Pub Date : 2020-06-01DOI: 10.1109/ICPHM49022.2020.9187063
Jackson Cornelius, Blake Brockner, Seong Hyeon Hong, Yi Wang, K. Pant, J. Ball
Many researchers in the prognostics and health management community have begun exploring the use of deep neural networks for predicting remaining useful life (RUL) of mechanical systems. These models have consistently reestablished the state-of-the-art in RUL prediction performance on common benchmarks, such as the NASA C-MAPSS Aircraft Engine dataset. However, they do not attempt to capture the multiple sources of uncertainty that are inherent in their predictions. This paper presents an approach for estimating both epistemic and heteroscedastic aleatoric uncertainties that emerge in deep neural network models that are trained for RUL prediction and demonstrates that quantifying their overall impact on predictions can be extremely valuable in real-world systems, where decisions are sometimes made during uncertain operating conditions. First, a novel deep neural network architecture is proposed that demonstrates competitive performance on the NASA C-MAPSS FD001 and FD003 datasets. Then, this network is adapted to estimate epistemic and heteroscedastic aleatoric uncertainties in the RUL prediction problem. Finally, a study is carried out to observe the effects that augmenting the RUL truth data, i.e. utilizing piecewise linear truth curves in place of the actual truth data, have on the perceived uncertainties in the system. Case studies on the C-MAPSS FD001 dataset will show that utilizing the actual RUL truth data can yield more meaningful uncertainty estimates and more insight into the relationship between sensor data and an engine's time-to-failure.
{"title":"Estimating and Leveraging Uncertainties in Deep Learning for Remaining Useful Life Prediction in Mechanical Systems","authors":"Jackson Cornelius, Blake Brockner, Seong Hyeon Hong, Yi Wang, K. Pant, J. Ball","doi":"10.1109/ICPHM49022.2020.9187063","DOIUrl":"https://doi.org/10.1109/ICPHM49022.2020.9187063","url":null,"abstract":"Many researchers in the prognostics and health management community have begun exploring the use of deep neural networks for predicting remaining useful life (RUL) of mechanical systems. These models have consistently reestablished the state-of-the-art in RUL prediction performance on common benchmarks, such as the NASA C-MAPSS Aircraft Engine dataset. However, they do not attempt to capture the multiple sources of uncertainty that are inherent in their predictions. This paper presents an approach for estimating both epistemic and heteroscedastic aleatoric uncertainties that emerge in deep neural network models that are trained for RUL prediction and demonstrates that quantifying their overall impact on predictions can be extremely valuable in real-world systems, where decisions are sometimes made during uncertain operating conditions. First, a novel deep neural network architecture is proposed that demonstrates competitive performance on the NASA C-MAPSS FD001 and FD003 datasets. Then, this network is adapted to estimate epistemic and heteroscedastic aleatoric uncertainties in the RUL prediction problem. Finally, a study is carried out to observe the effects that augmenting the RUL truth data, i.e. utilizing piecewise linear truth curves in place of the actual truth data, have on the perceived uncertainties in the system. Case studies on the C-MAPSS FD001 dataset will show that utilizing the actual RUL truth data can yield more meaningful uncertainty estimates and more insight into the relationship between sensor data and an engine's time-to-failure.","PeriodicalId":148899,"journal":{"name":"2020 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129741418","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 : 2020-06-01DOI: 10.1109/ICPHM49022.2020.9187034
Subrata Mukherjee, Xuhui Huang, V. Rathod, L. Udpa, Y. Deng
Pipe infrastructure systems in service continue to degrade with passage of time. As the defects grow with time, for safety purposes, they have to be inspected periodically for detection of harmful defects. This paper presents development of a novel method for identifying defect growth using dynamically updated transfer learning technique on data from magnetic flux leakage (MFL) sensors. The operation of pipeline inspection gauge (PIG) within the pipeline to collect accurate, low noise readings for defect detection is expensive and time-consuming. Running probes within the operational pipeline produces noisy data. In this paper we consider a less noisy and time-consuming baseline readings within pipelines taken in the beginning. Using the baseline data, our goal is to first automatically detect the defective areas during inspection and thereafter monitor the growth of those defects. Based on the baseline data, a bivariate function was estimated using a function estimation method based on mixture regression framework to compute posterior probabilities of the defects at each scanning point. Thereafter, it is seen that applying direct function estimation with noisy field data on subsequent inspections is not effective. We use transfer learning perspectives by leveraging the defect probabilities and location from the previous inspections, and then consequently update those probabilities based on current data by applying a dynamically updated transfer learning technique. The defect growth is dynamically tracked and characterized with high accuracy and sensitivity.
