The industrial internet of things (IIoT) is disrupting industries through greater automation, efficiency and connectivity. But they also make IIoT systems vulnerable to advanced cyber threats like denial-of-service (DoS) attacks and infiltration, necessitating strong security measures to protect these platforms. This paper proposes a cognition computing based intrusion detection system (IDS) for IIoT security. The proposed IDS has the ability to not only detect known threats, but also emerging threats, by incorporating ensemble learning models, together with cognitive principles to achieve computational efficiency. To improve the performance of detection, the system uses feature selection methods such as Recursive Feature Elimination (RFE) and Mutual Information (MI) to select only a refined input data and remove noise. Results: The IDS evaluated on the UNSW-NB15-v2 dataset shows very competitive results compared to the state-of-the-art in terms of accuracy, recall, and ROC-AUC. Ensemble methods including XGBoost and Random Forest surpassed classical methods, with accuracy as high as 98% and a marked decrease in false alarms. We believe that supporting cognitive computing and ensemble learning techniques can provide over IIoT consistent, scalable, and adaptive security solutions. Results illustrate how this method can tackle the distinct security issues of IIoT, leading to more intelligent and robust industrial practices.