Abstract:Industrial Multivariate Time Series (MTS) is a critical view of the industrial field for people to understand the state of machines. However, due to data collection difficulty and privacy concerns, available data for building industrial intelligence and industrial large models is far from sufficient. Therefore, industrial time series data generation is of great importance. Existing research usually applies Generative Adversarial Networks (GANs) to generate MTS. However, GANs suffer from unstable training process due to the joint training of the generator and discriminator. This paper proposes a temporal-augmented conditional adaptive diffusion model, termed Diff-MTS, for MTS generation. It aims to better handle the complex temporal dependencies and dynamics of MTS data. Specifically, a conditional Adaptive Maximum-Mean Discrepancy (Ada-MMD) method has been proposed for the controlled generation of MTS, which does not require a classifier to control the generation. It improves the condition consistency of the diffusion model. Moreover, a Temporal Decomposition Reconstruction UNet (TDR-UNet) is established to capture complex temporal patterns and further improve the quality of the synthetic time series. Comprehensive experiments on the C-MAPSS and FEMTO datasets demonstrate that the proposed Diff-MTS performs substantially better in terms of diversity, fidelity, and utility compared with GAN-based methods. These results show that Diff-MTS facilitates the generation of industrial data, contributing to intelligent maintenance and the construction of industrial large models.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to solve several key challenges in the generation of industrial multivariate time - series (MTS) data to support intelligent maintenance and the construction of large - scale industrial models. Specifically, the paper mainly focuses on the following problems: 1. **Data shortage and privacy issues**: - Industrial multivariate time - series data is crucial for understanding machine states. However, due to difficulties in data collection and privacy protection, the amount of data available for building industrial intelligence and large - scale models is far from sufficient. - Data shortage seriously hinders intelligent maintenance research and the construction of large - scale industrial models. 2. **Limitations of existing generation methods**: - Existing generative adversarial networks (GANs) face the problem of unstable training when generating multivariate time - series, resulting in low - quality generated data. - The joint training of the generator and discriminator in GANs easily leads to non - convergence and an unstable training process, especially in cases of high sampling frequencies and strong noise. - Other generative models such as variational auto - encoders (VAEs) perform poorly in generating realistic samples because they rely on reconstruction loss functions. 3. **Conditional consistency issues**: - Multivariate time - series data has different conditions (such as fault categories, health indicators), and it is difficult for generative models to generate data consistent with specific conditions. - Some methods achieve control generation by introducing classifiers or discriminators, but this requires additional network structures for joint training, reducing conditional consistency. 4. **Complex time - dependencies**: - MTS data is usually not smooth and involves complex temporal dependency relationships, which makes it difficult to generate high - quality time - series data. - The time - series at the current moment is related to the time - series at previous moments. This relationship is non - stationary and difficult to predict, causing generative models to have difficulty in extracting trend information. To solve these problems, the paper proposes an improved method based on the diffusion model - Diff - MTS (Temporal - Augmented Conditional Diffusion - based AIGC for Industrial Time Series). This method improves the quality and stability of generating multivariate time - series by introducing conditional adaptability and time - decomposition - reconstruction mechanisms. Specific contributions include: - Proposing an enhanced temporal - conditional diffusion model, combining conditional adaptability and time - decomposition - reconstruction, which solves the shortcomings of traditional diffusion models in generating MTS with complex temporal dependencies. - Introducing a classifier - free conditional adaptive maximum mean discrepancy (Ada - MMD) diffusion method, which enhances conditional consistency. - Designing a time - decomposition - reconstruction U - Net (TDR - UNet) for denoising and restoring MTS data, which improves the fidelity of the generated data. Through these improvements, the experimental results of Diff - MTS on multiple datasets show that it is significantly superior to GAN - based methods in terms of diversity, fidelity, and practicality, providing a new solution for industrial data generation.

Diff-MTS: Temporal-Augmented Conditional Diffusion-based AIGC for Industrial Time Series Towards the Large Model Era

Diff-MTS: Temporal-Augmented Conditional Diffusion-Based AIGC for Industrial Time Series Toward the Large Model Era

A Novel Time Diffusion Model for Industrial Time Series Data Generation

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

TriD-MAE: A Generic Pre-trained Model for Multivariate Time Series with Missing Values

TS-Diffusion: Generating Highly Complex Time Series with Diffusion Models

Auto-Regressive Moving Diffusion Models for Time Series Forecasting

MG-TSD: Multi-Granularity Time Series Diffusion Models with Guided Learning Process

MTS-CycleGAN: An Adversarial-based Deep Mapping Learning Network for Multivariate Time Series Domain Adaptation Applied to the Ironmaking Industry

Temporal Disentangled Contrastive Diffusion Model for Spatiotemporal Imputation

Score-CDM: Score-Weighted Convolutional Diffusion Model for Multivariate Time Series Imputation

Robust Unsupervised Anomaly Detection Via Multi-Time Scale DCGANs with Forgetting Mechanism for Industrial Multivariate Time Series

A Novel Encoder-Decoder Model for Multivariate Time Series Forecasting

DiffsFormer: A Diffusion Transformer on Stock Factor Augmentation

Deep Temporal Deaggregation: Large-Scale Spatio-Temporal Generative Models

Time-Varying Gaussian Markov Random Fields Learning for Multivariate Time Series Clustering

Anomaly Detection Model for Large-scale Industrial Systems Using Transfer Entropy and Graph Attention Network

MV-Diffusion: Motion-aware Video Diffusion Model

AFSTGCN: Prediction for multivariate time series using an adaptive fused spatial-temporal graph convolutional network

Diff-MGR: Dynamic Causal Graph Attention and Pattern Reproduction Guided Diffusion Model for Multivariate Time Series Probabilistic Forecasting

Robust Multivariate Time Series Forecasting against Intra- and Inter-Series Transitional Shift