Establishing a health baseline for reciprocating machinery is one of the effective methods to detect and identify equipment anomalies, which is of great significance for ensuring the safe operation. However, the current methods face challenges due to the incompleteness of mechanistic feature sets and the unclear boundaries in the feature distribution of multi-class samples, making it difficult to construct health baselines using traditional single-model approaches. To address these issues, a method based on hybrid intelligent models for constructing health baselines of reciprocating machinery is proposed. This method utilizes variational time domain decomposition to extract time-domain impulse mechanistic features and combines denoising autoencoder and Kullback-Leibler divergence methods to complete data-driven feature extraction. A discrepancy loss function between deep features and mechanistic features is established, and principal component analysis is employed for feature fusion and dimensionality reduction. Furthermore, a health baseline construction method based on Gaussian mixture model (GMM) and support vector machines (SVMs) is designed. The GMM is used to calculate the cluster centers and covariance matrix eigenvalues of normal samples to estimate the feature distribution boundaries of normal samples. The SVM model is used to calculate the decision boundaries between normal and multiple fault samples, and the intersection of the boundaries from the two models is used to construct the health baseline. The proposed method achieves performance in terms of early warning accuracy, anomaly detection rate and false alarm rate, with results of 95.29%, 91.52%, and 0.13% on 12-cylinder and results of 96.10%, 93.71%, and 1.16% on 6-cylinder diesel engine.

To address the challenge of multi-sensor information fusion in rotating machinery fault monitoring and diagnosis, this paper proposes a novel fault diagnosis method based on multi-sensor feature fusion using a heterogeneous neural network. First, after performing signal denoising preprocessing, both traditional knowledge-based features and deep learning features are integrated through an attention mechanism combined with KL divergence. This integration effectively reduces redundant fault information while enhancing the diversity and discriminative capability of the extracted features. Next, the fused features are fed into a heterogeneous neural network, in which multi-sensor signals are modeled as different types of nodes. A dynamically constructed adjacency matrix is employed to capture the complex correlations among multi-source data, thereby improving the collaborative representation ability between heterogeneous sensor features. Experiments conducted on a public bearing fault dataset and real-world motor data-covering both vibration and electrical signals-demonstrate the superiority of the proposed method. Compared with conventional CNN and CNN-GCN models, the proposed approach achieves higher accuracy, precision, recall and F1-score, validating its effectiveness and robustness in multi-sensor fault diagnosis tasks.

This paper proposes an adaptive robust integral sliding-mode control (ARISMC) strategy centered on dynamic decoupling to address control-performance degradation in electromagnetic actuators for reciprocating compressors caused by coil aging and spring fatigue. First, a nonlinear electromechanically coupled model incorporating electromagnetic, mechanical, and circuit dynamics is developed and validated using test-rig data; the correlation coefficients between the experimental and simulated displacement responses are all above 0.95. A controller is then constructed by integrating backstepping-based dynamic decoupling, nonsingular terminal integral sliding mode, dynamic-surface filtering, and adaptive disturbance compensation. Conditions are established under voltage, current, sampling, and boundary-layer constraints to ensure bounded closed-loop signals and finite-time entry of the displacement- and current-loop sliding variables into a compact neighborhood. Under the 50% constant-load condition, ARISMC achieves an MAE of 0.0013–0.0021 mm under the tested degradation conditions, with a detected timing offset below 0.1 ms; under constant loads of 30%, 70%, and 90%, its maximum RMSE is 0.004600 mm. Under random variable-load operation and eight degradation conditions, its MAE remains within 0.001–0.002 mm, the detected timing offset is below 0.1 ms, and no sustained loss of tracking occurs. Ablation and post-tuning sensitivity results show that dynamic decoupling/compensation is the principal source of performance improvement, the integral sliding mode further reduces residual error, and adaptive compensation provides a modest, condition-dependent gain. The results demonstrate high tracking accuracy and parameter robustness within the degradation and load-switching simulation scenarios considered and provide a basis for subsequent closed-loop hardware validation.

