A fundamental requirement for IIoT is fault detection, especially in industrial applications. Fault detection is defined as detecting the fault occurrence or identifying the root causes of the system out of control status. The fault diagnosis includes two processes that interpret the current status of plant sensors and process the observed knowledge to the server. The fault diagnosis is a primary process of automation-based industrial IoT application systems. Reliability and security are the two main concerns of fault diagnosis. Previously, several works have been developed by focusing n resource utilization and workload allocation. The fault-detection towards IIoT is still immature stage. The industrial control equipment and their parameters are changed dynamically. Moreover, it is impractical to customize the learning algorithm specific to each parameter. Learning algorithms combined with dynamic voting weight have been used to implement fault-type prediction over IIoT.