Linear Unit Research Articles

An Adaptive Multi-D-Norm-Driven Sparse Unfolding Deconvolutional Network for Bearing Fault Diagnosis.

Impulsive blind deconvolution (IBD) is a popular method to recover impulsive sources for bearing fault diagnosis. Its underpinnings are in the design of objective functions based on prior knowledge of impulsive sources and a transfer function to describe transmission path influences. However, popular objective functions cannot retain waveform impulsiveness and periodicity cyclostationarity simultaneously, and the single convolution operation of IBD methods is insufficient to describe transmission paths composed of multiple linear and nonlinear units. Inspired by the MaxPooling period modulation intensity (MPMI) and convolutional sparse learning (CSL), an adaptive multi-D-norm-driven sparse unfolding deconvolution network (AMD-SUDN) is proposed in this paper. The core strategy is that one target vector with simultaneous impulsiveness and cyclostationarity is constructed automatically through the MPMI; then, this vector is substituted into the multi D-norm to design objective functions. Moreover, an iterative soft threshold algorithm (ISTA) for the CSL model is derived, and its iterative steps are unfolded into one deconvolution network. The algorithm's performance and the hyperparameter configuration are investigated by a set of numerical simulations. Finally, the proposed AMD-SUDN is applied to detect the impulsive features of bearing faults. All comparative results verify that the proposed AMD-SUDN achieves a better deconvolution accuracy than state-of-the-art IBD methods.

GLUENet: An Efficient Network for Remote Sensing Image Dehazing with Gated Linear Units and Efficient Channel Attention

GLUENet: An Efficient Network for Remote Sensing Image Dehazing with Gated Linear Units and Efficient Channel Attention

Defect Detection of Cell Phone Screen Using a Faster Regional Convolutional Neural Network with Multi-head Attention Mechanism

Since the glass outer screen of a cell phone is the main sensory part of the human eye when using a cell phone, the advantages and disadvantages of the cell phone screen directly affect people's sense of use. Therefore, the defect detection requirements for cell phone screens are high and need to meet the needs of high-volume factory inspection. Most of the traditional defect detection methods use visual methods, the detection results are overly dependent on the subjectivity and experience of workers, the efficiency of this method is low, and the accuracy is poor. Currently, machine learning-based detection methods are applied in numerous industries. In this paper, a faster Regional Convolutional Neural Network (R-CNN) with multi-head attention mechanism for defect detection of cell phone screen is proposed. To enhance the network's capability in extracting feature information, a four-head attention mechanism is added to the last convolutional layer of the ResNet50 network. An improved Region of Interest (ROI) Align is proposed to replace the original ROI Pooling to reduce the localization error of cell phone screen defects. Replace the original Rectified Linear Unit (ReLU) activation function with the Copy Exponential Linear Unit (CELU) activation function to expedite the convergence capability of the network. Finally, by comparing with other classical model training, the evaluation results indicate that the proposed method achieved an average accuracy of 95.71%, which is a 5.34% improvement compared to the original faster R-CNN network.

Instantaneous frequency measurement system based on quantum dash mode-locked laser

We present the theory and experimental results of a microwave photonic (MWP) filter based instantaneous frequency measurement system. A quantum dash mode-locked laser is used as an optical frequency comb source. With up to 41 flat comb lines and a real-time feedback loop for comb shaping, a set of MWP filters with linear frequency responses for either linear unit or dB unit are experimentally demonstrated. The maximum measurement frequency can be up to 20 GHz limited by the available test-and-measurement instruments. By using one MWP filter, the root-mean-square error is 51∼66 MHz, which can be improved to 42.2 MHz for linear unit, and 30.7 MHz for dB unit by using two MWP filters together.

