Threshold Modulation Research Articles

Unsupervised Learning for Machinery Adaptive Fault Detection Using Wide-Deep Convolutional Autoencoder with Kernelized Attention Mechanism

Applying deep learning to unsupervised bearing fault diagnosis in complex industrial environments is challenging. Traditional fault detection methods rely on labeled data, which is costly and labor-intensive to obtain. This paper proposes a novel unsupervised approach, WDCAE-LKA, combining a wide kernel convolutional autoencoder (WDCAE) with a large kernel attention (LKA) mechanism to improve fault detection under unlabeled conditions, and the adaptive threshold module based on a multi-layer perceptron (MLP) dynamically adjusts thresholds, boosting model robustness in imbalanced scenarios. Experimental validation on two datasets (CWRU and a customized ball screw dataset) demonstrates that the proposed model outperforms both traditional and state-of-the-art methods. Notably, WDCAE-LKA achieved an average diagnostic accuracy of 90.29% in varying fault scenarios on the CWRU dataset and 72.89% in the customized ball screw dataset and showed remarkable robustness under imbalanced conditions; compared with advanced models, it shortens training time by 10–26% and improves average fault diagnosis accuracy by 5–10%. The results underscore the potential of the WDCAE-LKA model as a robust and effective solution for intelligent fault diagnosis in industrial applications.

Network traffic anomaly detection model based on feature grouping and multi‐autoencoders integration

AbstractThis paper presents a network traffic anomaly detection model based on feature grouping and multiple autoencoders (multi‐AEs) integration. This model comprises four modules: feature grouping module, feature learning module, AUC and optimal threshold calculation module, and anomaly detection application module. In the feature grouping module, multiple group features are constructed by selecting the different features according to their attributes and variances. In the feature learning module, the group features of normal traffic are learned based on multi‐AEs. In the AUC and optimal threshold calculation module, the AUC of each AE is calculated according to the ROC curve of the verification data, and the optimal thresholds for each AE are determined using the Youden index. In the anomaly detection application module, the AEs that participated in fusion are selected and their weights are calculated by analysing AUC value, and the scores of unknown traffic in each AE are evaluated considering both the reconstruction error distribution and the optimal threshold. Finally, the anomaly detection result can be obtained by the fusion of these multiple scores. Through validation on the UNSW‐NB15 and CICIDS2017 datasets, the accuracy of the proposed model is improved by 12.04% and 10.52%, respectively, compared to the baseline model.

Enhancing the Segmentation Accuracy of the Otsu Method for Calculating Blood Vessel Rupture Volumes in CT-Scan Images

The precise quantification of cerebral hemorrhage volume assumes paramount significance, given its consequential impact on subsequent medical interventions. While antecedent scholarly endeavors have undertaken cerebral hemorrhage volume computations, the attendant accuracy remains amenable to refinement. The main contribution to this research is the development of a rectangle algorithm using the Otsu method, which can be used to calculate the total volume of brain hemorrhage in CT-Scan images of the human brain. The present investigation seeks to augment the segmentation paradigm inherent to the Otsu method, thereby advancing the precision of cerebral hemorrhage volume assessments. The Otsu method is conceived with the purpose of automating the segmentation of the hemorrhagic domain through the determination of threshold values. The refinement of the Otsu method’s accuracy is effectuated by the formulation of an algorithm aimed at maximizing the mean discrepancy between object and background intensities within the object image, predicated upon the ascertained threshold values. These determined thresholds subsequently govern the transformation of grayscale images into binary format, achieved through the quotient of the threshold value and the mean variance, thereby amplifying segmentation precision. The incorporation of variance and mean values in the binarization of grayscale images affords adaptability in threshold modulation commensurate with the intensity spectrum of the image. The research corpus leverages CT-Scan images sourced from RSUP M. Djamil Padang, West Sumatra, encompassing a compendium of 100 image slices culled from a quintet of patients. Empirical findings attest to the fortification of the Otsu method in the segmentation milieu, substantiating the augmentation of threshold value accuracy. This enhancement is conspicuously manifest in the outcomes of segmentation tests, underscoring the superior efficacy of the Otsu method in discerning between objects and backgrounds. The corollary effect is a discernibly more accurate computation of volumes employing the Otsu-T methodology. The import of this research rests in its successful establishment of automated threshold values and the concomitant elevation of segmentation precision.

