To accurately diagnosemisfire faults in automotive engines,we propose a Channel Attention Convolutional Model,specifically the Squeeze-and-Excitation Networks(SENET),for classifying engine vibration signals and precis...To accurately diagnosemisfire faults in automotive engines,we propose a Channel Attention Convolutional Model,specifically the Squeeze-and-Excitation Networks(SENET),for classifying engine vibration signals and precisely pinpointing misfire faults.In the experiment,we established a total of 11 distinct states,encompassing the engine’s normal state,single-cylinder misfire faults,and dual-cylinder misfire faults for different cylinders.Data collection was facilitated by a highly sensitive acceleration signal collector with a high sampling rate of 20,840Hz.The collected data were methodically divided into training and testing sets based on different experimental groups to ensure generalization and prevent overlap between the two sets.The results revealed that,with a vibration acceleration sequence of 1000 time steps(approximately 50 ms)as input,the SENET model achieved a misfire fault detection accuracy of 99.8%.For comparison,we also trained and tested several commonly used models,including Long Short-Term Memory(LSTM),Transformer,and Multi-Scale Residual Networks(MSRESNET),yielding accuracy rates of 84%,79%,and 95%,respectively.This underscores the superior accuracy of the SENET model in detecting engine misfire faults compared to other models.Furthermore,the F1 scores for each type of recognition in the SENET model surpassed 0.98,outperforming the baseline models.Our analysis indicated that the misclassified samples in the LSTM and Transformer models’predictions were primarily due to intra-class misidentifications between single-cylinder and dual-cylinder misfire scenarios.To delve deeper,we conducted a visual analysis of the features extracted by the LSTM and SENET models using T-distributed Stochastic Neighbor Embedding(T-SNE)technology.The findings revealed that,in the LSTMmodel,data points of the same type tended to cluster together with significant overlap.Conversely,in the SENET model,data points of various types were more widely and evenly dispersed,demonstrating its effectiveness in distinguishing between different fault types.展开更多
Misfire is a common fault in compression ignition engines,characterized by the absence or flame loss due to insufficient fuel in the cylinders.This fault is difficult to diagnose and resolve due to its multiple potent...Misfire is a common fault in compression ignition engines,characterized by the absence or flame loss due to insufficient fuel in the cylinders.This fault is difficult to diagnose and resolve due to its multiple potential causes.This study focuses on identifying misfires in a 12-cylinder V-type marine diesel engine by analyzing vibration data collected from 15 accelerometers mounted on the engine block.Three machine learning algorithms—K-Nearest Neighbors(K-NNs),support vector machines(SVMs),and random forests(RFs)—were employed to classify engine conditions using 18 time-domain features.Results showed that the K-NN,SVM and RF algorithms achieved F1 scores of 99.87%,100%,and 99.87%,respectively,when using 18 time-domain features and all 15 accelerometers mounted on the engine block.Additionally,the study evaluated classification performance while reducing the number of accelerometers and features using two methods:Relief-F and general combinatory analysis(GCA).Although the GCA method yields better results when using only two accelerometers and nine features for misfire classification,its overall process required substantially more computational time compared to Relief-F.The best result obtained with Relief-F was achieved using 3 accelerometers and 18 features.Therefore,Relief-F proved to be more practical and take less overall computational time within the proposed framework.展开更多
Due to non-stationary characteristics of the vibration signal acquired from cylinder head,a misfire fault diagnosis system of automobile engines based on correlation coefficient gained by wavelet packet and extreme le...Due to non-stationary characteristics of the vibration signal acquired from cylinder head,a misfire fault diagnosis system of automobile engines based on correlation coefficient gained by wavelet packet and extreme learning machine(ELM)is proposed.Firstly,the original signal is decomposed by wavelet packet,and correlation coefficients between the reconstructed signal of each sub-band and the original signal as well as the energy entropy of each sample are obtained.Then,the eigenvectors established by the correlation coefficients method and the energy entropy method fused with kurtosis are inputted to the four kinds of classifiers including BP neural network,KNN classifier,support vector machine and ELM respectively for training and testing.Experimental results show that the method proposed in this paper can effectively reflect the differences that the fault produces and identify the single-cylinder misfire accurately,which has the advantages of higher accuracy and shorter training time.展开更多
