Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FD-LLM: Large Language Model for Fault Diagnosis of Machines

Hamzah A. A. M. Qaid, Bo Zhang, Dan Li, See-Kiong Ng, Wei Li

TL;DR

FD-LLM introduces a fault-diagnosis-specific large language model framework that treats fault prediction as a multi-class classification problem. It encodes time-series vibration data into text via FFT-based vectors or statistical feature summaries, supplemented by machine specifications and operation conditions in prompts, and fine-tunes open-source LLMs using LoRA. The approach achieves standout performance with Llama3 family on FFT representations, showing strong adaptability across operational conditions but limited cross-component generalization, and demonstrates that prompt-contextual information substantially boosts statistical-data performance. The work highlights the potential of LLMs for integrated textual and numerical diagnostics and suggests future directions in cross-component robustness and reasoning-enabled inference to further enhance reliability in industrial fault diagnosis.

Abstract

Large language models (LLMs) are effective at capturing complex, valuable conceptual representations from textual data for a wide range of real-world applications. However, in fields like Intelligent Fault Diagnosis (IFD), incorporating additional sensor data-such as vibration signals, temperature readings, and operational metrics-is essential but it is challenging to capture such sensor data information within traditional text corpora. This study introduces a novel IFD approach by effectively adapting LLMs to numerical data inputs for identifying various machine faults from time-series sensor data. We propose FD-LLM, an LLM framework specifically designed for fault diagnosis by formulating the training of the LLM as a multi-class classification problem. We explore two methods for encoding vibration signals: the first method uses a string-based tokenization technique to encode vibration signals into text representations, while the second extracts statistical features from both the time and frequency domains as statistical summaries of each signal. We assess the fault diagnosis capabilities of four open-sourced LLMs based on the FD-LLM framework, and evaluate the models' adaptability and generalizability under various operational conditions and machine components, namely for traditional fault diagnosis, cross-operational conditions, and cross-machine component settings. Our results show that LLMs such as Llama3 and Llama3-instruct demonstrate strong fault detection capabilities and significant adaptability across different operational conditions, outperforming state-of-the-art deep learning (DL) approaches in many cases.

FD-LLM: Large Language Model for Fault Diagnosis of Machines

TL;DR

FD-LLM introduces a fault-diagnosis-specific large language model framework that treats fault prediction as a multi-class classification problem. It encodes time-series vibration data into text via FFT-based vectors or statistical feature summaries, supplemented by machine specifications and operation conditions in prompts, and fine-tunes open-source LLMs using LoRA. The approach achieves standout performance with Llama3 family on FFT representations, showing strong adaptability across operational conditions but limited cross-component generalization, and demonstrates that prompt-contextual information substantially boosts statistical-data performance. The work highlights the potential of LLMs for integrated textual and numerical diagnostics and suggests future directions in cross-component robustness and reasoning-enabled inference to further enhance reliability in industrial fault diagnosis.

Abstract

Large language models (LLMs) are effective at capturing complex, valuable conceptual representations from textual data for a wide range of real-world applications. However, in fields like Intelligent Fault Diagnosis (IFD), incorporating additional sensor data-such as vibration signals, temperature readings, and operational metrics-is essential but it is challenging to capture such sensor data information within traditional text corpora. This study introduces a novel IFD approach by effectively adapting LLMs to numerical data inputs for identifying various machine faults from time-series sensor data. We propose FD-LLM, an LLM framework specifically designed for fault diagnosis by formulating the training of the LLM as a multi-class classification problem. We explore two methods for encoding vibration signals: the first method uses a string-based tokenization technique to encode vibration signals into text representations, while the second extracts statistical features from both the time and frequency domains as statistical summaries of each signal. We assess the fault diagnosis capabilities of four open-sourced LLMs based on the FD-LLM framework, and evaluate the models' adaptability and generalizability under various operational conditions and machine components, namely for traditional fault diagnosis, cross-operational conditions, and cross-machine component settings. Our results show that LLMs such as Llama3 and Llama3-instruct demonstrate strong fault detection capabilities and significant adaptability across different operational conditions, outperforming state-of-the-art deep learning (DL) approaches in many cases.

Paper Structure

This paper contains 22 sections, 11 equations, 2 figures, 16 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Methods
  4. Problem definition
  5. FD-LLM framework
  6. Data pre-processing
  7. Time series data encoding
  8. Instruction fine-tuning
  9. Post-processing and evaluation
  10. Experiments
  11. Dataset source and tasks
  12. Models
  13. Settings
  14. Metrics
  15. Results and discussions
  16. ...and 7 more sections

Figures (2)

  • Figure 1: An illustration of the FD-LLM framework using both FFT-processed and statistically processed data pipelines.
  • Figure 2: Left: Illustrates the FFT pre-processing steps. For illustration purposes, we use a complete vibration signal and its corresponding FFT transformation from an outer race fault of the bearings. We select to use a length of 512 for each segment, and the number of decimal places $D$ is set to 3. Right: Presents the statistically processed data, where each table contains the statistical features of segments from a certain signal collected under certain machine health state. Each feature name and its corresponding value are serialized into the input element of the prompt template for every segment in the tables.