Under combustion impact loads, the small-end bearing in diesel engines operates under severe conditions, making it prone to wear. Most existing studies focus on the dynamics of a single friction pair, neglecting the interactions among multiple friction pairs. A novel rigid-flexible coupled dynamic model of the crank-connecting rod-piston mechanism (CRPM) with multi-clearance lubrication is developed, including the small-end bearing, piston-pin bearings and big-end bearing. Critical components are modeled as modal flexible bodies. The lubrication behavior of bearings is calculated using a transient mixed elasto-hydrodynamic lubrication model. The accuracy of the simulation model is validated through a small-end bearing wear experiment. Furthermore, the wear mechanisms of the small-end bearing are then analyzed under different wear profiles and depths. Experiments and simulations demonstrate that small-end bearing wear induces a new impact in the exhaust stroke. The lubrication performance of the small-end bearing exhibits a nonlinear dependence on the wear profile and depth, deteriorating sharply when the ratio of wear depth to wear profile is large. These findings offer a new theoretical basis for diagnosing small-end bearing wear in diesel engines.

Remaining useful life (RUL) prediction of wind-turbine bearings is challenged by nonstationary wind loads, multistage degradation, substantial lifetime dispersion, and strict deployment constraints. Conventional single-task regressors apply a unified feature-to-RUL mapping over the entire life cycle and therefore struggle to characterize stage transitions and bearing-specific degradation progress. To address these issues, this paper proposes a stage-aware multi-task RUL prediction method with causal degradation-prior fusion and collaborative distillation. During training, a high-capacity reference representation path transfers inter-sample relational structures and task-level degradation knowledge to a compact feature encoding path, while only the compact path is retained for inference. Based on the compact representation, a multi-task module jointly performs four-stage classification, stage-conditioned RUL regression, and continuous remaining-life estimation; predicted stage probabilities softly fuse the stage-conditioned outputs. A causal prior-fusion module further integrates a bearing-specific healthy-state anchor, causally identified first prediction time, cumulative damage, and the stage-aware prediction to adapt the RUL trajectory to individual degradation processes. Experiments on the IEEE PHM 2012 and XJTU-SY datasets demonstrate that the proposed method provides accurate and robust RUL prediction across different bearing degradation processes. Moreover, the compact inference path maintains efficient implementation, supporting its potential use in practical wind-turbine condition-monitoring applications.

Ensuring the safe and stable operation of industrial systems depends crucially on fault diagnosis models with strong generalization capabilities. However, there are significant differences in the distribution of data under different working conditions or equipment, which poses a fundamental challenge. Factors such as fluctuations in operating conditions result in models being ineffective for cross domain fault diagnosis, thereby limiting the practical value of intelligent diagnostic methods in engineering equipment. In addition, effectively integrating complementary but redundant multi-channel data to extract discriminative features remains a key technical obstacle. To solve these issues, this paper proposes a Boundary-Aware Hierarchical Subdomain Adaptation Network (BA-HSAN) for cross domain industrial fault diagnosis. Firstly, multi-channel vibration signals converted into greyscale images and use spatiotemporal attention mechanism next to deep convolutional layers to mine high-dimensional features, achieving robust multi-source data fusion. Secondly, a dual-classifier adversarial adaptation module is introduced to detect and reduce decision-boundary inconsistency between the source and target domains. Finally, LMMD-based subdomain alignment is applied to the outputs of the last two deep feature layers, enabling progressively improved class-level alignment of semantically similar samples across domains. Comprehensive experiments on multiple cross domain fault diagnosis tasks have shown that the proposed method consistently outperforms state-of-the-art methods in terms of diagnostic accuracy and robustness. This work provides a scalable framework for intelligent fault diagnosis in industrial big data environments.

