Theses/Dissertations/Projects

Permanent URI for this collection

https://ir.bellsuniversity.edu.ng/handle/123456789/52

Browse

Browsing Theses/Dissertations/Projects by Subject "Fit Regression"

Now showing 1 - 1 of 1
  • No Thumbnail Available
    ItemMetadata only
    A Comparative Analysis of Mathematical and Machine Learning Based Models for Predictive Maintenance in a Production Line.
    (Bells University of Technology, Ota., 2022) AINA, Yusuf Amuda; Dr. Mrs. Olamide O. Olusanya
    Maintenance is an activity to ensure that production line equipment are in satisfactory condition and reliable. Thus, the optimum goal of maintenance is to ensure that the performance of the equipment is satisfactory by minimizing downtime and reducing cost related to downtime. The complexity of production line has changed the tradition decision- making process regarding maintenance management through the4 use of artificial intelligent based predictive maintenance in order to make more efficient real-life decision. This study developed a mathematical and machine learning predictive maintenance models using production line parameters to predict machine failure. Fit retrogression model was used as the mathematical model and adaptive neuro fuzzy inference system (ANFIS based) was used as the adopted machine learning model. Data obtained from the production line revealed that the fault is subject to production line parameters such as main pressure, operating voltage, current consumption, vibration and temperature. Performing statistical analysis revealed the adequacy of the models with coefficient of determination (R2) of 83.99% for number of faults. Training of datasets with ANFIS revealed a convergence at epoch number of 3, and with minimum root mean square error (RMSE) of approximately zero (0.000161), indicating the significant training approach of the ANFIS training of the dataset. Evaluation of the two models (mathematical and machine learning) showed they give better prediction of the number of faults, as their predicting values were closer to that obtained from the equipment. Comparatively, machine learning (ANFIS based model) showed better predicting strength compared to the mathematical model. This is because ANFIS adopts the learning and performing system of artificial neural network (ANN) and fuzzy logic.