In this study, a mixed flow closed cooling tower was tested with a control variable method. The factors affecting outlet water temperature were screened with a grey correlation analysis method. Five most important factors were taken as input features in a gray_BP neural network which was developed to predict outlet water temperature of the mixed flow closed cooling tower. These factors include inlet water temperature, wet bulb temperature, water refill temperature, circulating water flow rate and air volume, and the prediction output is outlet water temperature. The network adopts a three-layer structure, four-hidden layer neurons, and 30 000 iterations. Experimental data that involve no training set were used to validate the developed model. It shows that the gray neural network model outperforms the traditional BP neural network model. The correlation coefficient, average relative error, root mean square error are 0.998 9, 0.293 4% and 0.152 9, respectively. We concluded that the gray_BP neural network is a promising algorithm for predicting water temperature of a mixed flow closed cooling tower.

To explore motion characteristics of indoor respirable particles, Brownian force on particles is considered in motion probability model. Motion characteristics of 0.01 μm, 0.1 μm and 1 μm particles under conditions of up supply with up return and up supply with side return air forms were simulated with lattice Boltzmann method taking Re as 400, 1 000 and 2 000, respectively. It shows that range of particle spacial distribution increases with Re, and smaller particles affected more obvious by air turbulence and diffusion effect. Mean square displacement (MSD) of particles is inversely proportional to Re and diameter of particles. At same Re, MSD of particles is greater under up supply with side return air form, thus lower suspended particles and higher indoor air quality appears in up supply with side return air form.