Hyperparameter on-line learning of stochastic resonance based threshold networks

Aiming at training the feed-forward threshold neural network consisting of nondifferentiable activation functions, the approach of noise injection forms a stochastic resonance based threshold network that can be optimized by various gradient-based optimizers. The introduction of injected noise extends the noise level into the parameter space of the designed threshold network, but leads to a highly non-convex optimization landscape of the loss function. Thus, the hyperparameter on-line learning procedure with respective to network weights and noise levels becomes of challenge. It is shown that the Adam optimizer, as an adaptive variant of stochastic gradient descent, manifests its superior learning ability in training the stochastic resonance based threshold network effectively. Experimental results demonstrate the significant improvement of performance of the designed threshold network trained by the Adam optimizer for function approximation and image classification.     

An efficient edge based data structure for the compressible Reynolds-averaged Navier–Stokes equations on hybrid unstructured meshes

An efficient edge based data structure has been developed in order to implement an unstructured vertex based finite volume algorithm for the Reynolds-averaged Navier–Stokes equations on hybrid meshes. In the present approach, the data structure is tailored to meet the requirements of the vertex based algorithm by considering data access patterns and cache efficiency. The required data are packed and allocated in a way that they are close to each other in the physical memory. Therefore, the proposed data structure increases cache performance and improves computation time. As a result, the explicit flow solver indicates a significant speed up compared to other open-source solvers in terms of CPU time. A fully implicit version has also been implemented based on the PETSc library in order to improve the robustness of the algorithm. The resulting algebraic equations due to the compressible Navier–Stokes and the one equation Spalart–Allmaras turbulence equations are solved in a monolithic manner using the restricted additive Schwarz preconditioner combined with the FGMRES Krylov subspace algorithm. In order to further improve the computational accuracy, the multiscale metric based anisotropic mesh refinement library PyAMG is used for mesh adaptation. The numerical algorithm is validated for the classical benchmark problems such as the transonic turbulent flow around a supercritical RAE2822 airfoil and DLR-F6 wing-body-nacelle-pylon configuration. The efficiency of the data structure is demonstrated by achieving up to an order of magnitude speed up in CPU times.     

Optimization of Transducer Location for Novel Non-Intrusive Methodologies of Diagnosis in Diesel Engines

The health monitoring has been studied to ensure integrity of design of engine structure by detection, quantification, and prediction of damages. Early detection of faults may allow the downtime of maintenance to be rescheduled, thus preventing sudden shutdown of machines. In cylinder pressure developed, vibrations and noise emissions data provide a rich source of information about condition of engines. Monitoring of vibrations and noise emissions are novel non-intrusive methodologies for which positioning of various transducers are important issue. The presented work shows applicability of these diagnosis methodologies adopted in case of diesel engines. The effects of changing various fuel injection parameters was analyzed. Scope of using non-intrusive technique has been analyzed by changing locations of microphone. Novelty of this worklies in exploring signal processing methods for various locations around the engine test set up. Various frequency ranges of contributing noise and vibration sources were identified. Time-Frequency analysis showed the onset of various cyclic. Based on the identification of various frequency bands, it is possible to device suitable filters in order to extract more information.     

Prediction of the drag reduction effect of pulsating pipe flow based on machine learning

Prediction of drag reduction effect caused by pulsating pipe flows is examined using machine learning. First, a large set of flow field data is obtained experimentally by measuring turbulent pipe flows with various pulsation patterns. Consequently, more than 7000 waveforms are applied, obtaining a maximum drag reduction rate and maximum energy saving rate of 38.6% and 31.4%, respectively. The results indicate that the pulsating flow effect can be characterized by the pulsation period and pressure gradient during acceleration and deceleration. Subsequently, two machine learning models are tested to predict the drag reduction rate. The results confirm that the machine learning model developed for predicting the time variation of the flow velocity and differential pressure with respect to the pump voltage can accurately predict the nonlinearity of pressure gradients. Therefore, using this model, the drag reduction effect can be estimated with high accuracy.     

Assessment of Noise Exposure of Sawmill Workers in Southwest,Nigeria

Economic wood processing employs the use of industrial machines for cutting, shaping, milling, and sawing timber, thereby leading to the generation of high levels of noise. Published data from empirical studies have categorized noise as an environmental hazard of global significance. Furthermore, noise exposure limits for different industries and all the industrial machines available has not been formally established as it presently exists in developed nations around the world. Therefore, this study assessed the daily exposure of sawmills workers to noise in Southwestern Nigeria. Reconnaissance surveys were first carried out in Osun, Oyo, Ondo, Ekiti, Lagos, and Ogun States to select sawmills that were fully operational and fit for the study. Two fully functional sawmills in two cities of each State were eventually selected for data collection, making a total of 24 sawmills, while the Circular Machines (CM), Planer Machines (PM), and Band-saw Machines (BM) were the machines in each sawmill considered. Two machines each of CM, PM, and BM were considered in each sawmill, making a total of forty-eight (48) machines each of CM, PM, and BM. Sound data were collected between 7 am and 7 pm each day for six days (between Monday and Saturday) using Extech 407732 sound level meter and all stabilized measurements were taken three times at different intervals. The data collected were in three different periods: Machine No-work Period (NP_m), Machine Idle Period (IP_m), and Machine Working Period (WP_m). A two–way Analysis of Variance (ANOVA) was carried out at P < 0.05 to determine whether there is a significant difference in the sound level average before and after the break, for both the idle and working periods of the three machines considered. This was also done to determine whether there is a significant difference between the sound level average of the results collected during idle and working periods of the three machines. Noise Pollution Levels (L_np) ranged from 83.20 dB (PM) to 107.65 (BM) and 93.42 (CM and PM) – 116.00 (BM) respectively, while IP_m also gave the least noise pollution level of 95.79 dB and WP_m gave the highest level of 102.88 dB. The results revealed that all the machines’ L_np values in the working period are more than the 90 dB acceptable limit the recommendation value of 90 dB while 89.6% of CMs, 75% of PMs, and 89.6% of BM had their L_np above 90 dB in the idle period respectively. The minimum and the maximum noise dose levels for IP_m, WP_m and overall are 0.09 (BM) and 2.37 (CM), 0.50 (CM), and 4.77 (PM) and 0.69 (BM) and 6.64 (PM) respectively. The study found out that the fundamental contributing factors to the high noise levels in sawmills are poor machine maintenance, use of old and obsolete machines, poor housekeeping strategy, limited space, workers’ negligence, lack of PPE, and lack of occupational safety training. The study recommends that proper workplace practices such as use of personal protective equipment, new and modern machines, training, and occupational safety programmes be implemented in the considered sawmills.     

Fluid‐structure coupling of linear elastic model with compressible flow models

Cavitation erosion is caused in solids exposed to strong pressure waves developing in an adjacent fluid field. The knowledge of the transient distribution of stresses in the solid is important to understand the cause of damaging by comparisons with breaking points of the material. The modeling of this problem requires the coupling of the models for the fluid and the solid. For this purpose, we use a strategy based on the solution of coupled Riemann problems that has been originally developed for the coupling of 2 fluids. This concept is exemplified for the coupling of a linear elastic structure with an ideal gas. The coupling procedure relies on the solution of a nonlinear equation. Existence and uniqueness of the solution is proven. The coupling conditions are validated by means of quasi‐1D problems for which an explicit solution can be determined. For a more realistic scenario, a 2D application is considered where in a compressible single fluid, a hot gas bubble at low pressure collapses in a cold gas at high pressure near an adjacent structure.