With the application of stepless capacity control systems, an increasing number of reciprocating compressor suction valves have shifted from automatic to controlled operation. Inconsistencies in the modification of suction valves, along with various fault states during operation, have reduced the adaptability of traditional integrated valve operation models, making it difficult to meet the analysis requirements for the transient thermodynamic processes of compressors. This study introduces a new independent operation model for the suction valves of a reciprocating compressor. The flow channel structure, flow performance parameters of different suction valves, and the structure and motion parameters of the unloaders were designed as independent variables to perform numerical simulations of compressor operation under complex working conditions. To address the nonlinear relationship between the mechanical structure of the suction valve and the forward and reverse flow performance of the capacity control, a three-dimensional simulation model was constructed to calculate the flow coefficient. Using experimental data, the accuracy of the theoretical model was validated, and the flow coefficients under various states were integrated into the new independent operation model for the suction valves. A detailed thermodynamic performance analysis of the compressor was conducted under abnormal conditions, such as inconsistent suction valve lift, unloader fork wear, and unloader retraction delay. Variations in thermodynamic parameters, including dynamic pressure, exhaust volume, and indicated work, were analysed, providing technical guidance for suction valve design, fault monitoring, and the optimisation control of reciprocating compressors.

To solve the problem of extracting the impact component from the complex time-domain vibration signal of reciprocating machinery vibration signals, a differential evolution (DE)-based time domain decomposition method is proposed to achieve adaptive extraction of impact components. The method establishes new decomposition window containing three adjustment parameters to adapt to multiple forms of impact components. Furthermore, with the optimization objectives of minimizing reconstruction loss, amplitude moment loss, and similarity loss, a decomposition parameter optimization algorithm based on DE is established to achieve the optimization process of decomposition parameters. The results of processing simulated and actual vibration signals of diesel engines show that the new method can adaptively and accurately identify the impact component and impact time center in the vibration component, with a signal reconstruction loss of less than 2.5% and a decomposition time of only 54.1 s.

Vibration sensor network optimization increases monitoring effectiveness and reduces sensor quantity and transmission burden. However, the traditional model-driven methods depend on precise finite element models, challenging for complex machinery. This paper presents a data-driven approach using the Sparse Regularized Graph Pooling Network (SRGPN), which conceptualizes sensor networks as graphs and uses graph pooling to identify optimal sensor combinations. A sparsity regularization term related to the number of sensors is included in the loss function, aiming for the minimal yet effective sensor combination. Additionally, a monitoring capability metric suited for diesel engines with multi-source impulse signals is proposed, reflecting the sensors’ monitoring capacity. Validated through simulations and tests on a diesel engine test bench, the results show that SRGPN optimally places sensors near excitation sources, balancing sensor count and monitoring needs. This approach shows potential for optimizing sensor placements in condition monitoring.

Performance degradation assessment (PDA) is a critical component of predictive health management (PHM). The mixed multi-source impulse characteristics of diesel engine vibration signals make PDA more challenging compared to rotating machinery. To address the unique characteristics of diesel engine signals, this study proposes a Subspace-Whitening Support Vector Data Description (S-WhiteSVDD) feature fusion approach that combines knowledge-based features with deep learning features. The method tracks cross-cycle variations of multiple homologous impulses and constructs a feature subspace for each impulse. Whitening transformation ensures balanced stretching and compression of subspace data across all components, preventing features with large variances from dominating the decision boundary. This approach aligns more closely with the data manifold and enables precise control of anomaly boundaries. To overcome the challenge of significant health indicator (HI) fluctuations that hinder early fault detection, the method integrates the interpretability of knowledge-based features with the complex mapping capabilities of deep features. This fusion enhances the richness of feature representation, facilitating the detection of early fault onset. The effectiveness and superiority of the proposed method are demonstrated through both valve degradation simulations and nozzle degradation engineering case studies. The constructed HI effectively indicates component degradation. The proposed approach shows strong potential for practical engineering applications.

