An Artificial Neural Network-Based Quality Control Framework for Nigerian Manufacturing Industries

• Musibaudeen O. IDRIS

  Author

• Busayo S. ADEBOYE

  Author

• Rasheedah A. AYOOLA

  Author

• Abiodun G. ABIOYE

  Author

• Abideen T. OYEWO

  Author

Manufacturing industries in developing economies face persistent challenges of inconsistent product quality, high defect rates, and the limited responsiveness of conventional inspection systems. This study develops and validates an Artificial Neural Network (ANN)-based quality control framework for Nigerian manufacturing industries. Three ANN implementations were examined across sectorally different companies: a supervised Multi-Layer Perceptron (MLP) binary classifier for Diaper Company and a Pharmaceutical Manufacturing Company, and an unsupervised Autoencoder for Paint Company where no historical quality labels existed. Models were evaluated using accuracy, precision, recall, F1-score, ROC-AUC, and Matthews Correlation Coefficient (MCC). The pharmaceutical ANN achieved a ROC-AUC of 0.9098 and 92.5% accuracy. In the diaper case, the meta-classifier achieved perfect performance on the held-out study sample when highly informative sub-system diagnostic flags were incorporated as inputs, indicating the value of integrated machine signals rather than implying raw-sensor generalisation. The paint Autoencoder detected an anomalous batch with a reconstruction error 13.73 times the normal mean without any labelled training data. These results were synthesised into a five-stage, multi-modal quality control framework covering data infrastructure assessment, preprocessing, architecture selection, threshold calibration, and deployment. Overall, the study shows that ANN-based quality control is technically viable and practically accessible for Nigerian manufacturers using open-source tools, modest datasets, and context-appropriate modelling pathways.

• FJET_21_90_90

Copyright (c) 2026 FUDMA Journal of Engineering and Technology

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.