Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AIGC for Industrial Time Series: From Deep Generative Models to Large Generative Models

Lei Ren, Haiteng Wang, Jinwang Li, Yang Tang, Chunhua Yang

TL;DR

Industrial time-series data are often scarce, noisy, and privacy-sensitive, hindering AI-driven insights. The authors review DGMs and LGMs for industrial time series, propose a DGM-based AIGC framework, and outline a four-part roadmap: dataset, architecture, self-supervision, and fine-tuning. They provide a unified evaluation benchmark for fidelity, diversity, and utility, and present a case study on engine predictive maintenance demonstrating improved performance with diffusion-based data generation. The work offers practical guidance for deploying generative models in industry and highlights open challenges and future directions.

Abstract

With the remarkable success of generative models like ChatGPT, Artificial Intelligence Generated Content (AIGC) is undergoing explosive development. Not limited to text and images, generative models can generate industrial time series data, addressing challenges such as the difficulty of data collection and data annotation. Due to their outstanding generation ability, they have been widely used in Internet of Things, metaverse, and cyber-physical-social systems to enhance the efficiency of industrial production. In this paper, we present a comprehensive overview of generative models for industrial time series from deep generative models (DGMs) to large generative models (LGMs). First, a DGM-based AIGC framework is proposed for industrial time series generation. Within this framework, we survey advanced industrial DGMs and present a multi-perspective categorization. Furthermore, we systematically analyze the critical technologies required to construct industrial LGMs from four aspects: large-scale industrial dataset, LGMs architecture for complex industrial characteristics, self-supervised training for industrial time series, and fine-tuning of industrial downstream tasks. Finally, we conclude the challenges and future directions to enable the development of generative models in industry.

AIGC for Industrial Time Series: From Deep Generative Models to Large Generative Models

TL;DR

Industrial time-series data are often scarce, noisy, and privacy-sensitive, hindering AI-driven insights. The authors review DGMs and LGMs for industrial time series, propose a DGM-based AIGC framework, and outline a four-part roadmap: dataset, architecture, self-supervision, and fine-tuning. They provide a unified evaluation benchmark for fidelity, diversity, and utility, and present a case study on engine predictive maintenance demonstrating improved performance with diffusion-based data generation. The work offers practical guidance for deploying generative models in industry and highlights open challenges and future directions.

Abstract

With the remarkable success of generative models like ChatGPT, Artificial Intelligence Generated Content (AIGC) is undergoing explosive development. Not limited to text and images, generative models can generate industrial time series data, addressing challenges such as the difficulty of data collection and data annotation. Due to their outstanding generation ability, they have been widely used in Internet of Things, metaverse, and cyber-physical-social systems to enhance the efficiency of industrial production. In this paper, we present a comprehensive overview of generative models for industrial time series from deep generative models (DGMs) to large generative models (LGMs). First, a DGM-based AIGC framework is proposed for industrial time series generation. Within this framework, we survey advanced industrial DGMs and present a multi-perspective categorization. Furthermore, we systematically analyze the critical technologies required to construct industrial LGMs from four aspects: large-scale industrial dataset, LGMs architecture for complex industrial characteristics, self-supervised training for industrial time series, and fine-tuning of industrial downstream tasks. Finally, we conclude the challenges and future directions to enable the development of generative models in industry.
Paper Structure (48 sections, 8 equations, 7 figures, 3 tables)

This paper contains 48 sections, 8 equations, 7 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Deep Generative Models
  3. Autoregressive Model
  4. Variational Autoencoder (VAE)
  5. Generative Adversarial Network (GAN)
  6. Diffusion Model
  7. The definition, characteristic, and demand of industrial time series generation
  8. The Definition of Industrial Time Series Generation
  9. The Characteristic of Industrial Time Series Generation
  10. Challenges in Industrial Time Series Acquisition
  11. The proposed DGM-based AIGC framework and the current state-of-the-art DGMs
  12. Limited Sample Augmentation
  13. Imbalanced Data Synthesis
  14. Sensor Signal Imputation
  15. Sensor Signal Denoising
  16. ...and 33 more sections

Figures (7)

  • Figure 1: The comparison between DGMs and LGMs.
  • Figure 2: The comparison of mainstream DGMs and their introduction.
  • Figure 3: The DGM-baed AIGC framework of industrial time series generation. First, the generative model learns the latent representations of industrial time series by modeling dynamic processes and correlations between variables. Second, five specific generative tasks are executed in a downstream network. Finally, the generated samples will be applied to various industrial scenarios to address challenges such as sample scarcity in industrial scenarios.
  • Figure 4: From DGMs to LGMs, the research methods of generative models have undergone profound changes. The roadmap for building large generative models from four aspects: Large industrial dataset, LGMs architecture for complex industrial characteristics, self-supervised training for industrial time series, fine-tuning for industrial downstream tasks.
  • Figure 5: Workflow of implementing diffusion model in this case study
  • ...and 2 more figures