Deep learning techniques for quality of transmission estimation in optical networks

A large body of research has recently examined the estimation of the quality of transmission (QoT) in optical networks with deep learning. This paper discusses a lightpath’s quality of transmission to design fiber-optic communication and networks using deep learning algorithms. We need different major estimation parameters for advanced optical fiber communication and networks, i.e., modulation formats, baud rate, and code rate. Currently, the quality of transmission for unspecified optical paths depends on different estimation techniques i.e., (1) analytical models estimating physical layer impairments (PLIs) and (2) margined formulas. This paper focuses on deep-learning techniques that can be applied to optimization and complex systems. The deep learning algorithms contain different classifiers that can simulate results and estimate the bit-error rate, and signal-to-noise ratio of unspecified optical paths with threshold values, traffic volume, and modulation format. We must train and test the datasets for various classifiers, and classification features using Korean network topology. The classifier accuracy and Area Under the ROC Curve (AUC) simulation results are carried out using MATLAB.

Pain sensitization and pain-related psychological factors in patients with temporomandibular disorders: an observational cross-sectional study.

There is conflicting evidence on how central processing impairments affect patients with temporomandibular disorders (TMD). Moreover, there is sparse research on the assessment of endogenous pain modulation in this population through conditioned pain modulation (CPM) testing. The main objective of this observational study was to evaluate the possible differences between myofascial TDM patients and healthy pain-free controls on psychophysical variables suggestive of central processing impairments (including temporal summation (TSP), pressure pain threshold (PPT) and conditioned pain modulation (CPM)). This is a cross-sectional observational study including a sample of patients with TMD and pain-free controls recruited from private and university clinics in Spain. Outcome measures included local and distal PPTs, temporal summation, conditioned pain modulation and psychological factors of depression, anxiety, kinesiophobia, fear avoidance beliefs and pain catastrophizing. Fifty-nine patients with TMD of myofascial origin (32 years [IR: 25-43]) and 30 healthy, pain-free controls (29.5 years [IR: 25-41]) participated in the study and completed the evaluations. Patients with TMD showed significantly reduced CPM (p = 0.001; t = 3.31) and both local and distal PPTs (p < 0.05) when compared with controls, after adjusting for the influence of age and sex. TSP did not show any difference between the groups (p = 0.839; Z = 0.20). All psychological factors were higher in patients with TMD (p < 0.005), except for anxiety (p = 0.134). Patients with myofascial TMD included in this study exhibited signs of altered central processing, linked to impaired descending pain modulation, distal hyperalgesia and psychological factors like depression, kinesiophobia, fear avoidance beliefs and pain catastrophizing but not anxiety.

Value of the central sensitisation inventory in patients with axial spondyloarthritis.

In many patients with axial spondyloarthritis (axSpA), pain persists despite anti-inflammatory medication. Quantitative sensory testing (QST) indirectly assesses altered somatosensory function, though its clinical practicality is limited. The Central Sensitisation Inventory (CSI) could be an alternative in the initial assessment of central sensitisation (CS). This study aimed to investigate the value of the CSI in evaluating CS in patients with axSpA by (1) assessing somatosensory function related to CS with QST and (2) exploring associations between CSI, QST, patient and disease characteristics and pain-related psychosocial factors. Consecutive outpatients from the Groningen Leeuwarden AxSpA cohort underwent QST, including pressure pain threshold (PPT), temporal summation (TS) and conditioned pain modulation (CPM). Participants completed questionnaires assessing CS (CSI), illness perception (Revised Illness Perception Questionnaire, IPQ-R), pain-related worrying (Pain Catastrophising Scale, PCS), fatigue (Modified Fatigue Impact Scale, MFIS), anxiety/depression (Hospital Anxiety and Depression Scale, HADS) and coping. QST measurements were stratified for CSI≥40. 201 patients with axSpA were included; 63% male, 64% radiographic axSpA, median symptom duration 12 years (IQR 5-24), mean Axial Spondyloarthritis Disease Activity Score 2.1±1.0. Patients with CSI≥40 had significantly lower PPTs and higher TS than CSI<40 (p<0.004). No significant differences in CPM were observed. In multivariable linear regression, sex, PCS, IPQ-R Identity, MFIS and HADS anxiety were independently associated with CSI (78% explained variance). In this large cross-sectional study in patients with axSpA, the CSI appears as a useful initial CS assessment questionnaire. When CSI scores indicate CS, considering pain-related psychosocial factors is important. These results emphasise the need for a biopsychosocial approach to manage chronic pain in patients with axSpA.

Altered Sensory Processing in People Attending Specialist Orthopaedic Consultation for Management of Persistent Shoulder Pain: An Observational Cross-Sectional Study.