In this paper,transient electromagnetic method was used to carry out the feasibility study on the detection and recognition of chamber blasting misfire.Firstly,an electromagnetic background field was established in th...In this paper,transient electromagnetic method was used to carry out the feasibility study on the detection and recognition of chamber blasting misfire.Firstly,an electromagnetic background field was established in the test;secondly,a benign conductor was preset in the chamber,and then the background field was eliminated after the electromagnetic field was measured;thirdly,the transient electromagnetic field was measured again after blasting;at last,the chamber blasting misfire was detected and recognized by comparing the change of eddy current field of the preset benign conductor before and after blasting.The test results showed that:When the buried depth of aluminum box target was no more than 30 m,transient electromagnetic method can clearly identify the position of the aluminum box;when the buried depth of aluminum box was more than30 m,the buried depth and position of the aluminum box was not sure due to the unknown level of secondary eddy current field generated by aluminum box.展开更多
Diesel engines are critical power sources widely used in marine,transportation,and industrial applications,where reliable operation is essential for safety and economic efficiency.However,traditional signal processing...Diesel engines are critical power sources widely used in marine,transportation,and industrial applications,where reliable operation is essential for safety and economic efficiency.However,traditional signal processing and many machine learning methods face challenges in extracting generalized fault features and accurately diagnosing misfires under complex,noisy operating conditions.To address these challenges,this paper proposes a novel multi-scale bottleneck attention and mixupbased domain adaptation network(MBA-MDAN)for reliable misfire detection across varying noise levels and working conditions.The approach integrates a denoising convolutional neural network(DnCNN)to suppress noise unrelated to fault characteristics,enabling clearer fault signal extraction.A parallel multi-scale convolution module captures fault features at different time scales,while a bottleneck attention module(BAM)selectively emphasizes critical features for deep fault representation.To preserve important information,time-domain statistical features are also incorporated.During training,metric learning minimizes feature discrepancies between source and target domains,and adversarial training between a domain discriminator and the fault classifier enhances domain adaptation.Additionally,domain mixup is applied to augment discriminator samples,further improving diagnostic performance.On the real-world datasets,compared to several state-of-the-art methods—including ShuffleNetV2,DenseNet,ANMCNN,MCBACNN,DANN,and WDAN—MBA-MDAN improves average diagnostic accuracy by 30.372%,16.407%,15.410%,16.483%,5.200%,and 4.725%,respectively.These results confirm the effectiveness of MBA-MDAN and indicate its strong potential for integration into intelligent operation and maintenance(O&M)systems for diesel engines,facilitating early fault detection and maintenance decision-making in complex industrial environments.展开更多
The misfiring fault of internal combustion engine was detected by using the vibration signals of cylinder-head.Based on the data acquisition system built with LabVIEW,the cylinder-head vibration signals were detected ...The misfiring fault of internal combustion engine was detected by using the vibration signals of cylinder-head.Based on the data acquisition system built with LabVIEW,the cylinder-head vibration signals were detected with an accelerometer while the engine was rapidly accelerating from idle speed to high speed,at which time the engine was running under four working conditions of normal and single cylinder misfiring,double cylinders continuously misfiring and double cylinders alternately misfiring.After decomposing the vibration signals with db3 wavelet,whether the engine was misfiring or not,and what type of misfiring,were judged by comparing the decomposing results.The resultd showrf that,the low-frequency vibration of the engine cylinder head was related to the rotation of the principal shaft,and the high-frequency vibration was related to the combustion in the cylinder.There were certain corresponding relationships between wave crests of high-frequency vibration and wave crests of low-frequency under the four conditions of normal and faults when engine runs in idle segment,accelerating segment,and high-speed segment.Thus,the misfiring fault and type can be detected by analyzing the corresponding relations.Detection of the misfiring fault by using wavelet analysis was effective and feasible.展开更多
