.png){kind=icon}
.png){kind=icon}
Return
Multi-scale diffusion model for gas–solid flow mass flow rate prediction based on electrostatic sensing
.png){kind=icon}
PRE
AI
DOI：10.1016/j.measurement.2026.122860.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
•
Ring electrostatic sensor optimized by simulation for gas–solid flow sensitivity.
•
New decomposition–diffusion–reconstruction separates trend, transient, and residual.
•
Adaptive gating and cross-modal attention fuse conditions during diffusion.
•
Model achieves MSE 0.0298 and MAPE 5.43%.
Abstract
Electrostatic sensors provide a non-intrusive and low-cost approach to gas–solid flow monitoring, but their output is jointly affected by particle concentration, velocity, stochastic charging, and intermittent particle transport. Consequently, the mapping from a single electrostatic signal to solid mass flow is nonlinear and non-stationary, and conventional statistical or deterministic sequence models may inadequately represent abrupt particle-packet arrivals and multiscale fluctuations. This study proposes a temporal multi-scale conditional diffusion model for solid mass-flow forecasting from one differential ring-electrode electrostatic sensing channel. The model decomposes the historical signal into trend, transient, and residual representations and injects them into the diffusion denoiser through cross-modal attention and a diffusion-step-dependent adaptive gate. Experiments were conducted with cement ash in a DN100 pneumatic conveying rig over a mass-flow range of approximately 20–100 t/h, using five pressure levels and two conveying directions. Results show that the proposed model significantly outperforms baseline methods, achieving an MSE of 0.0298 and a MAPE of 5.43%.
Keywords:
Gas–solid mass flow rate prediction
Electrostatic sensor
Multi-scale diffusion model
Journal
IF:
5.6
Papers:
1.9W
Citations:
5.4W