NB-IoT Application on Decision Support System of Building Information Management

Internet of Things (IoT) is a popular system architecture for monitoring application such as building, industrial or environment. IoT system produces amount of data that is difficult for operator to process. Decision support system is an information that assists the system administrator to decide a decision when facing a problem. Moreover, the common wireless communication technology to build the IoT system is Wi-Fi, ZigBee and Bluetooth that have weakness in the coverage area. The weak signals were usually found when implement in smart building application. In this research, we applied Narrow Band Internet of Things (NB-IoT) to create a building information management system as well as used Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) to build the decision support system for building information management. The proposed system was applied in a building which has basement, first floor, and second floor. Each floor was installed end node which consist of sensors, esp-32 and M3510 (NB). Those are three kinds of nodes function in our proposes system, (1) nodes for building, (2) nodes for equipment, and (3) nodes for human activity. The sensors array for node building are placed on windows, doors and glass wall. The human activity nodes recorded from sensor on front door, Passive Infrared sensors and sensor on back door. For equipment management, sensors were placed to monitor pump and water level. The Decision System in this research was built by using the SVM and KNN. Both of SVM and KNN analyzed and decided the decisions based on data from end node. Based on experiment, the proposed NB-IoT design was able to solve the coverage area problem by replacing the Wi-Fi, ZigBee and Bluetooth. The sensor measurements were perfectly transmitted through NB-IoT and completely recorded in server. The proposed system was work perfectly to monitor, record and classify the normal and abnormal condition when received the alert information from conventional monitoring system. The accuracy of proposed SVM and KNN methods are 96.9% and 94.1%, respectively. The SVM performance is higher than KNN.