In an aerospace manufacturing project, engineers are preparing data to train an AI system for predictive maintenance. They need to transform the data from multiple sensors and ensure it is consistent and accurate before building the model. What should the project manager do to handle the inconsistencies?

The question addresses the problem of "inconsistencies" within an existing dataset prepared for an AI model. The most direct and fundamental action is to identify and reconcile these conflicting data points. This is a core task within the data cleaning and preparation phase of any AI project. Inconsistencies, such as contradictory readings from different sensors at the same timestamp, must be resolved to ensure the data's integrity. Training a model on inconsistent data would lead it to learn incorrect patterns, severely compromising the reliability of the predictive maintenance system, which is unacceptable in a critical domain like aerospace.

A. Enhancing data with new sources is data enrichment; it adds context but does not resolve internal conflicts within the original dataset.
B. Data augmentation creates synthetic data to increase the training set's size, which is useful for preventing overfitting but does not fix inconsistencies.
C. A validation protocol is a forward-looking governance measure to ensure the quality of future data, not a method for cleaning an existing dataset.

1. Project Management Institute (PMI). (2024). AI in Project Management: A Practical Guide. Section 3.4, "Data Preparation," emphasizes that this phase involves cleaning the data, which includes tasks like handling missing values, correcting errors, and resolving inconsistencies to create a high-quality dataset suitable for model training.

2. García, S., Luengo, J., & Herrera, F. (2015). Data Preprocessing in Data Mining. Springer International Publishing. Chapter 2, "Data Cleaning," explicitly details the process of resolving inconsistencies. It states, "Inconsistencies may arise... for instance, due to data entry errors... The typical solution is to correct these values manually using external references or automatic rules" (p. 28). DOI: https://doi.org/10.1007/978-3-319-10247-4

3. Carvalho, T. P., et al. (2019). A systematic literature review of machine learning methods applied to predictive maintenance. Computers & Industrial Engineering, 137, 106024. This review highlights that data preprocessing is a critical stage in predictive maintenance projects. The paper discusses the necessity of "data cleaning" to handle issues like noise and inconsistencies from sensor data to ensure the robustness of the final model (Section 4.1, Data Preprocessing). DOI: https://doi.org/10.1016/j.cie.2019.106024