{"title":"Defects Tracking via NDE Based Transfer Learning","authors":"Subrata Mukherjee, Xuhui Huang, V. Rathod, L. Udpa, Y. Deng","doi":"10.1109/ICPHM49022.2020.9187034","DOIUrl":"https://doi.org/10.1109/ICPHM49022.2020.9187034","url":null,"abstract":"Pipe infrastructure systems in service continue to degrade with passage of time. As the defects grow with time, for safety purposes, they have to be inspected periodically for detection of harmful defects. This paper presents development of a novel method for identifying defect growth using dynamically updated transfer learning technique on data from magnetic flux leakage (MFL) sensors. The operation of pipeline inspection gauge (PIG) within the pipeline to collect accurate, low noise readings for defect detection is expensive and time-consuming. Running probes within the operational pipeline produces noisy data. In this paper we consider a less noisy and time-consuming baseline readings within pipelines taken in the beginning. Using the baseline data, our goal is to first automatically detect the defective areas during inspection and thereafter monitor the growth of those defects. Based on the baseline data, a bivariate function was estimated using a function estimation method based on mixture regression framework to compute posterior probabilities of the defects at each scanning point. Thereafter, it is seen that applying direct function estimation with noisy field data on subsequent inspections is not effective. We use transfer learning perspectives by leveraging the defect probabilities and location from the previous inspections, and then consequently update those probabilities based on current data by applying a dynamically updated transfer learning technique. The defect growth is dynamically tracked and characterized with high accuracy and sensitivity.","PeriodicalId":148899,"journal":{"name":"2020 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"335 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124714346","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 : 2020-06-01DOI: 10.1109/ICPHM49022.2020.9187042
Md Shahin Alam, S. A. Arefifar
Energy storage systems (ESSs) facilitate high penetration and stable operation of renewable energy sources (RESs) in power distribution grids. ESSs could reduce the distribution system operational costs, decrease power and energy losses, reduce emissions, and increase the system efficiency. This research proposes an optimal energy management approach that considers both ESSs capacities and renewable energy resources impacts on distribution system operational performances. Different capacities of ESS along with various penetration level of PV and wind generators are considered and the system performance is evaluated. The system performance is investigated in terms of operational costs, power losses and emissions. The well-known PG&E 69-bus power distribution system is chosen for analysis. For more accurate results, the uncertain characteristics of PV and wind are taken into account while applying EMS during optimal operation of ESSs. Several case studies are created considering PVs and wind generations separately and collectively for ESSs optimal operations. Moreover, sensitivity studies has been done for calculating yearly system performance improvements in dollar values to validate the proposed EMS approach with ESS. The results demonstrate that efficient operation of ESSs along with EMS can considerably reduce distribution system operational costs, system losses, and environmental emissions.
{"title":"Optimal Operation of Energy Storage Units in PV and Wind Integrated Smart Distribution Systems","authors":"Md Shahin Alam, S. A. Arefifar","doi":"10.1109/ICPHM49022.2020.9187042","DOIUrl":"https://doi.org/10.1109/ICPHM49022.2020.9187042","url":null,"abstract":"Energy storage systems (ESSs) facilitate high penetration and stable operation of renewable energy sources (RESs) in power distribution grids. ESSs could reduce the distribution system operational costs, decrease power and energy losses, reduce emissions, and increase the system efficiency. This research proposes an optimal energy management approach that considers both ESSs capacities and renewable energy resources impacts on distribution system operational performances. Different capacities of ESS along with various penetration level of PV and wind generators are considered and the system performance is evaluated. The system performance is investigated in terms of operational costs, power losses and emissions. The well-known PG&E 69-bus power distribution system is chosen for analysis. For more accurate results, the uncertain characteristics of PV and wind are taken into account while applying EMS during optimal operation of ESSs. Several case studies are created considering PVs and wind generations separately and collectively for ESSs optimal operations. Moreover, sensitivity studies has been done for calculating yearly system performance improvements in dollar values to validate the proposed EMS approach with ESS. The results demonstrate that efficient operation of ESSs along with EMS can considerably reduce distribution system operational costs, system losses, and environmental emissions.","PeriodicalId":148899,"journal":{"name":"2020 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"120 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132553641","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 : 2020-06-01DOI: 10.1109/ICPHM49022.2020.9187050
Lindy Williams, Caleb Phillips, S. Sheng, A. Dobos, Xiupeng Wei
Operation and maintenance (O&M) costs for wind turbines pose a risk to competitiveness and asset owners. With machine-learning technologies and digitalization rapidly maturing, the wind industry is actively investigating these new technologies to optimize O&M practices and reduce costs. This paper reviews recent work on machine-learning approaches to generator bearing failure prediction and presents a relevant real-world case study through a collaboration between the National Renewable Energy Laboratory and Envision Digital Corporation. In the case study, we evaluate the performance of representative machine-learning algorithms for predicting wind turbine generator bearing failures. Operational supervisory control and data acquisition data from one wind power plant was used to train and test the machine-learning models. The investigated data channels are chosen based on whether physically they reflect the failed generator bearing conditions and the component historical usage, including both environmental and operational conditions. Benefits and drawbacks of different methods are identified.