Exam Assessor Tool: An Automated System for Efficient Answer Sheet Evaluation

With Education 4.0 and four quadrant approach number of innovations have gone into academics for efficient, experiential, and outcome-based education however assessment schemes are still very much dependent on manual assessment methods which are time-consuming and cumbersome. The grading system can sometimes be irrational, with diversified schemes for the same course and can also be biased. Covid 19 pandemic caused a global economic avalanche like we’ve never experienced in our lifetime. Many countries have implemented control measures such as blockades and curfews. The education system in this chaos saw a silver lining with academics shifting to online mode, with paradigm shift in teaching, assessment techniques too need to evolve. Work done is an effort to ease the process of assessment, a machine learning assisted model is developed that automates subjective answer evaluation in the education sector. Our project involved several crucial steps, including grayscale conversion, Natural Language Processing (NLP) for data cleansing, data splitting, and training an artificial neural network (ANN) to predict scores based on extracted features. ANN-based system grades subjective responses without human intervention, reducing the workload of teachers and professors. Model constructed an ANN architecture with three layers using Rectified Linear Activation Unit (ReLU) and Sigmoid activation functions. Trained model was incorporated into a user-friendly web application using the Streamlit library. Model design gives a major boost in grading efficiency and accuracy while providing valuable feedback to students. Research surveys were conducted, and a dataset was constructed for training and testing the model. study yielded an accuracy of 83.14% after employing techniques such as text cleaning, preprocessing, and feature extraction.

Predicting total biogas potential of food waste using the initial output of biogas potential tests as input data to train an artificial neural network

Quantification of biogas potential is important for predicting anaerobic digestion operability and price. This study uses data from 446 biogas potential tests to train and test a multilayer perceptron artificial neural network (ANN) to forecast total biogas production using the evolution at the start of the experiment (3–14 days) as input data. ANN architecture (training algorithm, activation function, hidden nodes, regularisation, and input data) was optimised using response surface methodology. Best conditions (accuracy/computational speed) were obtained using adaptive moment estimation (adam) training algorithm and rectified linear unit (ReLU) activation function. When using three days of biogas production data, the accuracy of the model was reasonable (r2test = 0.881, r2validation = 0.879), although this increased significantly for 7 days (r2test = 0.953, r2validation = 0.925), or 14 days (r2test = 0.971, r2validation = 0.953). The highest accuracy was reported for readily digestible substrates (sugars and carbohydrates) and macronutrient mixtures. The methodology could be used to shorten prediction times of biogas potential tests.

Regulating the Phase and Optical Properties of Mixed‐Halide Perovskites via Hot‐Electron Engineering

AbstractThe rapid development of mixed‐halide perovskites has established a versatile optoelectronic platform owing to their extraordinary physical properties, but there remain challenges toward achieving highly reliable synthesis and performance, in addition, post‐synthesis approaches for tuning their photoluminescence properties after device fabrication remain limited. In this work, an effective approach is reported to leveraging hot electrons generated from plasmonic nanostructures to regulate the optical properties of perovskites. A plasmonic metasurface composed of Au nanoparticles can effectively tailor both photoluminescence and location‐specific phase segregation of mixed‐halide CsPbI2Br thin films. The ultrafast transient absorption spectroscopy measurements reveal hot electron injection on the timescale of hundreds of femtoseconds. Photocurrent measurements confirm the hot‐electron‐enhanced photon‐carrier conversion, and in addition, gate‐voltage tuning of phase segregation is observed because of correlated carrier injection and halide migration in the perovskite films. Finally, the characteristics of the gate‐modulated light emission are found to conform to a rectified linear unit function, serving as nonlinear electrical‐to‐optical converters in artificial neural networks. Overall, the hot electron engineering approach demonstrated in this work provides effective location‐specific control of the phase and optical properties of halide perovskites, underscoring the potential of plasmonic metasurfaces for advancing perovskite technologies.

Visual Insights in Lung Cancer Prediction: Leveraging Data Visualization and Analysis for Building and Tuning Machine Learning Models

Lung cancer stands as a significant global health challenge with its high occurrence rates and fatal lethality. Given lung cancer's severity and complicated symptoms in later stages, healthcare facilities must detect them early. Traditional diagnostic methodologies are often burdened by high cost and specialized expertise, prompting exploration of alternative methods, such as machine Learning. This paper utilizes a dataset from a Nature Medicine study involving 462,000 participants in China and employs machine learning to predict lung cancer risk. Through the application of Exploratory Data Analysis (EDA), this paper explored the correlation between demographic variables, environmental factors, and lifestyle habits with the probability of lung cancer occurrence. The neural network architecture incorporates dynamic layers, rectified linear unit (ReLU) activation, and the Adam optimizer with dropout regularization. Results from EDA reveal correlations between lifestyle factors and lung cancer risk. The machine learning model achieved 70% accuracy on predictions and was refined to 90% through EDA and utilizing the LIME interpreter. This study aimed to advance lung cancer prediction by applying an amalgamation of techniques and technologies, like data visualization, extensive data analysis, and machine learning. This research aimed to contribute to the showcase of applying the latest technological advancements for the era of big data within the medical research and healthcare industry. The success of this model suggests the viability of creating machine learning models targeted for cancer predictions and indicates further advancements, such as personalized prediction models.

