Pub Date : 2022-08-24DOI: 10.1109/IAI55780.2022.9976748
J. Leventides, E. Melas, C. Poulios
We apply the Koopman operator theory and Extended Dynamic Mode Decomposition in a pair of forward and reverse chemical reactions which occur simultaneously with comparable speeds. The system of ODES which governs the evolution of the concentration of the reactants constitutes a nonlinear dynamical system with an interesting feature: It possesses uncountable infinite equilibria which reside on an algebraic surface. Koopman operator captures the dynamics of a nonlinear system, however it is infinite dimensional. In this study, we approximate the chemical reaction dynamics with a data-driven finite dimensional linear system which is defined on some augmented state space. We approximate so, with given initial conditions, the trajectories of the system and obtain an alternative description of the system based on Koopman operator theory, Extended Dynamic Mode Decomposition, and Machine Learning.
{"title":"Koopman operators and Extended Dynamic Mode Decomposition for a pair of forward and reverse chemical reactions which occur simultaneously","authors":"J. Leventides, E. Melas, C. Poulios","doi":"10.1109/IAI55780.2022.9976748","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976748","url":null,"abstract":"We apply the Koopman operator theory and Extended Dynamic Mode Decomposition in a pair of forward and reverse chemical reactions which occur simultaneously with comparable speeds. The system of ODES which governs the evolution of the concentration of the reactants constitutes a nonlinear dynamical system with an interesting feature: It possesses uncountable infinite equilibria which reside on an algebraic surface. Koopman operator captures the dynamics of a nonlinear system, however it is infinite dimensional. In this study, we approximate the chemical reaction dynamics with a data-driven finite dimensional linear system which is defined on some augmented state space. We approximate so, with given initial conditions, the trajectories of the system and obtain an alternative description of the system based on Koopman operator theory, Extended Dynamic Mode Decomposition, and Machine Learning.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129755038","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}
The electric drive system provides traction power for the entire high-speed train system, and its fault detection and diagnosis (FDD) has been widely studied. In this paper, a new method called just-in-time-learning multi-block dynamic independent comparative analysis (JITL-MBDICA) is proposed. The significant advantages of the FDD method based on JITL-MBDICA are: 1) It improves the matching ability of offline models with online data; 2) lt accurately detects faults through multiple modules; 3) It uses Support Vector Data Description (SVDD) to comprehensively analyze the detection results. The false alarms are reduced, The fault detection rate (FDR) is improved; 4) It is suitable for a non-Gaussian electric drive system. the effectiveness of JITL-MBDICA is verified on the high-speed train electric drive system.
{"title":"Just-In-Time-Learning Multi-Block Dynamic Independent Component Analysis for Electrical Drive Systems of High-Speed Trains","authors":"Xin Wang, Chao Cheng, Sheng Yang, Xiaoyue Yang, Hongtian Chen","doi":"10.1109/IAI55780.2022.9976655","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976655","url":null,"abstract":"The electric drive system provides traction power for the entire high-speed train system, and its fault detection and diagnosis (FDD) has been widely studied. In this paper, a new method called just-in-time-learning multi-block dynamic independent comparative analysis (JITL-MBDICA) is proposed. The significant advantages of the FDD method based on JITL-MBDICA are: 1) It improves the matching ability of offline models with online data; 2) lt accurately detects faults through multiple modules; 3) It uses Support Vector Data Description (SVDD) to comprehensively analyze the detection results. The false alarms are reduced, The fault detection rate (FDR) is improved; 4) It is suitable for a non-Gaussian electric drive system. the effectiveness of JITL-MBDICA is verified on the high-speed train electric drive system.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133919950","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 : 2022-08-24DOI: 10.1109/IAI55780.2022.9976816
Shuyu Wang, Shoujin Huang, N. Lu
As is well known, the cutting tool wear has a negative impact on machining precision. A precise tool wear monitoring method plays an important role in facilitating in-time cutting tool replacement, decreasing the risk of tool failure, and enhancing the machining precision. This work proposes an end-to-end approach for online tool wear monitoring based on deep learning. Firstly, a temporal convolutional network (TCN) is designed to extract features in time series from raw sensor data acquired during the cutting process. Secondly, a fully connected network is built to decode the extracted features into the exact value of tool wear. Finally, the approach is validated on PHM 2010 challenge dataset. Experimental studies show that the flank wear of the cutting tool can be monitored not only precisely, but also fast, indicating that the proposed approach has great prospects for application.
