Particle Flow Detection and Forecasting Solution
(REF : S-0593)
- Our solution provides indoor air flow direction detection and forecasting, in horizontal axis for all areas, and also in vertical axis for closed areas like hospital wards. Deep Learning (DL) on IoT sensor data and Computational Fluid Dynamics (CFD) simulation will be used.
- Additional sensors (air-pressure, particle sensors like PM1, PM2.5, PM10, temperature, humidity, etc.) can be installed on existing lighting sensor boxes for measuring the pressure gradient/contour across the area. This information is to calculate air flow and pollutants flow directions in the horizontal axis.
- Closed area air flows are often complicated and environmental turbulence severely degrades ventilation efficiency. To detect the air flow rate within hospital wards and measure the effectiveness of the air exchanger, 3D CFD simulation of indoor air dynamics will be conducted in the areas.
- Deep learning on simulation, together with the pressure, particle, etc. sensors data will also be employed in the experiments, enabling the learned DL model to predict the air flows in 3D from sensor data.
- With the deep learning on the CFD simulations, our solution can produce better detection and forecasting on the air flow direction, particle concentration (which can be further translated into aerosol concentrations), air exchange rate, and even extend to evaluate probability levels of infections.
|Trial Application and Expected Outcome
- Air pressure sensors and particle sensors will be installed at ceiling level and evenly distributed across short distances. Computing on the air pressures from these sensors will give the air pressure gradients and contours, which is further translated into air flow directions in the horizontal axis.
- For closed areas, CFD will be conducted on some chosen typical room setups regarding the inlet/outlet locations, inlet air directions and velocities, bed locations, etc.. Air pressure and particle sensors at the ceiling level, eye level and bed level will be installed from a sensor grid.
- Deep learning on data collected from the closed area experimental environment will be carried out to predict airflows and particle concentrations from various room setup parameters. External environmental parameters will also be used as input parameters. Temperature, humidity and particle concentrations will provide primary benchmarks.
- When used in actual environment, the deep learning models will provide forecasting on the airflows and particle concentrations, with room setup and external environmental parameters taken as input.
- Combining the horizontal airflow directions given by the air pressure contour and airflow vector chart, and the vertical airflow directions given by the DL models, our solution can measure and visualize the indoor air flow, and the probability levels of infections (if expert labelling are available).
|Additional Solution Information
EIP Overview with Focus on Contamination Analysis.pdf
|Info on I&T Solution Provider