Pub Date : 2023-05-01DOI: 10.1109/PHM58589.2023.00046
Bin Liao, Yi Hua, Shenrui Zhu, Fangyi Wan, X. Qing, Jie Liu
In this paper, we study a heterogeneous task assignment problem with a constraint on the number of collaborators. Existing work on task allocation pays little attention to the task’s requirement on the number of collaborators, so most algorithms may not work at all when this constraint is taken into account. First, this paper proposes a new task utility function that makes the traditional task allocation algorithm work properly. Then, this task allocation problem is modeled based on a game and an algorithm named IGreedyNE is proposed to solve this problem. IGreedyNE is a greedy strategy-based algorithm that allows multiple agents to change their game strategy simultaneously in each iteration, so it takes fewer iterations and less time to solve. Finally, we also show that the IGreedyNE algorithm converges in a finite number of iterations and returns a Nash equilibrium solution. We have performed numerous simulations, and the statistical results show that our proposed utility function can effectively handle the constraint on the number of cooperators, and our proposed IGreedyNE algorithm has a significant advantage in the speed of solving.
{"title":"An efficient algorithm for task allocation with multi-agent collaboration constraints","authors":"Bin Liao, Yi Hua, Shenrui Zhu, Fangyi Wan, X. Qing, Jie Liu","doi":"10.1109/PHM58589.2023.00046","DOIUrl":"https://doi.org/10.1109/PHM58589.2023.00046","url":null,"abstract":"In this paper, we study a heterogeneous task assignment problem with a constraint on the number of collaborators. Existing work on task allocation pays little attention to the task’s requirement on the number of collaborators, so most algorithms may not work at all when this constraint is taken into account. First, this paper proposes a new task utility function that makes the traditional task allocation algorithm work properly. Then, this task allocation problem is modeled based on a game and an algorithm named IGreedyNE is proposed to solve this problem. IGreedyNE is a greedy strategy-based algorithm that allows multiple agents to change their game strategy simultaneously in each iteration, so it takes fewer iterations and less time to solve. Finally, we also show that the IGreedyNE algorithm converges in a finite number of iterations and returns a Nash equilibrium solution. We have performed numerous simulations, and the statistical results show that our proposed utility function can effectively handle the constraint on the number of cooperators, and our proposed IGreedyNE algorithm has a significant advantage in the speed of solving.","PeriodicalId":196601,"journal":{"name":"2023 Prognostics and Health Management Conference (PHM)","volume":"234 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115660710","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/PHM58589.2023.00013
Xiaochen Gao, Xianghua Ma
To address the problem that dynamic objects, sparse environmental features, and blurred images in smart manufacturing workshops cause the performance degradation of robotic SLAM (Simultaneous Localization and Mapping) systems, semantic information and pixel-based direct method are introduced to improve the existing vision SLAM algorithm. The objects in the environment are discriminated by the target detection technique, and the results are put into the tracking thread, and the objects with high dynamic level in the results are screened twice dynamically, static points are incorporated into the matching, and dynamic points are further processed to solve the problem of effective data loss caused by the previous direct rejection of dynamic objects. To cope with the variable environment, the input data are pre-processed by an adaptive enhancement algorithm that limits the contrast, and then the camera motion is estimated by a semi-dense direct method that is insensitive to feature missing. The evaluation results on the dynamic dataset show that the error of the improved system is significantly reduced compared with ORB-SLAM2, and the estimated trajectory fits better with the real trajectory, indicating that the localization accuracy of the system is improved, and the stability and robustness are improved.