{"title":"Real-time Tool Wear Monitoring Based on A Temporal Convolutional Network","authors":"Shuyu Wang, Shoujin Huang, N. Lu","doi":"10.1109/IAI55780.2022.9976816","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976816","url":null,"abstract":"As is well known, the cutting tool wear has a negative impact on machining precision. A precise tool wear monitoring method plays an important role in facilitating in-time cutting tool replacement, decreasing the risk of tool failure, and enhancing the machining precision. This work proposes an end-to-end approach for online tool wear monitoring based on deep learning. Firstly, a temporal convolutional network (TCN) is designed to extract features in time series from raw sensor data acquired during the cutting process. Secondly, a fully connected network is built to decode the extracted features into the exact value of tool wear. Finally, the approach is validated on PHM 2010 challenge dataset. Experimental studies show that the flank wear of the cutting tool can be monitored not only precisely, but also fast, indicating that the proposed approach has great prospects for application.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"65 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132224816","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 : 2022-08-24DOI: 10.1109/IAI55780.2022.9976658
Zhiling Ren, Yun Dong, Dong Lin
This paper presents a hierarchical model predictive control method for the management of agricultural greenhouse systems. The proposed approach consists of an optimization layer and a control layer. At the optimization layer, an optimization strategy is proposed to minimize the total costs of greenhouse heating/cooling, ventilation, irrigation, carbon dioxide (CO2) supply and carbon emissions while maintaining greenhouse environmental factors, including temperature, humidity and CO2 concentration, within specified ranges. The proposed method is compared with a baseline method that minimizes greenhouse operating costs. At the control layer, a model predictive controller (MPC) is designed to track the reference trajectory obtained from the optimization layer. Simulation results show that the proposed method can reduce the total cost by R827 and the carbon emissions by 1.16 tons compared with the baseline method. Moreover, the designed MPC controller is verified to have good control performance under different levels of system disturbances. The proposed method is helpful to realize cleaner production and sustainable development of agricultural greenhouses.
{"title":"Hierarchical model predictive control for managing agricultural greenhouse systems","authors":"Zhiling Ren, Yun Dong, Dong Lin","doi":"10.1109/IAI55780.2022.9976658","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976658","url":null,"abstract":"This paper presents a hierarchical model predictive control method for the management of agricultural greenhouse systems. The proposed approach consists of an optimization layer and a control layer. At the optimization layer, an optimization strategy is proposed to minimize the total costs of greenhouse heating/cooling, ventilation, irrigation, carbon dioxide (CO2) supply and carbon emissions while maintaining greenhouse environmental factors, including temperature, humidity and CO2 concentration, within specified ranges. The proposed method is compared with a baseline method that minimizes greenhouse operating costs. At the control layer, a model predictive controller (MPC) is designed to track the reference trajectory obtained from the optimization layer. Simulation results show that the proposed method can reduce the total cost by R827 and the carbon emissions by 1.16 tons compared with the baseline method. Moreover, the designed MPC controller is verified to have good control performance under different levels of system disturbances. The proposed method is helpful to realize cleaner production and sustainable development of agricultural greenhouses.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129501226","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}
The fusion formation coefficient at the bottom of the weld bead is a key parameter to characterize the formation of a single-pass weld in ultra-narrow gap welding, and it is also an important content of welding quality control. Combined with the characteristics of the ultra-narrow gap welding method and the welding process, 14 characteristic parameters affecting the forming coefficient were extracted from the welding process signal and pre-welding preset parameters, and a convolutional neural network and a bidirectional long-short-term memory network (CNN-BILSTM-Attention) were established.) of the welding bead fusion forming coefficient prediction model, the results show that the model can effectively predict the welding bead fusion forming coefficient, and the mean square error of the prediction reaches 0.017, which provides a basis for further online control of welding quality.
{"title":"Research on prediction method of fusion forming coefficient at the bottom of ultra-narrow gap weld bead","authors":"Qian Ma, A. Zhang, Jing Ma, Yongqiang Ma, Yajun Zhang, Tingting Liang","doi":"10.1109/IAI55780.2022.9976604","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976604","url":null,"abstract":"The fusion formation coefficient at the bottom of the weld bead is a key parameter to characterize the formation of a single-pass weld in ultra-narrow gap welding, and it is also an important content of welding quality control. Combined with the characteristics of the ultra-narrow gap welding method and the welding process, 14 characteristic parameters affecting the forming coefficient were extracted from the welding process signal and pre-welding preset parameters, and a convolutional neural network and a bidirectional long-short-term memory network (CNN-BILSTM-Attention) were established.) of the welding bead fusion forming coefficient prediction model, the results show that the model can effectively predict the welding bead fusion forming coefficient, and the mean square error of the prediction reaches 0.017, which provides a basis for further online control of welding quality.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129068245","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 : 2022-08-24DOI: 10.1109/IAI55780.2022.9976699
Fan Wu, Lei Hao, Hongfeng Wang
Compared to traditional material yards with simple supply requirements and centralized material storage, intelligent material yards can significantly reduce storage space, improve material pickup efficiency, and reduce additional costs due to material mutual contamination. However, the current material delivery process is still dominated by a manual decision-making model, which is difficult to adapt to the complex and changing supply requirements. To this end, an integrated scheduling problem of material pickup and delivery considering multi-factory order requirements is proposed in this paper, which originates from a real-world scenario of Binxin intelligent material yard. By introducing the concept of spatio-temporal network flow, a discrete time-based integer linear programming model is established and then the CPLEX solver is used to solve the model. Compared with the traditional continuous-time based model, the established model shows significant advantages in terms of both solution quality and solution time, which can greatly improve the overall efficiency of the Binxin intelligent material yard.
