Deep learning and dempster shafer theory based decision-making in the industrial internet of things

One of the main goals of IIoT is to collect data from industrial environments to monitor and automatically detect events and unexpected changes and anomalies in industrial environments. Because the data collected in industrial environments is voluminous, it is needed to use efficient machine learning methods such as deep learning (DL) to analyze this voluminous data. According to the development of sensor technology, different sensor sources can be used by IIoT to obtain proper inference, which requires combining the resulting decisions to get a single and reliable decision. One of the most used methods for the fusion of many decisions is Dempster Shafer's (DS) combining theory. Therefore, two methods are proposed in this thesis to make decisions about IIoT anomalies. The first uses artificial neural networks (ANN) and support vector machine (SVM) for classification and DS theory to combine decisions. This method is suitable for small and medium-sized industrial units with relatively small data. The second proposed method uses one-dimensional convolutional neural networks (1DCNNs) for information classification and DS combination theory to combine decisions. This method is suitable for large industrial units due to the massive data processing power of deep learning networks.