针对智能制造车间中物体动态、环境特征稀疏、图像模糊等导致机器人SLAM (Simultaneous Localization and Mapping)系统性能下降的问题，引入语义信息和基于像素的直接方法对现有视觉SLAM算法进行改进。利用目标检测技术对环境中的目标进行判别，并将结果放入跟踪线程中，对结果中动态水平较高的目标进行二次动态筛选，将静态点纳入匹配，对动态点进行进一步处理，解决了之前直接拒绝动态目标导致的有效数据丢失问题。为了应对多变的环境，输入数据通过限制对比度的自适应增强算法进行预处理，然后通过对特征缺失不敏感的半密集直接方法估计相机运动。在动态数据集上的评估结果表明，与ORB-SLAM2相比，改进后的系统误差显著减小，估计轨迹与实际轨迹拟合更好，表明系统的定位精度得到提高，稳定性和鲁棒性得到提高。
{"title":"Robot Localization and Mapping Method in Dynamic Intelligent Manufacturing Shop Environment","authors":"Xiaochen Gao, Xianghua Ma","doi":"10.1109/PHM58589.2023.00013","DOIUrl":"https://doi.org/10.1109/PHM58589.2023.00013","url":null,"abstract":"To address the problem that dynamic objects, sparse environmental features, and blurred images in smart manufacturing workshops cause the performance degradation of robotic SLAM (Simultaneous Localization and Mapping) systems, semantic information and pixel-based direct method are introduced to improve the existing vision SLAM algorithm. The objects in the environment are discriminated by the target detection technique, and the results are put into the tracking thread, and the objects with high dynamic level in the results are screened twice dynamically, static points are incorporated into the matching, and dynamic points are further processed to solve the problem of effective data loss caused by the previous direct rejection of dynamic objects. To cope with the variable environment, the input data are pre-processed by an adaptive enhancement algorithm that limits the contrast, and then the camera motion is estimated by a semi-dense direct method that is insensitive to feature missing. The evaluation results on the dynamic dataset show that the error of the improved system is significantly reduced compared with ORB-SLAM2, and the estimated trajectory fits better with the real trajectory, indicating that the localization accuracy of the system is improved, and the stability and robustness are improved.","PeriodicalId":196601,"journal":{"name":"2023 Prognostics and Health Management Conference (PHM)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126738817","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/PHM58589.2023.00060
Kun Long, Rongxin Zhang, Jianyu Long, Ning He, Yu Liu, Chuan Li
Predicting remaining useful life of rotating machineries like gears / bearings accurately is vital to guarantee safe and reliable operation of equipments. With the development of sensor technology, more and more operation state signals of equipments could be collected effectively, thus enabling to achieve considerable development in data-driven prediction method of remaining useful life. Nevertheless, existing models only considered time sequences of sàmples, but ignores spatial information among sensors when processing health state degeneration data collected by multiple sensors. To address this problem, a deep adaptive spatial-temporal graph network model was proposed to predict remaining useful life of rotating machinery. Specifically, multiple state inspection information was preprocessed firstly through time window and each slice of each time window was divided into a remaining useful life value corresponding to one sample. Secondly, the model is divided into temporal convolution layer and graph convolution layer. The former one is composed of extended causal convolution and it is used to learn time sequence information. The later one contains the learnable adjacent matrix and it was used to learn spatial information of different-state detection data. After undergoing testing on a publicly available dataset, the model’s evaluation metrics were found to be inferior to those of other high-performing prediction models. Moreover, validity of the graph convolution layer was verified through an ablation experiment.