{"title":"Joint Scheduling of Material Pickup and Delivery Towards Intelligent Material Yard","authors":"Fan Wu, Lei Hao, Hongfeng Wang","doi":"10.1109/IAI55780.2022.9976699","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976699","url":null,"abstract":"Compared to traditional material yards with simple supply requirements and centralized material storage, intelligent material yards can significantly reduce storage space, improve material pickup efficiency, and reduce additional costs due to material mutual contamination. However, the current material delivery process is still dominated by a manual decision-making model, which is difficult to adapt to the complex and changing supply requirements. To this end, an integrated scheduling problem of material pickup and delivery considering multi-factory order requirements is proposed in this paper, which originates from a real-world scenario of Binxin intelligent material yard. By introducing the concept of spatio-temporal network flow, a discrete time-based integer linear programming model is established and then the CPLEX solver is used to solve the model. Compared with the traditional continuous-time based model, the established model shows significant advantages in terms of both solution quality and solution time, which can greatly improve the overall efficiency of the Binxin intelligent material yard.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128297240","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 : 2022-08-24DOI: 10.1109/IAI55780.2022.9976657
Fei Yan, G. Gu
This paper addresses the issue of full actuation control for a class of nonlinear systems, commonly seen in engineering applications. The class of nonlinear systems involves unknown and uncertain parameters, rendering the design of feed-back controllers very challenging, especially for the full actuation control. To tackle the design issue in the presence of parameter uncertainties, the asymptotic full actuation control is proposed, aimed at achieving the full actuation control asymptotically. We first develop an adaptive control algorithm, reminiscent to the well-known backstepping control, to achieve the asymptotic global stabilization for the class of nonlinear systems, in the absence of convergence for the parameter estimates to their respective true values. The well-known recursive least-squares algorithm is then employed to estimate system parameters via sampling the output and other system signals. The asymptotic convergence of the estimates to the true system parameters and hence the asymptotic full actuation are then shown to hold for the class of nonlinear systems under some mild assumptions.
{"title":"Asymptotic Full Actuation Control for A Class of Nonlinear Systems","authors":"Fei Yan, G. Gu","doi":"10.1109/IAI55780.2022.9976657","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976657","url":null,"abstract":"This paper addresses the issue of full actuation control for a class of nonlinear systems, commonly seen in engineering applications. The class of nonlinear systems involves unknown and uncertain parameters, rendering the design of feed-back controllers very challenging, especially for the full actuation control. To tackle the design issue in the presence of parameter uncertainties, the asymptotic full actuation control is proposed, aimed at achieving the full actuation control asymptotically. We first develop an adaptive control algorithm, reminiscent to the well-known backstepping control, to achieve the asymptotic global stabilization for the class of nonlinear systems, in the absence of convergence for the parameter estimates to their respective true values. The well-known recursive least-squares algorithm is then employed to estimate system parameters via sampling the output and other system signals. The asymptotic convergence of the estimates to the true system parameters and hence the asymptotic full actuation are then shown to hold for the class of nonlinear systems under some mild assumptions.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129534417","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}
Instead of the well-known three laws of robotics that seem difficult to be applied to solving the trolley problems in the context of frame problems, this paper proposes algebraic modeling of the trolley problems on a Boolean multivalued logic so that we can analyze psychologically any knowledge simply by quasi-optimizing the truth values of logic formulae for inference in a class of Boolean algebra. Some simulation results suggest a possibility that, by introducing an atom that takes the truth values of directly killing person(s), we can control the utilitarian over-rationalization of sacrificing person(s) on AI machines.