Vector of diagnostic signs(VDS)using torsional vibration(TV)signal on the main propulsion plant(MPP)is the vector of z maxima(or minima)values of the TV signal in accordance with the cylinder firing orders.The technic...Vector of diagnostic signs(VDS)using torsional vibration(TV)signal on the main propulsion plant(MPP)is the vector of z maxima(or minima)values of the TV signal in accordance with the cylinder firing orders.The technical states of the marine diesel engine(MDE)include R=z+1 classes and are presented in z-dimensional space coordinate of VDS.The presentation of Dk,k=1÷R using z diagnostic signs(Vi,i=1÷z)is nonfigurative and quite complicated.This paper aims to develop a new method for converting VDS from z-dimensional to 2-dimensional space(two-axes)based on the firing orders of the diesel cylinders,as an equivalent geometrical sign of the all diagnostic signs.The proposed model is useful for presenting a technical state Dk in two-dimensional space(x,y)for better visualization.The paper verifies the simulation of the classification illustration of the 7–state classes for the MDE 6S46-MCC,installed on the motor vessel(MV)34000DWT,using the new above mentioned method.The seven technical state classes(for 6-cylinder MDE,z=6)are drawn separately and visually in the Descartes.The received results are valuable to improve smart diagnostic system for analyzing normal/misfire states of cylinders in operation regimes.展开更多
Free piston linear generator(FPLG)is a promising range extender for the electrical vehicle with unparallel advantages,such as compact structure,higher system efficiency,and reduced maintenance cost.However,due to the ...Free piston linear generator(FPLG)is a promising range extender for the electrical vehicle with unparallel advantages,such as compact structure,higher system efficiency,and reduced maintenance cost.However,due to the lack of the mechanic crankshaft,the related piston motion control is a challenge for the FPLG which causes problems such as misfire and crash and limits its widespread commercialization.Aimed at resolving the problems as misfire,a single-piston FPLG prototype has been designed and manufactured at Shanghai Jiao Tong University(SJTU).In this paper,the development process and experimental validation of the related control strategies were detailed.From the experimental studies,significant misfires were observed at first,while the FPLG operated in natural-aspiration conditions.The root cause of this misfire was then identified as the poor scavenging process,and a compressed air source was leveraged to enhance the related scavenging pressure.Afterward,optimal control parameters,in terms of scavenging pressure,air-fuel equivalence ratio,and ignition position,were then calibrated in this charged-scavenging condition.Eventually,the FPLG prototype has achieved a continuous stable operation of over 1000 cycles with an ignition rate of 100%and a cycle-to-cycle variation of less than 0.8%,produced an indicated power of 2.8 kW with an indicated thermal efficiency of 26%and an electrical power of 2.5 kW with an overall efficiency of 23.2%.展开更多
基金Yongxian Huang supported by Projects of Guangzhou Science and Technology Plan(2023A04J0409)。
文摘To accurately diagnosemisfire faults in automotive engines,we propose a Channel Attention Convolutional Model,specifically the Squeeze-and-Excitation Networks(SENET),for classifying engine vibration signals and precisely pinpointing misfire faults.In the experiment,we established a total of 11 distinct states,encompassing the engine’s normal state,single-cylinder misfire faults,and dual-cylinder misfire faults for different cylinders.Data collection was facilitated by a highly sensitive acceleration signal collector with a high sampling rate of 20,840Hz.The collected data were methodically divided into training and testing sets based on different experimental groups to ensure generalization and prevent overlap between the two sets.The results revealed that,with a vibration acceleration sequence of 1000 time steps(approximately 50 ms)as input,the SENET model achieved a misfire fault detection accuracy of 99.8%.For comparison,we also trained and tested several commonly used models,including Long Short-Term Memory(LSTM),Transformer,and Multi-Scale Residual Networks(MSRESNET),yielding accuracy rates of 84%,79%,and 95%,respectively.This underscores the superior accuracy of the SENET model in detecting engine misfire faults compared to other models.Furthermore,the F1 scores for each type of recognition in the SENET model surpassed 0.98,outperforming the baseline models.Our analysis indicated that the misclassified samples in the LSTM and Transformer models’predictions were primarily due to intra-class misidentifications between single-cylinder and dual-cylinder misfire scenarios.To delve deeper,we conducted a visual analysis of the features extracted by the LSTM and SENET models using T-distributed Stochastic Neighbor Embedding(T-SNE)technology.The findings revealed that,in the LSTMmodel,data points of the same type tended to cluster together with significant overlap.Conversely,in the SENET model,data points of various types were more widely and evenly dispersed,demonstrating its effectiveness in distinguishing between different fault types.
