|
|
Train Identification System at Railway Switches And Crossings Using Advanced Machine Learning Methods
Krč, Rostislav ; Vorel,, Jan (referee) ; Plášek, Otto (referee) ; Podroužek, Jan (advisor)
This doctoral thesis elaborates possibilities of automatic train type identification in railway S&C using accelerometer data. Current state-of-the-art was considered, including requirements stated by research projects such as S-Code, In2Track or Turnout 4.0. Conducted experiments considered different architectures of artificial neural networks (ANN) and statistically evaluated multiple use case scenarios. The resulting accuracy reached up to 89.2% for convolutional neural network (CNN), which was selected as a suitable baseline architecture for further experiments. High generalization capability was observed as models trained on data from one location were able to classify locomotive types in the other location. Further experiments evaluated the effect of signal filtering and denoising. Evaluation of allocated memory and processing time for pre-trained models proved feasibility for in-situ application with regard to hardware restrictions. Due to a limited amount of available accelerometer data, distribution grid power demand data were utilized for further refinement of the proposed CNN architecture. Deep multi-layer architecture with regularization techniques such as dropout or batch normalization provides state-of-the-art performance for time series classification problems. Class activation mapping (CAM) allowed an explanation of decisions made by the neural network. Presented results proved that train type identification directly in the S&C is possible. The CNN was selected as optimal architecture for this task due to high classification accuracy, automatic filtration, and pattern recognition capabilities, allowing for the incorporation of the end-to-end learning strategy. Moreover, direct on-site application of pre-trained models is feasible with respect to limitations of in-situ hardware. This thesis contributes to understanding the train type identification problem and provides a solid theoretical background for future research.
|
|
|
Trully Smart Smart Socket
Valušek, Ondřej ; Zemčík, Pavel (referee) ; Materna, Zdeněk (advisor)
There is a large selection of so called smart sockets available on the market today. The possibilities of these sockets are sadly very limited. Typically, they can measure power consumption, be turned off and on remotely by mobile application and timer. This thesis deals with this problem by showing how a smart relay can be used to create a truly smart smart socket that can classify currently connected appliances using just short time window for up to three devices combined. The power consumption is measured using Shelly 1PM together for three plugs. Using time series feature extraction, unknown device detection with SVM and neural network classification, the accuracy was over 99%. on a dataset containing combinations of smart TV, lamp and a laptop consumption. Information about currently connected devices is displayed on a webpage and written to a database to be viewed later. The information about connecting and disconnecting a device can be further sent to a system for smart home management.
|
|
|
Trully Smart Smart Socket
Valušek, Ondřej ; Zemčík, Pavel (referee) ; Materna, Zdeněk (advisor)
There is a large selection of so called smart sockets available on the market today. The possibilities of these sockets are sadly very limited. Typically, they can measure power consumption, be turned off and on remotely by mobile application and timer. This thesis deals with this problem by showing how a smart relay can be used to create a truly smart smart socket that can classify currently connected appliances using just short time window for up to three devices combined. The power consumption is measured using Shelly 1PM together for three plugs. Using time series feature extraction, unknown device detection with SVM and neural network classification, the accuracy was over 99%. on a dataset containing combinations of smart TV, lamp and a laptop consumption. Information about currently connected devices is displayed on a webpage and written to a database to be viewed later. The information about connecting and disconnecting a device can be further sent to a system for smart home management.
|
|
|
Train Identification System at Railway Switches And Crossings Using Advanced Machine Learning Methods
Krč, Rostislav ; Vorel,, Jan (referee) ; Plášek, Otto (referee) ; Podroužek, Jan (advisor)
This doctoral thesis elaborates possibilities of automatic train type identification in railway S&C using accelerometer data. Current state-of-the-art was considered, including requirements stated by research projects such as S-Code, In2Track or Turnout 4.0. Conducted experiments considered different architectures of artificial neural networks (ANN) and statistically evaluated multiple use case scenarios. The resulting accuracy reached up to 89.2% for convolutional neural network (CNN), which was selected as a suitable baseline architecture for further experiments. High generalization capability was observed as models trained on data from one location were able to classify locomotive types in the other location. Further experiments evaluated the effect of signal filtering and denoising. Evaluation of allocated memory and processing time for pre-trained models proved feasibility for in-situ application with regard to hardware restrictions. Due to a limited amount of available accelerometer data, distribution grid power demand data were utilized for further refinement of the proposed CNN architecture. Deep multi-layer architecture with regularization techniques such as dropout or batch normalization provides state-of-the-art performance for time series classification problems. Class activation mapping (CAM) allowed an explanation of decisions made by the neural network. Presented results proved that train type identification directly in the S&C is possible. The CNN was selected as optimal architecture for this task due to high classification accuracy, automatic filtration, and pattern recognition capabilities, allowing for the incorporation of the end-to-end learning strategy. Moreover, direct on-site application of pre-trained models is feasible with respect to limitations of in-situ hardware. This thesis contributes to understanding the train type identification problem and provides a solid theoretical background for future research.
|
guest ::
