Intelligent Real-Time Production Scheduling Using Machine Learning Algorithms

Thesis Abstract

In the dynamic landscape of modern manufacturing, efficient production scheduling remains a critical challenge due to increasing complexity, variability in demand, and constraints in resource availability. Traditional scheduling methods often lack adaptability and real-time responsiveness, leading to suboptimal resource utilization, delays, and increased operational costs. This study aims to develop and evaluate an intelligent, real-time production scheduling system that leverages machine learning algorithms to enhance decision-making processes in manufacturing environments. Specifically, the research seeks to design a machine learning-driven scheduling framework, assess its performance against conventional methods, and identify key factors influencing its efficacy. The study adopts a mixed-methods research design, integrating quantitative analysis through statistical modeling with qualitative insights from industry practitioners. The target population comprises manufacturing firms in the automotive component sector within the region, with a sampling frame drawn from 50 companies identified through industry directories. A stratified random sampling technique selected a sample of twenty manufacturing plants, each equipped with computer-controlled machinery and existing automated scheduling tools, to ensure representative variability in operational scales and technological maturity. Data collection involved a combination of structured interviews, to gather qualitative perspectives on current scheduling practices, and the extraction of operational data logs over a six-month period for quantitative analysis. These logs included information on job sequences, processing times, machine states, and production disruptions. To develop the machine learning model, the study employed supervised learning techniques, notably random forest classifiers and support vector machines (SVM), trained on historical operational data. Feature selection focused on variables such as job priority levels, machine utilization rates, maintenance schedules, and order deadlines. The model's predictive performance was validated using cross-validation and metrics such as accuracy, precision, recall, and F1-score. An optimization component was integrated using reinforcement learning algorithms, specifically Q-learning, to adapt scheduling decisions dynamically in response to real-time data, disturbances, and unforeseen events. The analytical framework was underpinned by the Theory of Constraints and the Scheduling Theory, providing conceptual grounding for prioritization and resource allocation strategies. Expected findings anticipate that the machine learning-based scheduling system will significantly improve key performance indicators, including a reduction in makespan by an average of 15%, decreased machine idle time by 20%, and enhanced on-time delivery rates by 25% compared to traditional scheduling approaches. The system’s ability to adapt to real-time disruptions and learn from ongoing data streams is expected to outperform static heuristic methods, thereby increasing manufacturing agility and responsiveness. The study also aims to identify critical data features influencing scheduling efficacy and to explore the integration challenges faced by firms during system implementation. This research makes a substantial contribution to the field of industrial and production engineering by demonstrating the practical viability of intelligent, data-driven scheduling frameworks in manufacturing operations. It bridges the gap between theoretical machine learning models and their real-world application, providing a blueprint for industry adoption. The findings will inform managers and engineers on leveraging ICT solutions to automate and optimize scheduling processes dynamically, thereby improving operational efficiency. The study concludes with recommendations for developing scalable, user-friendly decision-support tools, and suggests further research into multi-objective optimization and the incorporation of IoT-enabled sensor data. Overall, the research affirms that integrating machine learning algorithms within real-time scheduling systems offers a transformative approach to tackling the complexities of modern manufacturing environments. It underscores the importance of continuous data collection and model refinement, advocating for industry-wide deployment of intelligent scheduling solutions as a means to achieve higher productivity, better resource management, and improved competitive advantage in manufacturing sectors.