{"title":"Scalable Wind Turbine Generator Bearing Fault Prediction Using Machine Learning: A Case Study","authors":"Lindy Williams, Caleb Phillips, S. Sheng, A. Dobos, Xiupeng Wei","doi":"10.1109/ICPHM49022.2020.9187050","DOIUrl":"https://doi.org/10.1109/ICPHM49022.2020.9187050","url":null,"abstract":"Operation and maintenance (O&M) costs for wind turbines pose a risk to competitiveness and asset owners. With machine-learning technologies and digitalization rapidly maturing, the wind industry is actively investigating these new technologies to optimize O&M practices and reduce costs. This paper reviews recent work on machine-learning approaches to generator bearing failure prediction and presents a relevant real-world case study through a collaboration between the National Renewable Energy Laboratory and Envision Digital Corporation. In the case study, we evaluate the performance of representative machine-learning algorithms for predicting wind turbine generator bearing failures. Operational supervisory control and data acquisition data from one wind power plant was used to train and test the machine-learning models. The investigated data channels are chosen based on whether physically they reflect the failed generator bearing conditions and the component historical usage, including both environmental and operational conditions. Benefits and drawbacks of different methods are identified.","PeriodicalId":148899,"journal":{"name":"2020 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133106647","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 : 2020-06-01DOI: 10.1109/ICPHM49022.2020.9187053
Mohamed Ragab, Zhenghua Chen, Min Wu, C. Kwoh, Xiaoli Li
Remaining useful life (RUL) prediction is a key task for realizing predictive maintenance for industrial machines/assets. Accurate RUL prediction enables prior maintenance scheduling that can reduce downtime, reduce maintenance costs, and increase machine availability. Data-driven approaches have a widely acclaimed performance on RUL prediction of industrial machines. Usually, they assume that data used in training and testing phases are drawn from the same distribution. However, machines may work under different conditions (i.e., data distribution) for training and testing phases. As a result, the model performing well during training can deteriorate significantly during testing. Naive recollection and re-annotation of data for each new working condition can be very expensive and obviously not a viable solution. To alleviate this problem, we rely on a transfer learning approach called domain adaptation to transfer the knowledge learned from one labelled operating condition (source domain) to another operating condition (target domain) without labels. Particularly, we propose a novel adversarial domain adaption approach for remaining useful life prediction, named ADARUL, which can work on the data from different working conditions or different fault modes. This approach is built on top of a bidirectional deep long short-term memory (LSTM) network that can model the temporal dependency and extract representative features. Moreover, it derives invariant representation among the working conditions by removing the domain-specific information while keeping the task-specific information. We have conducted comprehensive experiments among four different datasets of turbofan engines. The experiments show that our proposed method significantly outperforms the state-of-the-art methods..