{"title":"A Graph Neural Network-Based Method for Predicting Remaining Useful Life of Rotating Machinery","authors":"Kun Long, Rongxin Zhang, Jianyu Long, Ning He, Yu Liu, Chuan Li","doi":"10.1109/PHM58589.2023.00060","DOIUrl":"https://doi.org/10.1109/PHM58589.2023.00060","url":null,"abstract":"Predicting remaining useful life of rotating machineries like gears / bearings accurately is vital to guarantee safe and reliable operation of equipments. With the development of sensor technology, more and more operation state signals of equipments could be collected effectively, thus enabling to achieve considerable development in data-driven prediction method of remaining useful life. Nevertheless, existing models only considered time sequences of sàmples, but ignores spatial information among sensors when processing health state degeneration data collected by multiple sensors. To address this problem, a deep adaptive spatial-temporal graph network model was proposed to predict remaining useful life of rotating machinery. Specifically, multiple state inspection information was preprocessed firstly through time window and each slice of each time window was divided into a remaining useful life value corresponding to one sample. Secondly, the model is divided into temporal convolution layer and graph convolution layer. The former one is composed of extended causal convolution and it is used to learn time sequence information. The later one contains the learnable adjacent matrix and it was used to learn spatial information of different-state detection data. After undergoing testing on a publicly available dataset, the model’s evaluation metrics were found to be inferior to those of other high-performing prediction models. Moreover, validity of the graph convolution layer was verified through an ablation experiment.","PeriodicalId":196601,"journal":{"name":"2023 Prognostics and Health Management Conference (PHM)","volume":"67 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129088251","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/PHM58589.2023.00043
D. Jiang, Fangyi Wan, W. Cui, Shuhai Jiang, Yajie Han, Tian Chen
The drag parachute lock system is one of the most important parts of an aircraft’s take-off and landing system. By minimizing the landing glide distance and extending the life of the wheels, this system’s reliability has a direct impact on the safety of the aircraft’s landing. In order to construct a three-dimensional model of the drag bail lock system and a multi-body dynamics solution model using CATIA and LMS Virtual, this research undertakes an extensive investigation of its essential components and operating principle. Based on the multi-body dynamics model, lab software and a parameterized reliability simulation model are developed. The failure mode of unintentional parachute release is sampled using the Monte Carlo method, the model response amount is calculated using simulation, and the reliability study of the parachute lock system is then carried out using the findings of the simulation.
{"title":"Failure Simulation and Reliability Modelling Analysis of Aircraft Drag Parachute Lock System","authors":"D. Jiang, Fangyi Wan, W. Cui, Shuhai Jiang, Yajie Han, Tian Chen","doi":"10.1109/PHM58589.2023.00043","DOIUrl":"https://doi.org/10.1109/PHM58589.2023.00043","url":null,"abstract":"The drag parachute lock system is one of the most important parts of an aircraft’s take-off and landing system. By minimizing the landing glide distance and extending the life of the wheels, this system’s reliability has a direct impact on the safety of the aircraft’s landing. In order to construct a three-dimensional model of the drag bail lock system and a multi-body dynamics solution model using CATIA and LMS Virtual, this research undertakes an extensive investigation of its essential components and operating principle. Based on the multi-body dynamics model, lab software and a parameterized reliability simulation model are developed. The failure mode of unintentional parachute release is sampled using the Monte Carlo method, the model response amount is calculated using simulation, and the reliability study of the parachute lock system is then carried out using the findings of the simulation.","PeriodicalId":196601,"journal":{"name":"2023 Prognostics and Health Management Conference (PHM)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134131334","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/PHM58589.2023.00015
Yalong Feng, Xu-yun Fu, Lijun Wang, Z. Bai, Rui Wang, Hai Chen
Reasonable life cycle maintenance decision of aeroengine, determining the aeroengine maintenance interval and maintenance workscope, can effectively reduce aeroengine maintenance costs. To achieve this, an optimization model of maintenance decision of aeroengine life cycle is established, taking the lowest total maintenance cost in the life cycle as the optimization objective, and using the aeroengine life cycle maintenance interval and maintenance workscope as decision variables. In order to reduce the size of the model’s solution space, the maintenance interval and the maintenance workscope are decoupled, and the optimal maintenance strategy to determine the maintenance workscope is proposed. Subsequently, particle swarm optimization algorithm is used to search the global optimal solution of the model. Finally, the effectiveness of the model is evaluated according to relevant numerical experiments and real aeroengine data. The results show that a better solution can be obtained in a short time for problems within 30 life limited parts, 28 modules and 90000 flight cycles.
