Design and Implementation of Industrial Internet of Things Based Smart Factory

Abstract

With the emanation of Industry 4.0 doors to new research fields are open and many industries are moving towards Industry 4.0. Smart manufacturing is a part of Industry 4.0 that connects the factory assets with the digital world and responds in real-time to meet the changes in the manufacturing process. This structure is adopted to improve efficiency and increase productivity and growth rate. This research has been divided into two phases. The first phase consists of the design and implementation of Industrial Internet of Things (IIoT), and the second phase is to develop the controller to maintain the pH quality of the manufacturing industry. In the first phase of this research, an IIoT-based smart factory is designed in which all sensors, actuators, and controllers are connected through communication and sensor layer. The idea behind a smart factory is to manufacture the products for the consumer and stakeholders at the right time. A user interface is also designed to connect the consumers and stakeholders with the industry. Asset and inventory is also monitored with the help of this interface. This smart industry is designed according to the Industry 4.0 standard called Reference Architecture Model of Industrie 4.0 (RAMI4.0). That is a fully integrated, collaborative system to meet the changing demand and customer needs. The second phase is to control and monitor the pH process to produce quality products. The control of pH is very important in the different manufacturing industries. The pH control is considered a benchmark problem due to its complexities, nonlinearities, and frequently changing dynamics. In the chemical industry, pH controlling is very important to produce quality products. The pH controlling process is based on the chemical reactions that occur in Continuous Stirred Tank Reactor (CSTR). In this research, a stable system with dead-time is considered for the pH controlling process. An intelligent rule-based FUZZY logic controller and, conventional PID controller is designed using the Ziegler Nichols tuning method. These controllers are designed to improve the quality of the product. Intelligent machine learning-based controllers are also developed to improve the quality. Performance analysis of tested models shows that material cost for FUZZY logic was 2.96% less than PID controller and 6.22% less than the time-based model. These models reduced the production time by 45 min. Elman, Layer Recurrent, and Feed-Forwardrward neural networks are designed to improve product quality. During the comparison of the trained models, it is found that the performance of Layer Recurrent NN is much better than other methods with a mean square error of 4.8401, and a standard deviation of 1.6049. The performance analysis of the trained model shows that the material cost of the FFNN model was 2.49%, with an error of 2.7%, while the material cost of the RNN model was 1.76% with an error of 2%. The computational time of the RNN model to predict the dosing quantity is 835.6 msec, while FFNN takes 709.5 msec for the prediction of dosing quantity