Particle clogging in the artificial groundwater recharge process is one of the main factors influencing the artificial groundwater recharge efficiency, and particle deposition is the microscopic mechanism of the occurrence and development of particle clogging. Particle deposition in porous media changes the pore structure. The computed tomography (CT) scanning technique is a nondestructive testing method and determines the spatial distribution of pores in porous media. This study combines physical and CT scanning experiments to identify the change process of the pore structure in the artificial groundwater recharge process and compares the pore changes during recharge experiments between two columns containing different media. Porous media changes are observed with the CT scanning technique. The fractal theory is applied in the analysis of CT scan images and physical experiment results. The results of this study indicate that particle deposition can be examined by using CT scan images to obtain pore-related parameters, the internal pore structure of porous media determined through CT scan images can be applied in numerical simulation, and a mathematical model for particle deposition calculation in porous media is established. Compared to the physical experiment measurements, the spatial particle deposition information acquired with the CT scanning technique exhibits a higher accuracy and contains much more relevant data. Not only does this research reveal more clearly the particle clogging mechanism which is based on particle deposition, but also characterize, simulate and predict more accurately the development tendency of particle clogging during artificial groundwater recharge.
3D printing offers great potential for developing complex flexure mechanisms. Recently, thickness-correction factors (TCFs) were introduced to correct the thickness and stiffness deviations of powder-based metal 3D printed flexure hinges during design and analysis. However, the reasons for the different TCFs obtained in each study are not clear, resulting in a limited value of these TCFs for future design and fabrication. Herein, the influence of the porous layer of 3D printed flexure hinges on the hinge thickness is investigated. Samples of parallelogram flexure mechanisms (PFMs) were 3D printed using selective laser melting (SLM) and 316L stainless steel powder. A 3D manufacturing error analysis was completed for each PFM sample via 3D scanning, surface roughness measurement and morphological observation. The thickness of the porous layer of the flexure hinge was independent of the designed hinge thickness and remained close to the average powder particle diameter. The effective hinge thickness could be estimated by subtracting twice the value of the porous layer thickness from the designed value. Guidelines based on finite element analysis and stiffness experiments are proposed. The limitations of the presented method for evaluating the effective hinge thickness of flexure hinges 3D printed via SLM are also discussed.
This paper presents a short term load forecasting model based on Bayesian neural network (shorted as BNN) learned by the Hybrid Monte Carlo (shorted as HMC) algorithm. The weight vector parameter of the Bayesian neural network is a multi-dimensional random variable. In learning process, the Bayesian neural network is considered as a special Hamiltonian dynamical system, and the weights vector as the system position variable. The HMC algorithm is used to learn the weight vector parameter with respect to Normal prior distribution and Cauchy prior distribution, respectively. The Bayesian neural networks learned by Laplace algorithm and HMC algorithm and the artificial neural network (ANN) learned by the BP algorithm were used to forecast the hourly load of 25 days of April (Spring), August (Summer), October (Autumn) and January (Winter), respectively. The roots mean squared error (RMSE) and the mean absolute percent errors (MAPE) were used to measured the forecasting performance. The experimental result shows that the BNNs learned by HMC algorithm have far better performance than the BNN learned by Laplace algorithm and the neural network learned BP algorithm and the BNN learned by HMC has powerful generalizing capability, it can welly solve the overfitting problem. 相似文献