Reciprocating compressor is the key equipment in refrigeration system. The electromagnetic capacity control system plays an important role in energy conservation of the reciprocating compressor. The objective of this study is to design a control strategy that addresses several challenges, including large seating impacts, mismatch between load variations and actuator response capabilities, and insufficient robustness within the electromagnetic capacity control system. Firstly, a nonlinear model of the reciprocating compressor with stepless capacity control system is established. Secondly, an Integral Model Predictive Control-Adaptive Backstepping Integral Sliding Mode Control (IMPC-ABISMC) cascade controller is designed via three steps. Finally, the controller is comparatively tested. The results demonstrate that the control strategy can ensure that the reciprocating compressor buffer tank pressure is timely and accurately tracked the set value, reducing the steady-state error to a range of 0-0.13% and the expected load generated exhibits greater smoothness than that of Adaptive Fuzzy Sliding Mode Control (AFSMC). Additionally, the control strategy can control the actuator to track the target trajectory under different working loads with tracking accuracies of less than 0.09mm and maintains a low average seating velocity of 0.016m s-1, demonstrating strong robustness against the pressure disturbance and the load force disturbance. The innovation of this study lies in the design of the IMPC and ABISMC controllers, as well as the cascading of two controllers via intermediate transformation, forming a cascade control strategy that ensures precise load response while accommodating dynamic performance of the actuator. This control strategy is applicable to compressor experiencing frequent load variation, facilitating stepless capacity control and effectively addressing challenges such as excessive energy consumption.

Reciprocating machinery has compact and complex structures, many moving parts, and numerous vibration excitation sources. Impact signals caused by mechanical part faults can easily produce time-frequency coupling with multi-source impact signals from components normal movements. At the same time, variable operating conditions, such as different speed and load will lead to nonlinear changes in the time-frequency characteristics of all collected vibration signals. These problems make it difficult to extract fault features. In this study, a multi-impact time-domain adaptive decomposition method for multigroup signal under variable operating conditions was proposed to separate fault features from multi-source impact signals. Firstly, in order to reduce the information loss of decomposition and improve impact extraction integrity, a decomposition target was established to minimise the loss of a reconstructed source signal, the sum of the inner product of a sub-signal, and the amplitude moment of a sub-signal relative to an impact centre. A new bilateral adaptive decomposition window with three control parameters including window center, bilateral shape and peak is designed to adapt to the characteristics of an impact shape in a time-varying state. Aiming at solving the problems of noise interference and initial model parameters setting, residual energy spectrum is applied to adaptively estimate a noise spectrum distribution and a multi-impact time domain centre. Furthermore, with the aim of large amounts of sensors signal synchronous decomposition at variable operating conditions, a multigroup signal adaptive decomposition parameter optimisation scheme integrating a fully connected network and an ADAM algorithm is designed to considerably improve computational efficiency. Numerical simulations and engine test data are studied to show that the proposed method shortens calculation time on the basis of realising multi-impact adaptive decomposition. The average processing time of a single group signal is 8.6 s, and the average processing time of four groups of signal synchronization is 20.0 s, which is significantly faster than those of the existing time domain impact decomposition method.

In marine and off-road machinery, diesel engines operate under complex dynamic conditions. The superposition of high-pressure combustion shocks and variable conditions severely deteriorates the lubrication performance of connecting rod bearings, leading to localized wear and potential failures. To investigate the tribo-dynamic behavior under complex conditions, a rigid-flexible coupling dynamic model of the crank-connecting rod mechanism is established. This model considers the effects of combustion pressure, bearing deformation, and surface roughness. The crankshaft and big-end bearing are modeled as modal flexible bodies, and key lubrication characteristics are evaluated using a mixed elastohydrodynamic lubrication (MEHD) model. This study investigates the effects of rotational speed, external load, and wear clearance on the lubrication performance of the big-end bearing, focusing on peak oil film pressure, peak asperity contact pressure, and minimum oil film thickness. Simulation results reveal distinct influences of rotational speed, load and wear clearance on bearing behavior. Further, the simulation outcomes are validated through wear experiments conducted on the big-end bearing of a diesel engine. These findings offer theoretical insight and practical guidance for optimizing bearing design, improving lubrication performance, and extending service life.