{"title":"Problem Decoupling and Optimization of Aeroengine Life Cycle Maintenance Decision","authors":"Yalong Feng, Xu-yun Fu, Lijun Wang, Z. Bai, Rui Wang, Hai Chen","doi":"10.1109/PHM58589.2023.00015","DOIUrl":"https://doi.org/10.1109/PHM58589.2023.00015","url":null,"abstract":"Reasonable life cycle maintenance decision of aeroengine, determining the aeroengine maintenance interval and maintenance workscope, can effectively reduce aeroengine maintenance costs. To achieve this, an optimization model of maintenance decision of aeroengine life cycle is established, taking the lowest total maintenance cost in the life cycle as the optimization objective, and using the aeroengine life cycle maintenance interval and maintenance workscope as decision variables. In order to reduce the size of the model’s solution space, the maintenance interval and the maintenance workscope are decoupled, and the optimal maintenance strategy to determine the maintenance workscope is proposed. Subsequently, particle swarm optimization algorithm is used to search the global optimal solution of the model. Finally, the effectiveness of the model is evaluated according to relevant numerical experiments and real aeroengine data. The results show that a better solution can be obtained in a short time for problems within 30 life limited parts, 28 modules and 90000 flight cycles.","PeriodicalId":196601,"journal":{"name":"2023 Prognostics and Health Management Conference (PHM)","volume":"67 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115529012","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/PHM58589.2023.00010
Yingshun Li, Yanni Zhang, Zhannan Guo, Aina Wang
In order to efficiently diagnose the mechanical wear failure of aero-engine lubricating oil systems, a base KPCA-ABC-SVM fault diagnosis model is established based on the number of metal abrasive particles considering multiple indicators such as viscosity, temperature, moisture and density. Firstly, the fault detection results obtained by the feature extraction of multi-parameters by kernel principal component analysis (KPCA) method are used as a reference, and then the extracted feature values are classified by the support vector machine (SVM); finally, the penalty factor and kernel function parameters of SVM are optimally selected by using the artificial bee colony (ABC) algorithm to obtain the fault diagnosis with the highest accuracy. Experiments show that support vector machine classification modified by artificial bee colony algorithm can effectively improve the fault detection accuracy after feature extraction.
{"title":"Fault Diagnosis of Aero-engine Lubrication System Based on KPCA-ABC-SVM","authors":"Yingshun Li, Yanni Zhang, Zhannan Guo, Aina Wang","doi":"10.1109/PHM58589.2023.00010","DOIUrl":"https://doi.org/10.1109/PHM58589.2023.00010","url":null,"abstract":"In order to efficiently diagnose the mechanical wear failure of aero-engine lubricating oil systems, a base KPCA-ABC-SVM fault diagnosis model is established based on the number of metal abrasive particles considering multiple indicators such as viscosity, temperature, moisture and density. Firstly, the fault detection results obtained by the feature extraction of multi-parameters by kernel principal component analysis (KPCA) method are used as a reference, and then the extracted feature values are classified by the support vector machine (SVM); finally, the penalty factor and kernel function parameters of SVM are optimally selected by using the artificial bee colony (ABC) algorithm to obtain the fault diagnosis with the highest accuracy. Experiments show that support vector machine classification modified by artificial bee colony algorithm can effectively improve the fault detection accuracy after feature extraction.","PeriodicalId":196601,"journal":{"name":"2023 Prognostics and Health Management Conference (PHM)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115733290","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/PHM58589.2023.00041
A. Mosallam, Jinlong Kang, Fares Ben Youssef, L. Laval, James L. Fulton
This paper presents a data-driven fault diagnosis method for neutron generator systems in logging-while-drilling tools. Specifically, the nuclear system’s main failure modes and associated electronic boards are first identified, and then statistical features of the selected boards are extracted based on expert knowledge. The extracted features discriminate between healthy and faulty behavior for each board. Finally, machine learning models are used to map the relationship between the extracted features and the labels of the corresponding sensor data for each board. This method is validated using data collected from actual oil well drilling operations, and the experimental results show that the method is effective. This work is part of a long-term project aiming to construct a digital fleet management system for drilling tools.