OBJECTIVES: The primary objective was to compare sensory processing measures in people attending specialist orthopaedic consultation for management of persistent shoulder pain with control participants. The secondary objective was to compare the groups' sociodemographic, clinical, general health and lifestyle, and psychological characteristics. DESIGN: Observational cross-sectional. METHODS: Participants with shoulder pain for ≥3 months, who attended a public hospital orthopaedic department (n = 119), and community participants without shoulder pain (n = 44) underwent a standardized quantitative sensory testing protocol, measuring pressure pain threshold, temporal summation, and conditioned pain modulation. Sociodemographic, clinical, general health and lifestyle, and psychological characteristics were also collected. RESULTS: Participants with shoulder pain had significantly lower pressure pain thresholds at all sites (ie, local and widespread mechanical hyperalgesia) and significantly decreased conditioned pain modulation effect (ie, descending inhibition of nociception) than control participants. There was no significant difference between groups for temporal summation. Participants with shoulder pain had decreased general health and function, less healthy lifestyles, and poorer psychological health compared with controls. CONCLUSION: People referred to specialist orthopaedic care for management of persistent shoulder pain had clinical signs of altered sensory processing and poor health outcomes. J Orthop Sports Phys Ther 2024;54(10):1-10. Epub 25 July 2024. doi:10.2519/jospt.2024.12512.

An electronic energy compensation method for flattening the energy response of SiPM-GGAG:Ce,B based gamma detector

Abstract The energy response of gross gamma dose rate monitors needs to be flat in order to prevent overestimation of dose at low gamma energies. In this paper, a discriminator threshold modulation based electronic energy compensation algorithm has been proposed for SiPM-scintillator based gamma detectors. Theoretical simulation studies were carried out in order to optimize the parameters of the periodic ramp voltage used for modulation of the discriminator threshold of a SiPM-GGAG:Ce,B based gamma dose rate monitor. A customized threshold modulation circuit and signal processing electronics were developed for this gamma detector. For experimentally optimizing the parameters, the energy response studies of the detector, with and without the discriminator threshold modulation, were carried out. With the optimized parameters for a periodic ramp threshold, the count rates for 241Am (60 keV) and 60Co (1173 and 1332 keV) were observed to be within ±30% of the count rate obtained for 137Cs (662 keV). Using the electronic energy compensation techniques presented in this paper, a flat energy response of the SiPM-scintillator gamma detector for the energy range of 60 keV to 1.5 MeV could be achieved.

Reliability of Pressure Pain Threshold (PPT) and Conditioned Pain Modulation (CPM) in Participants with and without Chronic Shoulder Pain.

The objectives of this study were to estimate the intra-rater and inter-rater reliability of the Pressure Pain Threshold (PPT) and Conditioned Pain Modulation (CPM) in healthy participants and patients with chronic shoulder pain. Additionally, the Standard Error of Measurement (SEM) and Smallest Detectable Change (SDC) were calculated. Thirty-one healthy volunteers and twenty patients with chronic shoulder pain were assessed using the PPT and CPM by two raters, with a 24 h interval between sessions. Excellent intra-rater reliability was demonstrated for PPT, with similar SEM and SDC when assessed by the same rater. The inter-rater reliability for PPTs in patients was moderate to good (ICC = 0.59-0.89) with higher SEM (73.83-121.98 kPa) and SDC (61.58-97.59) values than the asymptomatic group (ICC = 0.92-0.96, SEM = 49.61-103.12 kPa, SDC = 42.01-56.30) respectively. CPM's intra-rater reliability was good (ICC = 0.82) in the patients and moderate (ICC = 0.67) in the asymptomatic group, while inter-rater reliability was low for the asymptomatic group (ICC = 0.37) and extremely low (ICC = 0.074) for the patients, with comparable SEM and SDC outcomes in both groups. PPT and CPM measurements are highly reliable when conducted by the same rater on the same day. Patients had lower inter-rater PPT reliability but better intra-rater CPM reliability. Clinicians need to be mindful of potential variability when interpreting these test results.