A stepless capacity-control system drives a compressor suction valve to operate independently at high frequencies and speeds. Failures, such as unloader wear, valve plate leakage, and control system performance degradation, significantly alter the thermodynamic processes of the compressor. This study established a new compressor model accounting for independent collaborative operation of multiple valves coupled with diverse fault modes. The model comprehensively considers unloader driving force, displacement, and action time parameters, incorporating independently parameterized valve dynamic motion equations and valve clearance-cylinder fluid flow equations. Simultaneously, it integrates fault influence parameters (leakage rate, unloader wear) with solving equations for multi-valve partial faults. The model was validated through a constructed compressor simulation monitoring test rig, demonstrating less than 5% error under normal operating conditions. Furthermore, the research revealed thermodynamic performance distortion characteristics in complex operating conditions and identified characteristic parameters of sudden multi-valve local faults under variable loads. The extracted fault patterns include valve unloader wear, valve leakage, and coupled faults. Analytical results demonstrate that the comprehensive impact intensity of concurrent multi-valve faults exhibits significant nonlinear superposition effects. Coupled fault modes display mutually inverse characteristics in fault parameters such as indicated work and discharge capacity, while specific fault categories can be differentiated through inward contraction or expansion trends in dynamic pressure profiles.

Existing dynamic models of the crank-connecting rod mechanism (CRM) primarily focus on single clearance lubrication in piston engines, making it difficult to analyze the coupled effects of crankshaft misalignment and mixed lubrication of bearings under multi-clearance collaboration. This study proposes a novel dynamic model that integrates multi-clearance lubrication. Based on the generalized coordinates of the crank and connecting rod, including misalignment angles in two directions, the dynamic boundary conditions of the three-dimensional lubrication fields of the big-end and main bearings are calculated synchronously, and incorporated into a mixed elasto-hydrodynamic lubrication (MEHD) model to evaluate the friction performance. The motion equations are derived using the Lagrange method with a variable-step fourth-order Runge-Kutta (VRK4) method to address numerical instability under multi-clearance misalignment. Furthermore, the coupled effects of operating parameters and misalignment on the friction and wear characteristics of the bearings are examined. The results show that crankshaft misalignment significantly increases the volumetric wear rate and friction power loss, with more pronounced effects under low-speed, high-load conditions. Appropriate bearing clearance and lubrication viscosity can help mitigate the adverse effects of misalignment. This study provides a high-precision simulation framework for analyzing and designing piston engine bearings and elucidates the misalignment fault mechanisms.

Reciprocating mechanical vibration signals are characterized by multi-impact source time--frequency coupling and non-stationary properties. These characteristics present significant challenges for signal processing and the extraction of weak fault features. Current methods for decomposing reciprocating mechanical signals often overlook the variations in impact energy and struggle to eliminate noise within the impact zones, thereby leaving room for improvement in impact feature description and boundary localization. In this study, impact time-frequency decomposition (ITFD), an adaptive method for decomposing time-frequency coupled signals from multiple impact sources is introduced. Initially, the time-frequency energy gradient (TFEG) is defined based on the signal's time-frequency properties, accurately depicting the variations in impact energy. Subsequently, a strategy for impact-adaptive localization is designed based on the TFEG characteristics, which efficiently and precisely captures the time-frequency filtering boundaries for sub-impacts by utilizing the properties of the noise's time-frequency distribution. Lastly, a weighted impact recognition threshold is proposed, adaptively set through a time--frequency noise estimation method, ensuring robust noise-resistant identification of impact components. Simulations and practical reciprocating machinery fault simulation signal tests demonstrate that ITFD significantly outperforms other methods in terms of decomposition accuracy and efficiency. It can precisely and efficiently decouple time-frequency coupled signals from multiple impact sources, reducing the signal processing time for a single working cycle to 3.52 s, thus showing substantial potential for engineering applications.