文摘Misfire is a common fault in compression ignition engines,characterized by the absence or flame loss due to insufficient fuel in the cylinders.This fault is difficult to diagnose and resolve due to its multiple potential causes.This study focuses on identifying misfires in a 12-cylinder V-type marine diesel engine by analyzing vibration data collected from 15 accelerometers mounted on the engine block.Three machine learning algorithms—K-Nearest Neighbors(K-NNs),support vector machines(SVMs),and random forests(RFs)—were employed to classify engine conditions using 18 time-domain features.Results showed that the K-NN,SVM and RF algorithms achieved F1 scores of 99.87%,100%,and 99.87%,respectively,when using 18 time-domain features and all 15 accelerometers mounted on the engine block.Additionally,the study evaluated classification performance while reducing the number of accelerometers and features using two methods:Relief-F and general combinatory analysis(GCA).Although the GCA method yields better results when using only two accelerometers and nine features for misfire classification,its overall process required substantially more computational time compared to Relief-F.The best result obtained with Relief-F was achieved using 3 accelerometers and 18 features.Therefore,Relief-F proved to be more practical and take less overall computational time within the proposed framework.
基金Tianjin Regional Demonstration Project of Marine Economy Innovation and Development(No.2015120024000473)
文摘Due to non-stationary characteristics of the vibration signal acquired from cylinder head,a misfire fault diagnosis system of automobile engines based on correlation coefficient gained by wavelet packet and extreme learning machine(ELM)is proposed.Firstly,the original signal is decomposed by wavelet packet,and correlation coefficients between the reconstructed signal of each sub-band and the original signal as well as the energy entropy of each sample are obtained.Then,the eigenvectors established by the correlation coefficients method and the energy entropy method fused with kurtosis are inputted to the four kinds of classifiers including BP neural network,KNN classifier,support vector machine and ELM respectively for training and testing.Experimental results show that the method proposed in this paper can effectively reflect the differences that the fault produces and identify the single-cylinder misfire accurately,which has the advantages of higher accuracy and shorter training time.
文摘In this paper,transient electromagnetic method was used to carry out the feasibility study on the detection and recognition of chamber blasting misfire.Firstly,an electromagnetic background field was established in the test;secondly,a benign conductor was preset in the chamber,and then the background field was eliminated after the electromagnetic field was measured;thirdly,the transient electromagnetic field was measured again after blasting;at last,the chamber blasting misfire was detected and recognized by comparing the change of eddy current field of the preset benign conductor before and after blasting.The test results showed that:When the buried depth of aluminum box target was no more than 30 m,transient electromagnetic method can clearly identify the position of the aluminum box;when the buried depth of aluminum box was more than30 m,the buried depth and position of the aluminum box was not sure due to the unknown level of secondary eddy current field generated by aluminum box.
基金supported by the National Natural Science Foundation of China(Grant No.52375255)。
文摘Diesel engines are critical power sources widely used in marine,transportation,and industrial applications,where reliable operation is essential for safety and economic efficiency.However,traditional signal processing and many machine learning methods face challenges in extracting generalized fault features and accurately diagnosing misfires under complex,noisy operating conditions.To address these challenges,this paper proposes a novel multi-scale bottleneck attention and mixupbased domain adaptation network(MBA-MDAN)for reliable misfire detection across varying noise levels and working conditions.The approach integrates a denoising convolutional neural network(DnCNN)to suppress noise unrelated to fault characteristics,enabling clearer fault signal extraction.A parallel multi-scale convolution module captures fault features at different time scales,while a bottleneck attention module(BAM)selectively emphasizes critical features for deep fault representation.To preserve important information,time-domain statistical features are also incorporated.During training,metric learning minimizes feature discrepancies between source and target domains,and adversarial training between a domain discriminator and the fault classifier enhances domain adaptation.Additionally,domain mixup is applied to augment discriminator samples,further improving diagnostic performance.On the real-world datasets,compared to several state-of-the-art methods—including ShuffleNetV2,DenseNet,ANMCNN,MCBACNN,DANN,and WDAN—MBA-MDAN improves average diagnostic accuracy by 30.372%,16.407%,15.410%,16.483%,5.200%,and 4.725%,respectively.These results confirm the effectiveness of MBA-MDAN and indicate its strong potential for integration into intelligent operation and maintenance(O&M)systems for diesel engines,facilitating early fault detection and maintenance decision-making in complex industrial environments.