{"title":"Adversarial Transfer Learning for Machine Remaining Useful Life Prediction","authors":"Mohamed Ragab, Zhenghua Chen, Min Wu, C. Kwoh, Xiaoli Li","doi":"10.1109/ICPHM49022.2020.9187053","DOIUrl":"https://doi.org/10.1109/ICPHM49022.2020.9187053","url":null,"abstract":"Remaining useful life (RUL) prediction is a key task for realizing predictive maintenance for industrial machines/assets. Accurate RUL prediction enables prior maintenance scheduling that can reduce downtime, reduce maintenance costs, and increase machine availability. Data-driven approaches have a widely acclaimed performance on RUL prediction of industrial machines. Usually, they assume that data used in training and testing phases are drawn from the same distribution. However, machines may work under different conditions (i.e., data distribution) for training and testing phases. As a result, the model performing well during training can deteriorate significantly during testing. Naive recollection and re-annotation of data for each new working condition can be very expensive and obviously not a viable solution. To alleviate this problem, we rely on a transfer learning approach called domain adaptation to transfer the knowledge learned from one labelled operating condition (source domain) to another operating condition (target domain) without labels. Particularly, we propose a novel adversarial domain adaption approach for remaining useful life prediction, named ADARUL, which can work on the data from different working conditions or different fault modes. This approach is built on top of a bidirectional deep long short-term memory (LSTM) network that can model the temporal dependency and extract representative features. Moreover, it derives invariant representation among the working conditions by removing the domain-specific information while keeping the task-specific information. We have conducted comprehensive experiments among four different datasets of turbofan engines. The experiments show that our proposed method significantly outperforms the state-of-the-art methods..","PeriodicalId":148899,"journal":{"name":"2020 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"88 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133645200","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 : 2020-06-01DOI: 10.1109/ICPHM49022.2020.9187026
Kesaaki Minemura, S. Yabui, Kohei Iwata, T. Inoue
A technique for visualization of a gear-motor shaft’s whirling feature is proposed based on time-series analysis for rotary machine component condition monitoring. It is necessary to develop many technological elements, including machine components, the Internet of Things (IoT), sensing, signal processing and modeling for machine component condition monitoring. When a machine component is connected to another device, the machine component’s features change because of the connection. Specifically, this work considers the case of a machine component where the shaft around the axis connecting the component to another device does not form a circular orbit. It is assumed that the shaft does not have a circular orbit and it is thus necessary to visualize the shaft using a signal processing technique based on this assumption. In general methods, however, because a constant speed and circular orbit are assumed, some errors occur because of the noncircular orbit. In this paper, we consider visualization using a signal processing technique that focuses on the rotational axis, particularly for connections between rotary machine components for condition monitoring. In the proposed method, a time waveform is converted into polar coordinates and expressed in terms of its amplitude and angular direction. By calculating the density distribution for each angle, the features are confirmed even if the shaft orbit does not become a circle. Furthermore, it aids in judging whether the feature change has followed a machine component condition change in the trajectory. Measurement data were obtained through verification experiments. It is confirmed that the density distribution’s relative standard deviation is less than approximately 0.05 and that the orbit is constant under normal conditions. From the experimental results, it is confirmed that the proposed signal processing method is thus effective for machine component condition monitoring.
{"title":"Visualization of gear-motor shaft whirling feature based on time-series analysis for rotary machine component condition monitoring","authors":"Kesaaki Minemura, S. Yabui, Kohei Iwata, T. Inoue","doi":"10.1109/ICPHM49022.2020.9187026","DOIUrl":"https://doi.org/10.1109/ICPHM49022.2020.9187026","url":null,"abstract":"A technique for visualization of a gear-motor shaft’s whirling feature is proposed based on time-series analysis for rotary machine component condition monitoring. It is necessary to develop many technological elements, including machine components, the Internet of Things (IoT), sensing, signal processing and modeling for machine component condition monitoring. When a machine component is connected to another device, the machine component’s features change because of the connection. Specifically, this work considers the case of a machine component where the shaft around the axis connecting the component to another device does not form a circular orbit. It is assumed that the shaft does not have a circular orbit and it is thus necessary to visualize the shaft using a signal processing technique based on this assumption. In general methods, however, because a constant speed and circular orbit are assumed, some errors occur because of the noncircular orbit. In this paper, we consider visualization using a signal processing technique that focuses on the rotational axis, particularly for connections between rotary machine components for condition monitoring. In the proposed method, a time waveform is converted into polar coordinates and expressed in terms of its amplitude and angular direction. By calculating the density distribution for each angle, the features are confirmed even if the shaft orbit does not become a circle. Furthermore, it aids in judging whether the feature change has followed a machine component condition change in the trajectory. Measurement data were obtained through verification experiments. It is confirmed that the density distribution’s relative standard deviation is less than approximately 0.05 and that the orbit is constant under normal conditions. From the experimental results, it is confirmed that the proposed signal processing method is thus effective for machine component condition monitoring.","PeriodicalId":148899,"journal":{"name":"2020 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122606181","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 : 2020-06-01DOI: 10.1109/ICPHM49022.2020.9187062
Pan Luo, Z. Li
This paper reviews and looks into three case studies of Internet of Things (IoT) based engineering applications, i.e., the artificial pancreas device, vehicle-to-vehicle communication, and the Boeing’s flight control system of maneuvering characteristics augmentation system (MCAS). These applications span from emerging medical devices to intelligent transportation to advanced system control domains. We observe that all three investigated applications are built on the five primitives of IoT, which are sensor, aggregator, communication channel, external utility, and decision trigger. Through a thorough investigation and abstraction of the three case studies of IoT applications, one key research question is identified as the fusion of multiple streams of different frequency data inputs. A comprehensive literature review on multi-rate multi-sensor data fusion is presented. Lastly, additional IoT induced research challenges and opportunities are discussed and summarized.
