Deep Learning-Based Defect Detection and Optimization in IoRT Using Metaheuristic Techniques and the Flower Pollination Algorithm

Abstract

The DL with IoRT has improved very much in the detection of automated defects in industrial settings. Classical approaches rely on manual feature extraction and computationally expensive preprocessing, which limits real-time applications. The proposed optimized deep learning framework is improved by the Flower Pollination Algorithm for hyperparameter tuning. Utilizing CNNs along with IoRT-enabled real-time monitoring, the system achieved a better accuracy (95%), precision (92%), and recall (94%). Comparing the model with the existing metaheuristic models shows that it converges faster, provides fewer false alarms, and requires less computational overhead, thus it is best suited for smart manufacturing applications.