Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evaluation of Artificial Intelligence Methods for Lead Time Prediction in Non-Cycled Areas of Automotive Production

Cornelius Hake, Jonas Weigele, Frederik Reichert, Christian Friedrich

TL;DR

This study evaluates AI approaches for predicting unknown lead times in a non-cycled automotive production area (TFC), framing lead time as the interval between the sixth and eighth reporting points. Through a structured data pipeline with one-hot encoding and a focus on supervised classification, ensemble methods (LightGBM, XGBoost, CatBoost) outperform regression for this non-linear, exponential distribution, with LightGBM chosen as the final model. Experiments across data subsets show up to 90% accuracy for coarse classings (2–4 classes) and demonstrate substantial improvements over a rule-based baseline, while also highlighting the need for periodic retraining to handle production changes. The work emphasizes practical value in production planning and control, and points to future work in explainable AI and automated retraining strategies to maintain high prediction quality in dynamic manufacturing environments.

Abstract

The present study examines the effectiveness of applying Artificial Intelligence methods in an automotive production environment to predict unknown lead times in a non-cycle-controlled production area. Data structures are analyzed to identify contextual features and then preprocessed using one-hot encoding. Methods selection focuses on supervised machine learning techniques. In supervised learning methods, regression and classification methods are evaluated. Continuous regression based on target size distribution is not feasible. Classification methods analysis shows that Ensemble Learning and Support Vector Machines are the most suitable. Preliminary study results indicate that gradient boosting algorithms LightGBM, XGBoost, and CatBoost yield the best results. After further testing and extensive hyperparameter optimization, the final method choice is the LightGBM algorithm. Depending on feature availability and prediction interval granularity, relative prediction accuracies of up to 90% can be achieved. Further tests highlight the importance of periodic retraining of AI models to accurately represent complex production processes using the database. The research demonstrates that AI methods can be effectively applied to highly variable production data, adding business value by providing an additional metric for various control tasks while outperforming current non AI-based systems.

Evaluation of Artificial Intelligence Methods for Lead Time Prediction in Non-Cycled Areas of Automotive Production

TL;DR

This study evaluates AI approaches for predicting unknown lead times in a non-cycled automotive production area (TFC), framing lead time as the interval between the sixth and eighth reporting points. Through a structured data pipeline with one-hot encoding and a focus on supervised classification, ensemble methods (LightGBM, XGBoost, CatBoost) outperform regression for this non-linear, exponential distribution, with LightGBM chosen as the final model. Experiments across data subsets show up to 90% accuracy for coarse classings (2–4 classes) and demonstrate substantial improvements over a rule-based baseline, while also highlighting the need for periodic retraining to handle production changes. The work emphasizes practical value in production planning and control, and points to future work in explainable AI and automated retraining strategies to maintain high prediction quality in dynamic manufacturing environments.

Abstract

The present study examines the effectiveness of applying Artificial Intelligence methods in an automotive production environment to predict unknown lead times in a non-cycle-controlled production area. Data structures are analyzed to identify contextual features and then preprocessed using one-hot encoding. Methods selection focuses on supervised machine learning techniques. In supervised learning methods, regression and classification methods are evaluated. Continuous regression based on target size distribution is not feasible. Classification methods analysis shows that Ensemble Learning and Support Vector Machines are the most suitable. Preliminary study results indicate that gradient boosting algorithms LightGBM, XGBoost, and CatBoost yield the best results. After further testing and extensive hyperparameter optimization, the final method choice is the LightGBM algorithm. Depending on feature availability and prediction interval granularity, relative prediction accuracies of up to 90% can be achieved. Further tests highlight the importance of periodic retraining of AI models to accurately represent complex production processes using the database. The research demonstrates that AI methods can be effectively applied to highly variable production data, adding business value by providing an additional metric for various control tasks while outperforming current non AI-based systems.
Paper Structure (11 sections, 4 figures, 5 tables)

This paper contains 11 sections, 4 figures, 5 tables.

Table of Contents

  1. INTRODUCTION AND LITERATURE REVIEW
  2. METHODS AND MATERIALS
  3. Experimental Setup
  4. Obtaining and Preprocessing the Training Data
  5. Selection Process and Methodology for AI Model Evaluation
  6. Labeling the Data
  7. Training Procedure for AI Methods
  8. RESULTS
  9. Training Results
  10. Evaluation of the Prediction Quality
  11. CONCLUSION AND FUTURE WORK

Figures (4)

  • Figure 1: Illustration of the data pipeline; dashed line: test, continuous line: series application
  • Figure 2: TFC process with explanation of data availability and lead time calculation
  • Figure 3: Results of the prediction accuracy (bins) and training times (lines) in the preliminary experiment
  • Figure 4: Results from the comparison with the actual, rule-based system