ANN inverse analysis based on stochastic small-sample training set simulation

A new approach of inverse analysis is proposed to obtain parameters of a computational model in order to achieve the best agreement with experimental data. The inverse analysis is based on the coupling of a stochastic simulation and an artificial neural network (ANN). The identification parameters play the role of basic random variables with a scatter reflecting the physical range of potential values. A novelty of the approach is the utilization of the efficient small-sample simulation method Latin Hypercube Sampling (LHS) used for the stochastic preparation of the training set utilized in training the artificial neural network. Once the network has been trained, it represents an approximation consequently utilized to solve the key task: To provide the best possible set of model parameters for the given experimental data. The efficiency of the approach is shown using numerical examples from civil engineering computational mechanics.     

Fabrication and frequency response of a complex ultrasonic transducer for multilayer evaluation

To characterize the thickness of a corrosive fluid system's coated facilities, such as pipes, tubes, tanks, and structural members, a complex ultrasonic transducer capable of measuring multilayered parts was fabricated and the time and frequency responses were evaluated. The target transducer was constructed with two active dual elements made from tape-cast PbNb₂O₆ sheets and an additional thin-film active element made from sol–gel spin-coated Pb(Zr_0.52Ti_0.48)O₃ films. After adjusting the properties of each active element, a complex transducer was assembled after matching the impedance of each element and considering the minimal interference between the active layers. The impulse response of the assembled complex transducer shows excellent characteristics. Moreover, the assembled transducer's capability of accurately measuring the thickness assures that it can be directly applied to related industries.     

Improved resolution of the multiple inverse method by eliminating erroneous solutions

The multiple inverse method is a numerical technique designed to separate stresses from heterogeneous fault–slip data. The method is one of the resampling methods based on the pattern recognition. Plotting solutions determined from k-fault subset into the parameter space, we have clusters representing significant stresses for the dataset. This paper presents a technique to improve the resolution of stress for the method. The regularity/singularity test of the subsets taken from the fault–slip data is the key for this purpose. The resolution and accuracy of the method are improved by eliminating erroneous stresses or artifacts that were yielded by the method. The performances of the method are demonstrated with a variety of artificial datasets.     

Simulation of Strength Distribution in Ground Ceramics by Incorporating Residual Stress Effect

Strength of ground ceramics may be affected by residual stress as well as surface flaws induced by grinding. Strength prediction for ground ceramics is convenient for mechanical design of ceramic components. In this article, a numerical procedure based on fracture mechanics was proposed to estimate strength distribution of ground ceramics by considering grinding-induced residual stress. Bending strength and residual stress of ground ceramics were measured for three grinding-conditions. By comparison of simulated results with experimental ones, it was revealed that strength characteristics in experiments were well simulated by using the proposed procedure.     

Modelling of heat transfer in an infiltrated granular bed in view of the difference of phase temperatures

Physically substantiated boundary conditions for problems of heat transfer in infiltrated granular beds based on the two-temperature model which allow for the absence of interphase heat transfer on boundaries are formulated. It is shown that classical Dankwerts conditions would be applicable for gas. The problem of porous cooling at the boundary conditions of the 2nd and 3rd kind on the outer boundary is solved in a new formulation.     

Pool boiling of saturated FC-72 on nano-porous surface

Pool boiling heat transfer from nano-porous surface immersed in a saturated FC-72 dielectric fluid has been experimentally studied at atmospheric pressure (101 kPa). The data obtained from nano-porous surface (Anodisc 25) of thickness about 70 μm made from aluminum oxide (Al₂O₃) obtained from Whatman, were compared to that of a plain surface (aluminum) of thickness about 105 μm. From the experimental data obtained it was evident that there is a reduction of about 30% in the incipient superheat for the applied power for nano-porous surface over plain surface. SEM photographs of the nano-porous coating were taken for determining the size of the pores.