An analytical approach for unsupervised learning rate estimation using rectified linear units.

Unsupervised learning based on restricted Boltzmann machine or autoencoders has become an important research domain in the area of neural networks. In this paper mathematical expressions to adaptive learning step calculation for RBM with ReLU transfer function are proposed. As a result, we can automatically estimate the step size that minimizes the loss function of the neural network and correspondingly update the learning step in every iteration. We give a theoretical justification for the proposed adaptive learning rate approach, which is based on the steepest descent method. The proposed technique for adaptive learning rate estimation is compared with the existing constant step and Adam methods in terms of generalization ability and loss function. We demonstrate that the proposed approach provides better performance.

Deep learning-based image annotation for leukocyte segmentation and classification of blood cell morphology

The research focuses on the segmentation and classification of leukocytes, a crucial task in medical image analysis for diagnosing various diseases. The leukocyte dataset comprises four classes of images such as monocytes, lymphocytes, eosinophils, and neutrophils. Leukocyte segmentation is achieved through image processing techniques, including background subtraction, noise removal, and contouring. To get isolated leukocytes, background mask creation, Erythrocytes mask creation, and Leukocytes mask creation are performed on the blood cell images. Isolated leukocytes are then subjected to data augmentation including brightness and contrast adjustment, flipping, and random shearing, to improve the generalizability of the CNN model. A deep Convolutional Neural Network (CNN) model is employed on augmented dataset for effective feature extraction and classification. The deep CNN model consists of four convolutional blocks having eleven convolutional layers, eight batch normalization layers, eight Rectified Linear Unit (ReLU) layers, and four dropout layers to capture increasingly complex patterns. For this research, a publicly available dataset from Kaggle consisting of a total of 12,444 images of four types of leukocytes was used to conduct the experiments. Results showcase the robustness of the proposed framework, achieving impressive performance metrics with an accuracy of 97.98% and precision of 97.97%. These outcomes affirm the efficacy of the devised segmentation and classification approach in accurately identifying and categorizing leukocytes. The combination of advanced CNN architecture and meticulous pre-processing steps establishes a foundation for future developments in the field of medical image analysis.

Enhanced Solubility and Bioavailability of Clotrimazole in Aqueous Solutions with Hydrophobized Hyperbranched Polyglycidol for Improved Antifungal Activity.

The poor solubility of clotrimazole in the aqueous medium and the uncontrolled removal of the drug-loaded suppository content limit its effectiveness in the treatment of vulvovaginal candidiasis. We present here the aqueous formulations of clotrimazole in the form of non-Newtonian structured fluids, i.e., Bingham plastic or pseudoplastic fluids constructed of hyperbranched polyglycidol, HbPGL, with a hydrophobized core with aryl groups such as phenyl or biphenyl. The amphiphilic constructs were obtained by the modification of linear units containing monohydroxyl groups with benzoyl chloride, phenyl isocyanate, and biphenyl isocyanate, while the terminal 1,2-diol groups in the shell were protected during the modification step, followed by their deprotection. The encapsulation of clotrimazole within internally hydrophobized HbPGLs using a solvent evaporation method followed by water addition resulted in structured fluids formation. Detailed Fourier-transform infrared spectroscopy (FTIR) and differential scanning calorimetry (DSC) analyses performed for aryl-HbPGLs with clotrimazole revealed the difference in drug compatibility among polymers. Clotrimazole in biphenyl-enriched HbPGL, unlike phenyl derivatives, was molecularly distributed in both the dry and the hydrated states, resulting in transparent formulations. The shear-thinning properties of the obtained fluid formulations make them injectable and thus suitable for the intravaginal application. Permeability tests performed with the usage of the Franz diffusion cell showed a 5-fold increase in the permeability constant of clotrimazole compared to drugs loaded in a commercially available disposable tablet and a 50-fold increase of permeability in comparison to the aqueous suspension of clotrimazole. Furthermore, the biphenyl-modified HbPGL-based drug liquid showed enhanced antifungal activity against both Candida albicans and Candida glabrata that was retained for up to 7 days, in contrast to the phenyl-HbPGL derivatives and the tablet. With their simple formulation, convenient clotrimazole/biphenyl-HbPGL formulation strategy, rheological properties, and enhanced antifungal properties, these systems are potential antifungal therapeutics for gynecological applications. This study points in the synthetic direction of improving the solubility of poorly water-soluble aryl-enriched pharmaceuticals.

