Early fault diagnosis of rotating machinery is crucial in the industry. The network parameters of the traditional deep learning-based fault diagnosis method are optimized only by the single loss function, and the extracted features are usually not the most valuable features of the input signal. This article proposes a novel method, attention recurrent autoencoder (AE) hybrid model classification algorithm, for early fault diagnosis and severity detection of rotating machinery. The AE can learn the most valuable features in an unsupervised way. By adjusting the weight proportion of the two loss functions and optimizing the multibranch network simultaneously, the proposed method enables the network to extract the most valuable features of the input signals. Moreover, the proposed method can take the raw 1-D vibration signal as the input and does not need time-frequency conversion. By introducing long short-term memory networks in AE, the time-dependent features of the data can be extracted effectively. The proposed method was verified by five kinds of gears at different pitting degrees under six load conditions. The results indicate that the method can provide simultaneous accurate fault diagnosis and severity detection for different pitting degrees.

With the rapid development of the aviation industry, videoscope inspection of aeroengines has become crucial for ensuring aircraft flight safety. Recently, deep learning, particularly convolutional neural networks (CNNs), have shown remarkable efficacy in videoscope inspection tasks. However, these methods usually require large-scale training labels with accurate annotation. In videoscope images, defects are often present at the micrometer level and require manual labeling by professional inspection personnel, posing further challenges to model training. To address these issues, weakly supervised deep level set (WDLS), a fully automatic framework for aeroengine defect segmentation, was proposed in this work. WDLS employs a multibranch structure and an iterative learning strategy to combine a high-performance CNN architecture with level set evolution. First, the similarity-guided region detector was designed to generate class-specific segmentation proposals and initialize the level set function. Second, the adaptive local parameter prediction algorithm was proposed to integrate local priors and constraints into the energy function and optimize the iterative process. Finally, a self-supervised objective function named convexified level set was introduced to represent an improved level set formulation and obtain the final segmentation. The proposed framework is continuously differentiable and unified. Thus, it can be seamlessly embedded in any CNN without postprocessing. Furthermore, the method was evaluated on a new aeroengine dataset Turbo19 and a benchmark dataset. Experimental results demonstrated that the proposed framework meets real-time requirements and achieves better performance than state-of-the-art methods.

In view of the fact that the mature eddy current sensors on the market have a small measuring range and a certain requirement for the measured conductor area, which cannot meet the requirement of TBM disc cutter Wear monitoring, Ansoft Maxwell simulation was used to optimize the coil geometry parameters and circuit structure of eddy current sensor in this study. A special sensor for measuring disc cutter wear was designed, and the feasibility of the design was verified by measuring a 17-inch disc cutter. The results indicate that the linear range of the eddy current sensor is 35 mm under the condition that the measured blade width is 19 mm, which can meet the requirement of disc cutter wear measurement and provide monitoring data for real-time disc cutter wear monitoring system.

The preheating of the substrate in laser additive superalloys can reduce residual stress and minimize cracking. However, this preheating process can lead to changes in the heat transfer conditions, ultimately affecting the resulting microstructure and mechanical properties. In order to explore the influence of substrate preheating on the formation of laser cladding, this research focuses on investigating the characteristics of Inconel 718, a nickel-based superalloy, as the subject of study. To simulate the temperature and flow field of laser cladding, a 3D computational fluid dynamics (CFD) model is employed. By varying the initial preheating conditions, an investigation is conducted into the distribution of the temperature field under different parameters. This leads to the acquisition of varying temperature gradients, G, and solidification speeds, R. Subsequently, an analysis is carried out on both the flow field and solidification microstructure in the melt pool. The results demonstrate that the preheating of the substrate results in a slower cooling rate, ultimately leading to the formation of a coarser microstructure.

Primary dendrite arm spacing (PDAS) is a crucial microstructural feature in nickel-based superalloys produced by laser cladding. In order to investigate the effects of process parameters on PDAS, a multi-scale model that integrates a 3D transient heat and mass transfer model with a quantitative phase-field model was proposed to simulate the dendritic growth behavior in the molten pool for laser cladding Inconel 718. The values of temperature gradient (G) and solidification rate (R) at the S/L interface of the molten pool under different process conditions were obtained by multi-scale simulation and used as input for the quantitative phase field model. The influence of process parameters on microstructure morphology in the deposition layer was analyzed. The result shows that the dendrite morphology is in good agreement with the experimental result under varying laser power (P) and scanning velocity (V). PDAS was found to be more sensitive to changes in laser scanning velocity, and as the scanning velocity decreased from 12 mm/s to 4 mm/s, the PDAS increased by 197% when the laser power was 1500 W. Furthermore, smaller PDAS can be achieved by combining higher scanning velocity with lower laser power.