{"title":"Data-Driven Fault Diagnostics for Neutron Generator Systems in Multifunction Logging-While-Drilling Service","authors":"A. Mosallam, Jinlong Kang, Fares Ben Youssef, L. Laval, James L. Fulton","doi":"10.1109/PHM58589.2023.00041","DOIUrl":"https://doi.org/10.1109/PHM58589.2023.00041","url":null,"abstract":"This paper presents a data-driven fault diagnosis method for neutron generator systems in logging-while-drilling tools. Specifically, the nuclear system’s main failure modes and associated electronic boards are first identified, and then statistical features of the selected boards are extracted based on expert knowledge. The extracted features discriminate between healthy and faulty behavior for each board. Finally, machine learning models are used to map the relationship between the extracted features and the labels of the corresponding sensor data for each board. This method is validated using data collected from actual oil well drilling operations, and the experimental results show that the method is effective. This work is part of a long-term project aiming to construct a digital fleet management system for drilling tools.","PeriodicalId":196601,"journal":{"name":"2023 Prognostics and Health Management Conference (PHM)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123585343","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/PHM58589.2023.00042
Jiaqi Shi, Hongmei Shi, Jianbo Li
Considering an operational scenario that a freight train composed of four wagons runs over a three-span heavy haul railway bridge, the feasibility of detecting bridge damage from bogie responses is investigated and a three-stage indirect damage diagnosis method is put forward. In the data preparation stage, the time-domain subtraction method (TSM) and the Empirical Mode Decomposition (EMD) algorithm are applied to suppress the adverse effect of track irregularity and the vibration coupling effect among spans on the bogie accelerations, respectively. In the damage detection stage, a damage indicator based on the Mahalanobis distance is used to describe the dissimilarity between the train crossings in baseline status and damage status, so as to detect the occurrence of damage. In the damage localization stage, the moving window strategy is exploited to complement preliminary diagnosis with locational information. In order to appraise the efficiency of the proposed method, a blind test is carried out using the supplied data measurements without awareness of the relevant damage information. Regardless of damage location and severity, the results indicate that the proposed method simultaneously has high efficiency and superiority for damage detection and localization.
{"title":"A three-stage damage diagnosis method for heavy haul railway bridge by bogie response measurements","authors":"Jiaqi Shi, Hongmei Shi, Jianbo Li","doi":"10.1109/PHM58589.2023.00042","DOIUrl":"https://doi.org/10.1109/PHM58589.2023.00042","url":null,"abstract":"Considering an operational scenario that a freight train composed of four wagons runs over a three-span heavy haul railway bridge, the feasibility of detecting bridge damage from bogie responses is investigated and a three-stage indirect damage diagnosis method is put forward. In the data preparation stage, the time-domain subtraction method (TSM) and the Empirical Mode Decomposition (EMD) algorithm are applied to suppress the adverse effect of track irregularity and the vibration coupling effect among spans on the bogie accelerations, respectively. In the damage detection stage, a damage indicator based on the Mahalanobis distance is used to describe the dissimilarity between the train crossings in baseline status and damage status, so as to detect the occurrence of damage. In the damage localization stage, the moving window strategy is exploited to complement preliminary diagnosis with locational information. In order to appraise the efficiency of the proposed method, a blind test is carried out using the supplied data measurements without awareness of the relevant damage information. Regardless of damage location and severity, the results indicate that the proposed method simultaneously has high efficiency and superiority for damage detection and localization.","PeriodicalId":196601,"journal":{"name":"2023 Prognostics and Health Management Conference (PHM)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132167898","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/PHM58589.2023.00030
J. Fang, Xiang Lin, Fengxiang Zhou, Yan Tian, Min Zhang
Whether employees wear safety helmets is an important safety issue in power related work scenarios, and various safety issues can be avoided by monitoring this situation. However, traditional target detection methods are vulnerable to interference due to the weather, light, personnel density, location of surveillance cameras and other problems in the working environment, and the recognition and detection effect of such small targets is not very good. Therefore, this paper uses the high-precision YOLOv5 (You Only Look Once) as the target detection framework, and modifies its backbone network to improve its ability in small target recognition. The original backbone structure is cut and compressed, and the SwinT (Swin Transformer) modules are added to improve the overall recognition accuracy based on its powerful small target recognition ability. At the same time, SE (Squeeze and Excitation) and CBAM (Convolutional Block Attention Module) modules are added to further improve the recognition accuracy of the entire network. Finally, experiments are conducted on the SHWD (Safety Helmet Wearing Dataset) dataset. The experimental results show that compared to the network before modification, the accuracy of the optimized YOLO structure proposed in this paper is significantly improved on the validation dataset, with an average recognition accuracy of 93%.