Patch ModernTCN-Mixer: a dual-task temporal convolutional network framework for hybrid implementation of first prediction time detection and remaining useful life prognosis

Abstract Over the past few years, notable advancements have been achieved in predicting the remaining useful life (RUL) of rotating equipment through deep learning methodologies. However, existing RUL prediction models tend to implement the determination of the first prediction time (FPT) for stage division separately from the RUL prediction, ignoring their potential correlation in the degradation process. In response to this issue, this paper proposed a dual-task prediction network framework based on Patch ModernTCN-Mixer (PMTCN-Mixer), which adaptively and jointly achieved FPT detection and RUL prediction. Firstly, the network designed a hard sharing parameter feature extractor module, Patch ModernTCN, which is used to learn the temporal dependence and spatial correlation of degradation features. Secondly, to eliminate redundant information and noise during the feature extraction phase while enhancing the precision of detection and prognosis, a dynamic semi-soft thresholding (DST) module was constructed. Lastly, the dual-task learning network PMTCN-Mixer was constructed by combining Patch ModernTCN with DST, utilizing GradNorm to balance the gradients between FPT detection and RUL prediction tasks to achieve fusion prediction. The performance of the PMTCN-Mixer framework was validated on the XJTU-SY Bearing Datasets and IEEE PHM 2012 Challenge Datasets, compared with the state-of-the-art network’s optimal results, the RUL prediction metrics root mean square error, mean absolute error, Score and R 2 were improved by 17.31%, 23.59%, 10.66%, and 4.76%, respectively. The findings confirm that the dual-task prediction model PMTCN-Mixer effectively captures both spatial and temporal semantic information from deterioration data, precisely accomplishes the integration of FPT detection and RUL prediction, and possess good generalization ability and superiority.

Effect of modulation factor and low dose threshold level on gamma pass rates of single isocenter multi-target SRT treatment plans.

SRS MapCHECK (SMC) is a commercially available patient-specific quality assurance (PSQA) tool for stereotactic radiosurgery (SRS) applications. This study investigates the effects of degree of modulation, location off-axis, and low dose threshold (LDT) selection on gamma pass rates (GPRs) between SMC and treatment planning system, Analytical Anisotropic Algorithm (AAA), or Vancouver Island Monte Carlo (VMC++ algorithm) system calculated dose distributions. Volumetric-modulated arc therapy (VMAT) plans with modulation factors (MFs) ranging from 2.7 to 10.2 MU/cGy were delivered to SMC at isocenter and 6cm off-axis. SMC measured dose distributions were compared against AAA and VMC++ via gamma analysis (3%/1mm) with LDT of 10% to 80% using SNC Patient software. Comparing on-axis SMC dose against AAA and VMC++ with LDT of 10%, all AAA-calculated plans met the acceptance criteria of GPR≥90%, and only one VMC++ calculated plan was marginally outside the acceptance criteria with pass rate of 89.1%. Using LDT of 80% revealed decreasing GPR with increasing MF. For AAA, GPRs reduced from 100% at MF of 2.7 MU/cGy to 57% at MF of 10.2 MU/cGy, and for VMC++ calculated plans, the GPRs reduced from 89% to 60% in the same MF range. Comparison of SMC dose off-axis against AAA and VMC++ showed more pronounced reduction of GPR with increasing MF. For LDT of 10%, AAA GPRs reduced from 100% to 83% in the MF range of 2.7 to 9.8 MU/cGy, and VMC++ GPR reduced from 100% to 91% in the same range. With 80% LDT, GPRs dropped from 100% to 42% for both algorithms. MF, dose calculation algorithm, and LDT selections are vital in VMAT-based SRT PSQA. LDT of 80% enhances sensitivity of gamma analysis for detecting dose differences compared to 10% LDT. To achieve better agreement between calculated and SMC dose, it is recommended to limit the MF to 4.6 MU/cGy on-axis and 3.6 MU/cGy off-axis.

The interactive effects of different exercises and hawthorn consumption on the pain threshold of TMT-induced Alzheimer male rats

Exercise increases the pain threshold in healthy people. However, the pain threshold modulation effect of exercise and hawthorn is unclear because of its potential benefits in people with persistent pain, including those with Alzheimer's disease. Accordingly, after the induction of Alzheimer's disease by trimethyl chloride, male rats with Alzheimer's disease were subjected to a 12-week training regimen consisting of resistance training, swimming endurance exercises, and combined exercises. In addition, hawthorn extract was orally administered to the rats. Then, their pain threshold was evaluated using three Tail-flick, Hot-plate, and Formalin tests. Our results showed that Alzheimer's decreased the pain threshold in all three behavioral tests. Combined exercise with hawthorn consumption had the most statistically significant effect on Alzheimer's male rats' pain threshold in all three experiments. A combination of swimming endurance and resistance exercises with hawthorn consumption may modulate hyperalgesia in Alzheimer's rats. Future studies need to determine the effects of these factors on the treatment and/or management of painful conditions.Graphical

