolling bearing performance often degrades due to extra wear introduced by sliding-rolling contacts. Thus, it is vital to monitor the evolution of such wear process through investigating the worn surface. Compared with traditional two-dimensional (2-D) methods, three-dimensional (3-D) measurements of worn surfaces can provide more information. Due to the lack of in-situ 3-D measurement techniques and the scale-dependency of characterization parameters, it remains a challenge to examine the wear evolution of sliding-rolling contact using 3-D surface topography features. A characterization framework is here developed for inspecting worn surface topography variations under sliding-rolling contact, by combing 3-D surface reconstruction and computed fractal dimension. The worn surface is firstly imaged by an in-situ microscope. Next, a virtual 3-D surface is obtained from 2-D images via topography reconstruction. Then, the fractal dimension of worn surfaces is calculated to characterize the wear state with 3-D root-mean-square. A wear test is carried out on a roller-ring test rig to verify the proposed method. Results indicate that the proposed 3-D characterization performance is comparable to laser scanning confocal microscopy (LSCM) and allows rapid description of the wear process.

Mechanism-based wear debris classification (WDC) is important for root cause analysis and prediction of wear related faults. Compared to manual classifications, automatic WDC is more efficient and often more reliable for a wide range of industrial applications. However, existing methods unavoidably encounter some difficulties when dealing with those wear particles with highly geometric similarity, especially for fatigue particles and severe sliding particles. To meet the requirement for automatic WDC, an integrated, automated method for identifying typical wear debris is proposed with a two-level classification procedure. By referring to the traditional ferrography - a widely used wear particle imaging and analysis technique, the first-level classification is performed by a general back-propagation (BP) neural network with selected particle's morphological features. By doing this, three types of wear particles including rubbing, cuffing, and spherical particles can be determined. In the second-level classification, a deep learning model of a 6-layer convolution neural network (CNN) is adopted to identify fatigue particles and severe sliding particles by analyzing their very slight surface details in pixel-level. The method is tested with over 100 images of real particles generated from an extruder machine in a petrochemical plant and identified by a ferrograph specialist. A high recognition rate of over 80% is achieved for the three types including rubbing, cuffing, and spherical particles with the first procedure. Further, the identification rates are 85.7% and 80% for fatigue particles and severe sliding particles, respectively, which is distinctly improved from the reported values (they are 45.5% and 36.4%, respectively) of other intelligent methods.

On-line detection of metallic wear debris is an effective approach for condition monitoring of mechanical systems. Existing on-line oil conditioning sensors are mainly based on ferrography and inductive techniques. However, ferrography technique needs a clean background and inductive technique requires a high cleanliness of lubricant. To solve these issues, in this paper a metallic wear debris sensor based on eddy current principle is developed. Both numerical simulations and prototype experiments are conducted to evaluate the capacity and feasibility of the new sensor for detecting wear debris. The analysis results demonstrate that: 1) A pulse is generated when the wear debris pass through the sensor, the amplitude and width of the pulse can be used to identify the material and size of the debris; 2) The developed sensor is able to detect copper debris with a diameter greater than 150 pm and iron debris greater than 60 pm. This work provides a new idea for detecting wear debris and a new method for obtaining the characteristics of wear debris.

Early and accurately detecting faults in rotating machineries is crucial for operation safety of modern manufacturing system. In this paper, we proposed a novel deep CNN method based on knowledge-transferring from shallow models for rotating machinery fault diagnosis with scarce labeled samples. It is based on the idea that shallow models trained with different hand-crafted features can reveal the latent prior knowledge or diagnostic expertise and have good generalization ability even with scarce labeled samples. First, The raw vibration signal is transformed into time-frequency domain by applying the short-time Fourier transform (STFT) to extract integral features accordingly. Then, we train the SVM model with scarce labeled samples and make predictions on unlabeled samples. The predicted labels can be regarded as the data format of expert knowledge learned by the SVM model, which are combined together with the scarce fine labeled samples. Finally, they are used to train a deep CNN model of better discriminative ability. Experimental results demonstrate the effectiveness the proposed method that it achieves better performance than SVM model and original deep CNN model trained with only scarce labeled samples. Moreover, it is computational efficient and is promising for real-time rotating machinery fault diagnosis.