员工是否戴安全帽是电力相关工作场景中一个重要的安全问题，通过监控这种情况可以避免各种安全问题。但是，传统的目标检测方法在工作环境中容易受到天气、光线、人员密度、监控摄像机位置等问题的干扰，对这类小目标的识别和检测效果不是很好。因此，本文采用高精度的YOLOv5 (You Only Look Once)作为目标检测框架，并对其骨干网进行修改，提高其对小目标的识别能力。对原有主干结构进行剪切压缩，并在其强大的小目标识别能力的基础上加入SwinT (Swin Transformer)模块，提高整体识别精度。同时，增加了SE (Squeeze and Excitation)和CBAM (Convolutional Block Attention Module)模块，进一步提高了整个网络的识别精度。最后，在SHWD (Safety Helmet Wearing Dataset)数据集上进行了实验。实验结果表明，与修改前的网络相比，本文提出的优化YOLO结构在验证数据集上的准确率显著提高，平均识别准确率达到93%。
{"title":"Safety Helmet Detection Based on Optimized YOLOv5","authors":"J. Fang, Xiang Lin, Fengxiang Zhou, Yan Tian, Min Zhang","doi":"10.1109/PHM58589.2023.00030","DOIUrl":"https://doi.org/10.1109/PHM58589.2023.00030","url":null,"abstract":"Whether employees wear safety helmets is an important safety issue in power related work scenarios, and various safety issues can be avoided by monitoring this situation. However, traditional target detection methods are vulnerable to interference due to the weather, light, personnel density, location of surveillance cameras and other problems in the working environment, and the recognition and detection effect of such small targets is not very good. Therefore, this paper uses the high-precision YOLOv5 (You Only Look Once) as the target detection framework, and modifies its backbone network to improve its ability in small target recognition. The original backbone structure is cut and compressed, and the SwinT (Swin Transformer) modules are added to improve the overall recognition accuracy based on its powerful small target recognition ability. At the same time, SE (Squeeze and Excitation) and CBAM (Convolutional Block Attention Module) modules are added to further improve the recognition accuracy of the entire network. Finally, experiments are conducted on the SHWD (Safety Helmet Wearing Dataset) dataset. The experimental results show that compared to the network before modification, the accuracy of the optimized YOLO structure proposed in this paper is significantly improved on the validation dataset, with an average recognition accuracy of 93%.","PeriodicalId":196601,"journal":{"name":"2023 Prognostics and Health Management Conference (PHM)","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120925968","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/PHM58589.2023.00036
Yingshun Li, Fansen Kong, Huanhuan Sui, De-biao Wang
With the rapid development of computer technology, the integration degree of weapon fire control system is getting higher and higher, and the health management of weapon fire control system is also put forward higher requirements. Fault prediction technology is an important link and key technology in weapon fire control system health management. It can improve the early fault identification and diagnosis ability of weapon fire control system by monitoring the fault characteristic and symptom information, and realize the effective fault prevention. Therefore, predictive fault detection algorithms for fire control have been proposed in recent years. This paper will summarize these fault detection algorithms.
{"title":"Summary of Fault Prediction Algorithms for Fire Control System","authors":"Yingshun Li, Fansen Kong, Huanhuan Sui, De-biao Wang","doi":"10.1109/PHM58589.2023.00036","DOIUrl":"https://doi.org/10.1109/PHM58589.2023.00036","url":null,"abstract":"With the rapid development of computer technology, the integration degree of weapon fire control system is getting higher and higher, and the health management of weapon fire control system is also put forward higher requirements. Fault prediction technology is an important link and key technology in weapon fire control system health management. It can improve the early fault identification and diagnosis ability of weapon fire control system by monitoring the fault characteristic and symptom information, and realize the effective fault prevention. Therefore, predictive fault detection algorithms for fire control have been proposed in recent years. This paper will summarize these fault detection algorithms.","PeriodicalId":196601,"journal":{"name":"2023 Prognostics and Health Management Conference (PHM)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127461541","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：