Demonstrating a new evaluation method on ReLU based Neural Networks for classification problems

Deep neural networks, which have proven to be effective methods to solve complex problems, can even be applied in decision systems controlling critical processes. However, in such applications the outcomes of the neural network must be checked if it we have a clear understanding regarding the operation of network at given input intervals. The most straightforward, though often computationally expensive approach for this checking process evaluates the network at discrete input points and estimates the expected outputs at the given interval. The present research aims to develop a novel approach that can identify in the case of specific input intervals whether the operation process and the output of a neural network can be considered as known or unknown. During the presented case study, we investigated the ReLU (Rectified Linear Unit) and Sigmoid activation functions, using the double moon and the Banknote Authentication classification problems for demonstration.Our method can be applied to identify certain input intervals where the given neural network cannot support critical decisions. The evaluation is performed based on a nonlinear system of equations and inequalities built on arbitrary continuous activation functions. To define the critical intervals of the input variables (i.e. where the decision-making system should not be relied on), those input variable combinations are identified which result in a non-expected output value. This inverse logic is intended to identify intervals of the input variables where the response of the system and the correct decision are not identical. The presented demonstration examples supported our assumption that the number of the neurons and the dimension of the decision space have a significant impact on the complexity of the evaluation process.

BrutNet: A novel approach for violence detection and classification using DCNN with GRU

AbstractAutomatic Violence Detection and Classification (AVDC) with deep learning has garnered significant attention in computer vision research. This paper presents a novel approach for combining a custom Deep Convolutional Neural Network (DCNN) with a Gated Recurrent Unit (GRU) in developing a new AVDC model called BrutNet. Specifically, a time‐distributed DCNN (TD‐DCNN) is developed to generate a compact 2D representation with 512 spatial features per frame from a set of equally‐spaced frames of dimension 16090 in short video segments. Further to leverage the temporal information, a GRU layer is utilised, generating a condensed 1D vector that enables binary classification of violent or non‐violent content through multiple dense layers. Overfitting is addressed by incorporating dropout layers with a rate of 0.5, while the hidden and output layers employ rectified linear unit (ReLU) and sigmoid activations, respectively. The model is trained on the NVIDIA Tesla K80 GPU through Google Colab, demonstrating superior performance compared to existing models across various video datasets, including hockey fights, movie fights, AVD, and RWF‐2000. Notably, the model stands out by requiring only 3.416 million parameters and achieving impressive test accuracies of 97.62%, 100%, 97.22%, and 86.43% on the respective datasets. Thus, BrutNet exhibits the potential to emerge as a highly efficient and robust AVDC model in support of greater public safety, content moderation and censorship, computer‐aided investigations, and law enforcement.

Tool Wear Prediction Based on LSTM and Deep Residual Network

To improve the accuracy and efficiency of tool wear predictions, this study proposes a tool wear prediction model called LSTM_ResNet which is based on the long short-term memory (LSTM) network and the Residual Network (ResNet). The model utilizes LSTM layers for processing, where the first block and loop blocks serve as the core modules of the deep residual network. The model employs a series of methods including convolution, batch normalization (BN), and Rectified Linear Unit (ReLU) to enhance the model’s expression and prediction capabilities. The performance of the LSTM_ResNet model was evaluated using experimental data from the PHM2010 datasets and two different depths (64 and 128 layers), training both LSTM_ResNet models for 200 epochs. The 64-layer model’s root mean square error (RMSE) values are 3.36, 4.35, and 3.59, and the mean absolute error (MAE) values are 2.42, 2.85, and 2.21; using 128 layers, the RMSE values are 3.66, 3.99, and 3.77, and the MAE values are 2.49, 2.73, and 3.01. The results indicate that the 64-layer LSTM has smaller average errors, suggesting that compared to other common network structures, the LSTM_ResNet network has a higher performance. This research provides an effective solution for tool wear prediction and helps to improve the technical level of tool wear prediction in China.

