Page 69 - IJAMD-2-2

International Journal of AI for
Materials and Design
A unified industrial AI foundation framework

practices for model evaluation and optimization strategies machinery. We utilized a publicly available gearbox
for efficient inference. As a result, maintenance engineers vibration dataset provided by the Prognostics and Health
can ask context-aware questions (e.g., “What should I Management Society. The dataset consists of 2,016 labeled
69
check if vibration levels exceed the threshold in a [specific] samples collected under 78 different operating conditions,
machine?”), and the ILKM-powered QA system provides covering seven distinct fault classes, including various gear
answers grounded in both structured knowledge and damage levels (class 1, 2, 3, 4, 6, and 8) and healthy state
validated historical data. (class 0). Each sample comprises time-domain vibration
signals recorded over 3 – 12 s at a sampling frequency of
5. Case study 20,480 Hz. The problem addressed in this study is evaluating
In this section, we demonstrate the application of the the multi-class classification performance of the model
proposed industrial AI foundation framework through using this dataset under various operating conditions.
a case study on intelligent fault diagnosis for rotating Figure 3 illustrates the workflow of applying the industrial

Figure 3. Case study of applying the industrial artificial intelligence (AI) foundation framework for AI-driven rotating machinery fault diagnosis
Abbreviations: CNN: Convolutional neural network; LLM: Large language model; LSTM: Long short-term memory; ML: Machine learning; RQ: Research
question.

Volume 2 Issue 2 (2025) 63 doi: 10.36922/IJAMD025080006