{"title":"A Review of Internet of Things (IoT) based Engineering Applications and Data Fusion Challenges for Multi-rate Multi-sensor Systems","authors":"Pan Luo, Z. Li","doi":"10.1109/ICPHM49022.2020.9187062","DOIUrl":"https://doi.org/10.1109/ICPHM49022.2020.9187062","url":null,"abstract":"This paper reviews and looks into three case studies of Internet of Things (IoT) based engineering applications, i.e., the artificial pancreas device, vehicle-to-vehicle communication, and the Boeing’s flight control system of maneuvering characteristics augmentation system (MCAS). These applications span from emerging medical devices to intelligent transportation to advanced system control domains. We observe that all three investigated applications are built on the five primitives of IoT, which are sensor, aggregator, communication channel, external utility, and decision trigger. Through a thorough investigation and abstraction of the three case studies of IoT applications, one key research question is identified as the fusion of multiple streams of different frequency data inputs. A comprehensive literature review on multi-rate multi-sensor data fusion is presented. Lastly, additional IoT induced research challenges and opportunities are discussed and summarized.","PeriodicalId":148899,"journal":{"name":"2020 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127841147","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 : 2019-12-31DOI: 10.1109/ICPHM49022.2020.9187036
W. Shalaby, Adriano Arantes, Teresa GonzalezDiaz, Chetan Gupta
Popular conversational agents frameworks such as Alexa Skills Kit (ASK) and Google Actions (gActions) offer unprecedented opportunities for facilitating the development and deployment of voice-enabled AI solutions in various verticals. Nevertheless, understanding user utterances with high accuracy remains a challenging task with these frameworks. Particularly, when building chatbots with large volume of domain-specific entities. In this paper, we describe the challenges and lessons learned from building a large scale virtual assistant for understanding and responding to equipment-related complaints. In the process, we describe an alternative scalable framework for: 1) extracting the knowledge about equipment components and their associated problem entities from short texts, and 2) learning to identify such entities in user utterances. We show through evaluation on a real dataset that the proposed framework, compared to off-the-shelf popular ones, scales better with large volume of entities being up to 30% more accurate, and is more effective in understanding user utterances with domain-specific entities.
{"title":"Building chatbots from large scale domain-specific knowledge bases: challenges and opportunities","authors":"W. Shalaby, Adriano Arantes, Teresa GonzalezDiaz, Chetan Gupta","doi":"10.1109/ICPHM49022.2020.9187036","DOIUrl":"https://doi.org/10.1109/ICPHM49022.2020.9187036","url":null,"abstract":"Popular conversational agents frameworks such as Alexa Skills Kit (ASK) and Google Actions (gActions) offer unprecedented opportunities for facilitating the development and deployment of voice-enabled AI solutions in various verticals. Nevertheless, understanding user utterances with high accuracy remains a challenging task with these frameworks. Particularly, when building chatbots with large volume of domain-specific entities. In this paper, we describe the challenges and lessons learned from building a large scale virtual assistant for understanding and responding to equipment-related complaints. In the process, we describe an alternative scalable framework for: 1) extracting the knowledge about equipment components and their associated problem entities from short texts, and 2) learning to identify such entities in user utterances. We show through evaluation on a real dataset that the proposed framework, compared to off-the-shelf popular ones, scales better with large volume of entities being up to 30% more accurate, and is more effective in understanding user utterances with domain-specific entities.","PeriodicalId":148899,"journal":{"name":"2020 IEEE International Conference on Prognostics and Health Management (ICPHM)","volume":"191 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-12-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133712229","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：