Frequency switching leads to distinctive fast–slow behaviors in Duffing system

The slowly forced Duffing system has been found to exhibit unique fast–slow behaviors in descending frequency switching scheme, the typical of which is the sliding bursting, which is instructive for understanding the dynamics of ubiquitous frequency switching systems in scientific research and practical engineering. This paper is devoted to further refining the fast–slow dynamics of the slowly forced Duffing system with another commonly encountered frequency switching scheme, i.e., the ascending frequency switching, characterized by switching the frequency synchronously according to the increase or decrease of state variable values. Taking the forcing amplitude as an example, this paper provides a theoretical way to fully summarize the fast–slow behaviors in frequency switching systems with the variation of parameter values based on the proposed superposition analysis of one- and two-parameter bifurcations of subsystems. As a result, two typical bifurcation structures and eight threshold windows with distinct switching vector fields therein are identified, inducing up to ten different bursting patterns. Among them, several novel fast–slow dynamics, such as multiple jumps hysteresis loop formed by boundary equilibrium bifurcations and the switching failure phenomena, are presented and investigated. In particular, the underlying mechanism of threshold modulation, i.e., the evolution of unconventional bifurcations, is also found. These findings contribute to complementing and contrasting the existing studies on the fast–slow dynamics of frequency switching systems.

Joint Optimization of Crack Segmentation With an Adaptive Dynamic Threshold Module

Joint Optimization of Crack Segmentation With an Adaptive Dynamic Threshold Module

DRLSTM-DA: a novel aeroengine remaining useful life prediction method based on deep optimization of multiple sensors information

Abstract One of the crucial tasks in Remaining Useful Life (RUL) prediction is to effectively extract key sensor information from numerous sensor signals. In this article, a novel dual-attention enhanced deep residual LSTM (DRLSTM-DA) is developed to deeply optimize multidimensional sensor signals. First, two LSTM layers are designed to compress and reconstruct representative degradation information from input multidimensional time series data, to generate a new feature space. Second, a novel channel adaptive soft threshold module is designed to assign different weights according to the importance of different sensor information, and simultaneously eliminate the noise information in the signal. Thirdly, a temporal attention mechanism is designed to automatically highlight moments containing important decay information while suppressing unimportant moments. Finally, compared with other SOTA methods, our architecture achieves RMSE of 11.55, 13.74, 11.25, and 14.19 on four sub-datasets of the C-MAPSS dataset, with scores of 234.24, 465.49, 202.23, and 537.66, respectively. Meanwhile, in real aeroengine operation dataset, our architecture achieved the smallest RMSE (8.62). These results all validate the good predictive performance of our model.

Charge trapping layer enabled high-performance E-mode GaN HEMTs and monolithic integration GaN inverters

In this work, high threshold voltage and breakdown voltage E-mode GaN HEMTs using an Al:HfOx-based charge trapping layer (CTL) are presented. The developed GaN HEMTs exhibit a wide threshold modulation range of ΔVTH ∼ 17.8 V, which enables the achievement of enhancement-mode (E-mode) operation after initialization process owing to the high charge storage capacity of the Al:HfOx layer. The E-mode GaN HEMTs exhibit a high positive VTH of 8.4 V, a high IDS,max of 466 mA/mm, a low RON of 10.49 Ω mm, and a high on/off ratio of ∼109. Moreover, the off-state breakdown voltage reaches up to 1100 V, which is primarily attributed to in situ O3 pretreatment effectively suppressing and blocking leakage current. Furthermore, thanks to the VTH of GaN HEMTs being tunable by initialization voltage using the proposed CTL scheme, we prove that the direct-coupled FET logic-integrated GaN inverters can operate under a variety of conditions (β = 10–40 and VDD = 3–15 V) with commendable output swing and noise margins. These results present a promising approach toward realizing the monolithic integration of GaN devices for power IC applications.

Estimation of electric field inside a neural spheroid by low‐frequency magnetic field exposure

AbstractExposure to time‐varying, low‐frequency and high‐intensity magnetic field (MF) induce electric field (EF) inside the human body, producing stimulus effects such as nerve fiber excitation or synaptic modulation. To measure such stimulus effects by low‐frequency MF exposure in real‐time, we developed a fluorescent recording system using optical fibers that is neither affected by the MF nor affects the MF distribution. In this study, a numerical calculation model composed of voxels with a 6.25 µm spatial resolution was developed. Using this numerical model, we evaluated the distribution of the EF generated inside three‐dimensional neuronal tissue called neural spheroid, under 50 Hz sinusoidal wave, 300 mT (root mean square) uniform MF exposure. We also investigated the influence of the optical fiber on the electric field distribution in neural spheroid. As a result, MF produced an induced EF in the neural spheroid of more than 4 V/m, well above the theoretical threshold of synaptic modulation. These results indicated that our experimental system was suitable for the evaluation of the threshold of stimulus effects using neural spheroid.