Angular misalignment in diesel engine shafting poses significant challenges to performance and reliability. This study develops a rigid-flexible coupled multibody dynamics model to investigate the vibration characteristics of diesel engine shafting under misalignment. Unlike conventional rotor models, the proposed model includes highelastic coupling effects, periodic gas forces, and reciprocating inertial forces, providing a more accurate representation of system behavior. Analysis reveals that misalignment causes changes in fundamental, second, and third harmonic frequencies, and distinct shaft center trajectory variations. An efficient online detection method is proposed, using minimal sensor instrumentation to identify and assess misalignment severity. Experimental validation on a six-degree-of-freedom test bench shows strong agreement with simulation results, confirming the robustness of the model and detection method. This study offers an effective approach for early misalignment detection in diesel engines, providing valuable insights for industrial health monitoring.

Adaptive decomposition methods for vibration signals have predominantly concentrated on the extraction of fault frequency components in rotational machinery, often overlooking the time-domain impact characteristics of reciprocating machinery. These methods struggle to effectively isolate the periodic, multi-source impact features of reciprocating machinery faults, such as those induced by valve opening or closing, and connecting rod bearing wear. To address the non-stationary, strongly coupled, and interference-prone time-frequency characteristics of casing vibration signals in reciprocating machinery, this study proposes a novel adaptive decomposition method tailored for multi-impact signal analysis. The approach leverages the time-frequency reassigned multisynchrosqueezing transform to enhance the representation of impacts in the time-frequency domain. An optimization criterion based on the squared L2-norm of time-frequency spectral moments is formulated to accurately locate the energy centroids of decomposed components. A recursive decomposition framework is introduced, eliminating the need to predefine decomposition levels while simultaneously improving computational efficiency. To further refine decomposition accuracy, a weight factor optimization strategy employing a q-order time-frequency distance metric is developed to adaptively shape decomposition windows according to impact characteristics. Validation on both simulated and experimental datasets involving diesel engine valve clearance faults demonstrates that the proposed method significantly improves the precision of adaptive impact component extraction and offers enhanced computational performance.

Accurately locating the fault impacts and extracting sensitive fault features of vibration signals are challenging problems in diesel engine fault diagnosis. To address the limited integration of existing attention mechanisms with the knowledge of diesel engine operating principles, black -box feature extraction and insufficient interpretability problems, a novel method called priori -distribution adaptive sparse attention (PASA) is devised. This method translates the established priori -distribution of impact features guided by mechanistic knowledge into target formulas learnable, driving the model to learn attention results aligned with priori -distribution. Based on the attention results from PASA, a cross -domain feature mining (CDFM) method is proposed. Leveraging traditional thermodynamic and dynamic features associated with diesel engine operations, it accomplishes crossdomain feature extraction in frequency, time, and envelope domains, constructing a fault -sensitive feature set. Furthermore, the model structure for feature extraction is optimized, reducing model parameter complexity, and leading to the establishment of a diagnostic model. Fault experiments are conducted on two diesel engines to verify the proposed models, including misfires, valve malfunctions, collisions, and bush faults. The results demonstrate that compared to existing methods for fault diagnosis in diesel engines, the proposed approach accurately identifies vibration signal fault characteristics conforming to prior distributions. It shows well performance in four diagnostic indicators of diagnostic accuracy, precision, recall, and F1 score.

Designing an efficient and simplified electromagnetic capacity control system (ECCS) for reciprocating compressors, replacing hydraulic systems, is a crucial focus for energy conservation and efficiency enhancement in industrial production. Major challenges lie in the compressor’s high gas force and rotational speed, coupled with the actuator coil’s self-inductance, leading to complex response characteristics and capacity control failures under traditional on-off drive modes. To address these issues, this paper introduces a coupling control model for capacity control under synchronous frequency with asynchronous actuation of the actuator and suction valve, considering periodic excitation, reverse attraction, and partial stroke contact. By simulating actuator response and compressor performance under various voltage drives, as well as conducting a sensitivity analysis of various control parameters, we develop a load regulation procedure using Fixed Duration Timing Shift and Variable Duration (FDTSVD). This procedure achieves precise 0%-100% load regulation, with discharge volume relative error <7.5%. When the exhaust load decreases from 100 % to 25 %, the one-cycle indicated work decreases from 593.6 J to 175.2 J, demonstrating its effectiveness.