OOA-modified Bi-LSTM network: An effective intrusion detection framework for IoT systems

Currently, the Internet of Things (IoT) generates a huge amount of traffic data in communication and information technology. The diversification and integration of IoT applications and terminals make IoT vulnerable to intrusion attacks. Therefore, it is necessary to develop an efficient Intrusion Detection System (IDS) that guarantees the reliability, integrity, and security of IoT systems. The detection of intrusion is considered a challenging task because of inappropriate features existing in the input data and the slow training process. In order to address these issues, an effective meta heuristic based feature selection and deep learning techniques are developed for enhancing the IDS. The Osprey Optimization Algorithm (OOA) based feature selection is proposed for selecting the highly informative features from the input which leads to an effective differentiation among the normal and attack traffic of network. Moreover, the traditional sigmoid and tangent activation functions are replaced with the Exponential Linear Unit (ELU) activation function to propose the modified Bi-directional Long Short Term Memory (Bi-LSTM). The modified Bi-LSTM is used for classifying the types of intrusion attacks. The ELU activation function makes gradients extremely large during back-propagation and leads to faster learning. This research is analysed in three different datasets such as N-BaIoT, Canadian Institute for Cybersecurity Intrusion Detection Dataset 2017 (CICIDS-2017), and ToN-IoT datasets. The empirical investigation states that the proposed framework obtains impressive detection accuracy of 99.98 %, 99.97 % and 99.88 % on the N-BaIoT, CICIDS-2017, and ToN-IoT datasets, respectively. Compared to peer frameworks, this framework obtains high detection accuracy with better interpretability and reduced processing time.

Infrared image target detection for substation electrical equipment based on improved faster region-based convolutional neural network algorithm.

Substation electrical equipment generates a massive number of infrared images during operation. However, the overall quality of the infrared images is low and it lacks image detail information. When using traditional target detection algorithms for detection, feature extraction poses great difficulties. Therefore, to address this problem, this paper proposes a target detection algorithm based on the improved faster region-based convolutional neural network (Faster R-CNN). It achieves the correct identification of different types of electrical equipment in infrared images. First, the algorithm improves the backbone network of Faster R-CNN for feature extraction. An InResNet structure is proposed to replace the residual block structure of the original ResNet-34 network, which enhances the richness of feature extraction. Second, the rectified linear unit activation function in the original feature extraction network is replaced by the exponential linear unit activation function, and group normalization is used instead of batch normalization as the network normalization method. Then, the dense connection structure is introduced into the ResNet-34 network, and the whole network is called residual dense connection network. Finally, the improved Faster R-CNN is compared to the original Faster R-CNN, a single-shot multibox detector, and you only look once v3 plus spatial pyramid pooling. The experimental results show that the improved algorithm has the highest mean average precision and average recall for most of the substation electrical equipment in infrared images. Moreover, from the confidence level of the detected electrical equipment and the accuracy of the prediction box, the improved Faster R-CNN has the best performance.

Deep learning model for classifying shoulder pain rehabilitation exercises using IMU sensor

BackgroundArtificial intelligence is being used for rehabilitation, including monitoring exercise compliance through sensor technology. AI classification of shoulder exercise wearing an IMU sensor has only been reported in normal (i.e. painless) subjects. To prove the feasibility of monitoring exercise compliance, we aimed to classify 11 types of shoulder rehabilitation exercises using an AI (artificial intelligence) algorithm in patients with shoulder pain. We had the patients wear an IMU-based sensor, collected data during exercise, and determined the accuracy of exercise classification.MethodsData were collected from 58 patients (27 males, 31 females, age range 37–82 years) diagnosed with shoulder diseases such as adhesive capsulitis and rotator cuff disease. 11 types of shoulder pain rehabilitation exercise programs were developed and repeated each exercise ten times per session while wearing an IMU sensor. The study applied the Rectified Linear Unit (ReLU) and the SoftMax as the activation function for hidden layers, the output layer.ResultsThe acquired data was used to train a DNN model using the multilayer perceptron algorithm. The trained model was used to classify 11 types of shoulder pain rehabilitation exercises. The training accuracy was 0.975 and the test accuracy was 0.925.ConclusionThe study demonstrates that IMU sensor data can effectively classify shoulder pain rehabilitation exercises, providing more appropriate feedback for patients. The model can be utilized to establish a system for remotely monitoring patients’ exercise performance. The use of deep learning in patient monitoring and rehabilitation has significant potential to bring innovative changes to healthcare service delivery.