{"title":"Algebraic Modeling of Trolley Problems on a Boolean Multivalued Logic","authors":"Jiaqi Peng, Rintaro Mizutani, Kujira Suzuki, Akira Midorikawa, Hisashi Suzuki","doi":"10.1109/IAI55780.2022.9976864","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976864","url":null,"abstract":"Instead of the well-known three laws of robotics that seem difficult to be applied to solving the trolley problems in the context of frame problems, this paper proposes algebraic modeling of the trolley problems on a Boolean multivalued logic so that we can analyze psychologically any knowledge simply by quasi-optimizing the truth values of logic formulae for inference in a class of Boolean algebra. Some simulation results suggest a possibility that, by introducing an atom that takes the truth values of directly killing person(s), we can control the utilitarian over-rationalization of sacrificing person(s) on AI machines.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129496571","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 : 2022-08-24DOI: 10.1109/IAI55780.2022.9976620
Huijuan Huang, Lin Pan
This study addresses the problem of simultaneous pickup and delivery of urban and rural networks. Big data is used to predict the congestion index of vehicles in different time periods under different road types to calculate the actual travel speed, and a time-varying road network model is established with the goal of minimizing the total cost. An improved adaptive genetic algorithm is designed and its effectiveness is verified by an example. Finally, the impact of traffic congestion level and carbon tax cost on the distribution scheme is discussed through sensitivity analysis. The results show that the improved adaptive genetic algorithm has better solution performance. Traffic congestion will affect speed, which in turn will affect the cost of delivery. The increase in carbon tax will not only affect the cost of carbon emissions but also have a negative impact on other costs and reduce corporate profits.
{"title":"Simultaneous Pickup and Delivery Vehicle Route Optimization with Time Windows under Time-varying Road Networks","authors":"Huijuan Huang, Lin Pan","doi":"10.1109/IAI55780.2022.9976620","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976620","url":null,"abstract":"This study addresses the problem of simultaneous pickup and delivery of urban and rural networks. Big data is used to predict the congestion index of vehicles in different time periods under different road types to calculate the actual travel speed, and a time-varying road network model is established with the goal of minimizing the total cost. An improved adaptive genetic algorithm is designed and its effectiveness is verified by an example. Finally, the impact of traffic congestion level and carbon tax cost on the distribution scheme is discussed through sensitivity analysis. The results show that the improved adaptive genetic algorithm has better solution performance. Traffic congestion will affect speed, which in turn will affect the cost of delivery. The increase in carbon tax will not only affect the cost of carbon emissions but also have a negative impact on other costs and reduce corporate profits.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121044152","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 : 2022-08-24DOI: 10.1109/IAI55780.2022.9976704
Yichen Zhong, Zhe Zhang, Gaochang Wu
Fused magnesium furnace (FMF) is an important equipment for producing magnesium oxide, which is prone to occurring the semi-molten abnormal condition during the production. If the abnormal condition is not predicted in time, the furnace shell will be burned through, endangering the personal safety of the staff on site. Therefore, it is necessary to predict the semi-molten abnormal condition in time and accurately. Existing machine learning-based methods adopt static models for recognizing and predicting anomaly. However, the model accuracy will degrade as data features shifting over time and melting processes. To address the above problems, this paper proposes a dynamic prediction method for semi-molten abnormal condition of multiple FMFs based on semi-supervised learning. We introduce a consistent regularization strategy and dynamically update the model weights by learning multiple FMF smelting process video data with a sparse set of condition labels. The algorithm is able to dynamically adapt to the shifted data features for accurate anomaly prediction. The proposed algorithm can predict the semi-molten abnormal condition in real time and accurately under the condition of only 1% label, enabling the safe and reliable operation of FMF.
{"title":"A Semi-Supervised Learning-based Dynamic Prediction Method for Semi-molten Condition of Fused Magnesium Furnace","authors":"Yichen Zhong, Zhe Zhang, Gaochang Wu","doi":"10.1109/IAI55780.2022.9976704","DOIUrl":"https://doi.org/10.1109/IAI55780.2022.9976704","url":null,"abstract":"Fused magnesium furnace (FMF) is an important equipment for producing magnesium oxide, which is prone to occurring the semi-molten abnormal condition during the production. If the abnormal condition is not predicted in time, the furnace shell will be burned through, endangering the personal safety of the staff on site. Therefore, it is necessary to predict the semi-molten abnormal condition in time and accurately. Existing machine learning-based methods adopt static models for recognizing and predicting anomaly. However, the model accuracy will degrade as data features shifting over time and melting processes. To address the above problems, this paper proposes a dynamic prediction method for semi-molten abnormal condition of multiple FMFs based on semi-supervised learning. We introduce a consistent regularization strategy and dynamically update the model weights by learning multiple FMF smelting process video data with a sparse set of condition labels. The algorithm is able to dynamically adapt to the shifted data features for accurate anomaly prediction. The proposed algorithm can predict the semi-molten abnormal condition in real time and accurately under the condition of only 1% label, enabling the safe and reliable operation of FMF.","PeriodicalId":138951,"journal":{"name":"2022 4th International Conference on Industrial Artificial Intelligence (IAI)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122522059","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：