Threshold bounce — occupancy-dependent modulation of the discriminating threshold in silicon detectors

The front-end electronics of silicon detectors are typically designed to ensure optimal noise performance for the expected input charge. A combination of preamplifiers and shaper circuits result in a nontrivial response of the front-end to injected charge, and the magnitude of the response may be sizeable in readout windows subsequent to that in which the charge was initially injected. The modulation of the discriminator threshold due to the superposition of the front-end response across multiple readout windows is coined “threshold bounce”. In this paper, we report a measurement of threshold bounce using silicon modules built for the Phase-II Upgrade of the ATLAS detector at the Large Hadron Collider. These modules utilize ATLAS Binary Chips for their hit readout. The measurement was performed using a micro-focused 15 keV photon beam at the Diamond Light Source synchrotron. The effect of the choice of photon flux and discriminator threshold on the magnitude of the threshold bounce is studied. A Monte Carlo simulation which accounts for the front-end behaviour of the silicon modules is developed, and its predicted hit efficiency is found to be in good agreement with the measured hit efficiency.

Leakage detection of an acoustic emission pipeline based on an improved transformer network

Pipeline leakage detection is an integral part of pipeline integrity management. Combining AE (Acoustic Emission) with deep learning is currently the most commonly used method for pipeline leakage detection. However, this approach is usually applicable only to specific situations and requires powerful signal analysis and computational capabilities. To address these issues, this paper proposes an improved Transformer network model for diagnosing faults associated with abnormal working conditions in acoustic emission pipelines. First, the method utilizes the temporal properties of the GRU and the positional coding of the Transformer to capture and feature extract the data point sequence position information to suppress redundant information, and introduces the largest pooling layer into the Transformer model to alleviate the overfitting phenomenon. Second, while retaining the original attention learning mechanism and identity path in the original DRSN, a new soft threshold function is introduced to replace the ReLU activation function with a new threshold function, and a new soft threshold module and adaptive slope module are designed to construct the improved residual shrinkage unit (ASB-STRSBU), which is used to adaptively set the optimal threshold. Finally, pipeline leakage is classified. The experimental results show that the NDRSN model is able to make full use of global and local information when considering leakage signals and can automatically learn and acquire the important parameters of the input features in the spatial and channel domains. By optimizing the GRU improved Transformer network recognition model, the method significantly reduces the model training time and computational resource consumption while maintaining high leakage recognition accuracy. The average accuracy reached 93.97%. This indicates that the method has good robustness in acoustic emission pipeline leakage detection.

Adaptive thresholding and coordinate attention-based tree-inspired network for aero-engine bearing health monitoring under strong noise

In engineering, due to strong noise interference, it is a challenging issue for decision-making of the neural network to accurately detect healthy conditions of aero-engine bearings. As such, a hierarchical health monitoring model, termed adaptive thresholding and coordinate attention-based tree-inspired network (ATCATN) is developed for the health monitoring of aero-engine bearings under strong background noise. The ATCATN integrates the advantages of the adaptive thresholding and coordinate attention-based network (ATCAN) as well as a tree structure-based method, and it is not merely a straightforward integration of two models. First, the ATCAN is constructed as the backbone network, where the coordinate attention module is developed to process one-dimensional vibration signals; meanwhile, the designed deep residual adaptive thresholding module is introduced to extract significant features from vibration signals with strong noise. Second, a two-layer tree-inspired decision layer, consisting of seed and leaf nodes, is developed to reconstruct the output layer of the ATCAN. Finally, the trained ATCATN is more consistent with maintenance cognition involving multilevel decision information by the progressive determination of fault locations and fault sizes. The experimental results and comparative analysis with seven advanced methods on two aero-engine bearing datasets show that the developed model has the following advantages: (a) model robustness under strong noise; (b) more accurate diagnosis result and (c) hierarchical diagnosis with fault locations and fault sizes. Based on the experimental results, the ATCATN model demonstrates high accuracy in progressively identifying fault locations and sizes of aero-engine bearings even under strong noise interference, which is beneficial for formulating maintenance strategies and aligns with the basic understanding of maintenance.