文摘The misfiring fault of internal combustion engine was detected by using the vibration signals of cylinder-head.Based on the data acquisition system built with LabVIEW,the cylinder-head vibration signals were detected with an accelerometer while the engine was rapidly accelerating from idle speed to high speed,at which time the engine was running under four working conditions of normal and single cylinder misfiring,double cylinders continuously misfiring and double cylinders alternately misfiring.After decomposing the vibration signals with db3 wavelet,whether the engine was misfiring or not,and what type of misfiring,were judged by comparing the decomposing results.The resultd showrf that,the low-frequency vibration of the engine cylinder head was related to the rotation of the principal shaft,and the high-frequency vibration was related to the combustion in the cylinder.There were certain corresponding relationships between wave crests of high-frequency vibration and wave crests of low-frequency under the four conditions of normal and faults when engine runs in idle segment,accelerating segment,and high-speed segment.Thus,the misfiring fault and type can be detected by analyzing the corresponding relations.Detection of the misfiring fault by using wavelet analysis was effective and feasible.
文摘Vector of diagnostic signs(VDS)using torsional vibration(TV)signal on the main propulsion plant(MPP)is the vector of z maxima(or minima)values of the TV signal in accordance with the cylinder firing orders.The technical states of the marine diesel engine(MDE)include R=z+1 classes and are presented in z-dimensional space coordinate of VDS.The presentation of Dk,k=1÷R using z diagnostic signs(Vi,i=1÷z)is nonfigurative and quite complicated.This paper aims to develop a new method for converting VDS from z-dimensional to 2-dimensional space(two-axes)based on the firing orders of the diesel cylinders,as an equivalent geometrical sign of the all diagnostic signs.The proposed model is useful for presenting a technical state Dk in two-dimensional space(x,y)for better visualization.The paper verifies the simulation of the classification illustration of the 7–state classes for the MDE 6S46-MCC,installed on the motor vessel(MV)34000DWT,using the new above mentioned method.The seven technical state classes(for 6-cylinder MDE,z=6)are drawn separately and visually in the Descartes.The received results are valuable to improve smart diagnostic system for analyzing normal/misfire states of cylinders in operation regimes.
基金supported by the project of Shanghai Science and Technology Commission(No.19511108500).
文摘Free piston linear generator(FPLG)is a promising range extender for the electrical vehicle with unparallel advantages,such as compact structure,higher system efficiency,and reduced maintenance cost.However,due to the lack of the mechanic crankshaft,the related piston motion control is a challenge for the FPLG which causes problems such as misfire and crash and limits its widespread commercialization.Aimed at resolving the problems as misfire,a single-piston FPLG prototype has been designed and manufactured at Shanghai Jiao Tong University(SJTU).In this paper,the development process and experimental validation of the related control strategies were detailed.From the experimental studies,significant misfires were observed at first,while the FPLG operated in natural-aspiration conditions.The root cause of this misfire was then identified as the poor scavenging process,and a compressed air source was leveraged to enhance the related scavenging pressure.Afterward,optimal control parameters,in terms of scavenging pressure,air-fuel equivalence ratio,and ignition position,were then calibrated in this charged-scavenging condition.Eventually,the FPLG prototype has achieved a continuous stable operation of over 1000 cycles with an ignition rate of 100%and a cycle-to-cycle variation of less than 0.8%,produced an indicated power of 2.8 kW with an indicated thermal efficiency of 26%and an electrical power of 2.5 kW with an overall efficiency of 23.2%.
