The Industrial Internet of Things (IIoT) is an essential factor behind the digital transformation of conventional industries by enabling the connection of sensors, instruments, and other industrial devices to the Internet for efficient data gathering and analysis. This connectivity facilitates greater financial gains, enhanced productivity, and improved operational performance. However, it poses some challenges, such as network latency, sensor sample latency, and dependability. To tackle these issues, a novel Belgua Optimized deep leARning framework for fault detection in IIoT (BOARD-IIoT) has been proposed in this paper for real time monitoring and fault prediction. The proposed BOARD-IIoT method utilizing the Sparse Principal Component Analysis (SPCA) for feature extraction which efficiently capturing the most relevant features from complex data. Beluga Whale Optimization (BWO) to select the most significant features that enhance the classification accuracy. The proposed system employs the Heterogeneous Neural Network (HNN) for classification. The efficacy of the suggested technique has been evaluated using specific metrics including F1score (F1S), accuracy, precision (PR) and recall (RC). The accuracy of the BOARD-IIoT approach in the machine failure dataset is 0.8%, 1.2% and 0.4% higher than existing AE-AnoWGAN, GA-Att-LSTM and DyEdgeGAT techniques respectively.