Testing the Validity of the Purchasing Power Parity (PPP) Hypothesis for Türkiye: Linear and Nonlinear Unit Root Tests

In this paper, we test the validity of the purchasing power parity (PPP) hypothesis between Türkiye and its trading partners the European Union, China, and the US for the period from January 2001 to January 2020. We test the stationarity of the real exchange rates for the validity of the PPP hypothesis by applying linear and nonlinear unit root tests. We also employ Fourier-based tests to account for the structural changes that occurred in the considered period. Test results indicate that shocks are temporary, and the PPP hypothesis is valid for Türkiye.

Easily accessible linear and hyperbranched polyesters as solid polymer electrolytes

Herein we report synthesis, characterization, thermal properties, ionic conductivity and oxidative stability of novel linear and hyperbranched polyesters. Tri-acrylate ester of commercially available 1,1,1-Tris(hydroxymethyl)propane (M1) was used as the primary building block for the synthesis of linear and hyperbranched polymers. A3 + B2 type polycondensation between M1 and dithiols of two different chain lengths (C3, C6) by 100 % atom efficient thiol-acrylate Michael addition reaction, followed by consumption of the unreacted acrylate esters (in the linear or terminal units) with 1-butane thiol produced two hyperbranched polymer namely HB-P1 and HB-P2. For synthesis of linear analogues, one of the equivalent acrylate esters was reacted first with 1-butane thiol and the resulting di-acrylate monomer was polymerized with the C3 and C6-dithiols, producing linear polymers L-P1 and L-P2. All the polymers showed molecular weight (Mn) in the range of 5000–6000 gmol−1 with low dispersity. TGA analysis revealed sufficient thermal stability of the polymers for the application as solid polymer electrolyte. All the polymers are amorphous, showing only a glass transition in the range of ∼ -45 °C (C3 spacer) to −50 °C (C6 spacer) and no crystallization peak. In presence of lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) salt the Tg increased by ∼ 10 °C. Solid polymer electrolytes were prepared by incorporating LiTFSI at a molar ratio of polymer carbonyl units to lithium of C=O/Li of 10 and ionic conductivity (σ) was evaluated by electrochemical impedance spectroscopy (EIS). Moderate values in the range of 10-7 S cm−1 were obtained at 25 °C with slightly higher values for the linear polymers. The ionic conductivity increased up to ∼ 2 orders of magnitude at higher temperatures. The oxidative stability of the polymer electrolytes against lithium-metal electrodes revealed onset potential for the first degradation in the range of 3.8 V, indicating moderate stability that might be suitable for testing applications in all-solid-state lithium-metal batteries using sulfur or lithium iron phosphate cathodes.

A small sample bearing fault diagnosis method based on novel Zernike moment feature attention convolutional neural network

Bearings are one of the core components of rotating machine machinery. Monitoring their health status can ensure the safe and stable operation of rotating machine equipment. The limited nature of bearing fault samples makes it difficult to meet the demand for sufficient samples based on deep learning methods. Therefore, how to solve the problem of small- samples is the key to achieving intelligent fault diagnosis. In bearing failures based on vibration signals, the complex operating environment causes the vibration signals to inevitably mix with noise. The mixing of fault signature features and noise intensifies the strong spatial coupling of different types of fault features. In addition, diagnosing bearing failures under different loads is challenging because of the complex working conditions of bearings. Given the above problems, a small sample-bearing fault diagnosis method based on a high and low-frequency layered algorithm (HLFLA) and a novel Zernike moment feature attention convolutional neural network (ZMFA-CNN) is proposed. First, the proposed HLFLA converts one-dimensional time series signals into two-dimensional signals distributed rectangularly according to different frequency bands, and is used to simplify network feature screening, reduce the impact of noise, and retain adjacent signal constraint information. In addition, a new ZMFA-CNN is proposed to further extract multi-order moment features and attention weights, and can significantly improve the model generalization ability without increasing model parameters. At the same time, it is combined with FilterResponseNorm2d and thresholded linear unit to further improve model performance. Finally, sufficient experiments verified that the algorithm proposed in this paper can solve the above problems and has excellent transfer generalization ability and noise robustness. In addition, the experimental results of applying the algorithm proposed in this article to gas turbine main bearing fault diagnosis prove the reliability of the algorithm proposed in this article.