Sensor Fault Detection Data
Help to limit the consequences of failures in automation systems
@kaggle.arashnic_sensor_fault_detection_data
About this Dataset
Sensor Fault Detection Data
Context
The current industrial trend regarding automatisms and regarding industrial plants leads us towards systems more and more complex mechatronics, working in an uncertain, evolutionary environment. It is so necessary to develop a diagnosis module to detect a fault (Fault Detection) that may affect the operation of these systems and to locate their causes (Fault Isolation). Therefore, a diagnosis module is needed to improve the performance and productivity of systems and limit the consequences of failures that can be catastrophic on human goods and life.
Content
Time series of measurements on sensors uniquely identified by a Sensor Id. During the serie of measurement the sensor is disconnected or on failure.
This Dataset is aimed to serve the Fault detection Analytic component.
- SensorID = 1 = PT100 temperature sensor, in an industrial environment, with dust & vibrations
Themes : Plant, Building, Power, IT, Machine
Keywords : Operational, Meters & sensors, Ambient, Electrical, per minute, over months
Modified: March 7, 2018
Acknowledgements
The Publisher of this dataset is Schneider-Electric.
Inspiration
Diagnosis consists of detecting abnormal functioning from sensor data. These data may be noisy or corrupt due to unpredictable events. That abnormal operation may be a failure of process equipment (a sensor, actuator or a component), control system failure (due to operator error or cyber-attack of
the system), or change of environment for example resources that are lacking (unavailable operators, exhausted stocks, etc.), or change due to non-conformity product etc. After detecting abnormal functioning, the cause can then be located and identified to make decisions (corrective actions or
reconfiguration of the system). The different type of faultsin the process is illustrated
The application of IoT systems in industries creates a huge amount of data. In addition, these industrial systems have become more and more complex and it is difficult to obtain an analytical model of the system. In this context, the use of ML tools comes out obvious and logic to cope with the challenges of
diagnosis in these systems. The goal of this dataset is to apply through several methods, the application of ML techniques on fault detection and diagnosis
problems. Among the machine learning techniques(may be traditional) , there are Support Vector Machine (SVM), Artificial Neural Network (ANN), Fuzzy
Neural Network (FNN), Decision Trees (DT), Bayesian Network (BN).
It could be a great idea to apply novel Deep Learning algorithms on this dataset such as:
- Denoising stacked auto-encode and Long Short-Term Memory Network
- Self-Attentive Convolutional Neural Networks