Acta Polytechnica

The classification of eye diseases from fundus images based on CNN and pretrained models

2023-11-28T11:46:50+01:00

Visual impairment affects more than a billion people worldwide due to insufficient care or inadequate vision screening. Computer-aided diagnosis using deep neural networks is a promising approach, it can analyse and process retinal fundus images, providing valuable reference data for doctors in clinical diagnosis or screening. This study aims to achieve an accurate classification of fundus images, including images of healthy patients as well as those with diabetic retinopathy, cataracts, and glaucoma, using a convolutional neural network (CNN) architecture and several pretrained models (AlexNet, GoogleNet, ResNet18, ResNet50, YOLOv3, and VGG 19). To enhance the training process, a mirror effect technique was applied to augment the volume of data. The experimental study resulted in very satisfactory outcomes, with the GoogleNet model paired with the SGDM optimiser achieving the highest accuracy (92.7 %).

Reduction of torque pulsation in ultralight helicopter transmission

2024-01-09T13:46:53+01:00

Usually, 1–2 seats helicopters use piston engines. One of their disadvantages is a significant torque pulsation on the engine shaft. The main reason for such fluctuations is the change in parameters during the engine’s operating cycle. The purpose of the investigation was to define the pulsation on the helicopter rotor shaft. An additional elastic clutch was used on a single-seat ultralight helicopter Rotorschmiede VA-115 of German firm RS Helikopter GmbH for pulsation dampening. Several flights were performed to assess the pulsation on the shafts. The results of the tests showed that the torque pulsation on the shaft in all modes does not exceed 30 %. This is 10 times less than the pulsation value of 300% specified by the airworthiness standards for this helicopter class.

On the vortex identification within a linear blade cascade - an experimental research

2023-12-18T12:27:07+01:00

This paper deals with the identification of the individual vortical structures within a linear blade cascade based on H criteria. The experimental data obtained from the pressure measurement at the cascade outlet were evaluated by a standard procedure to obtain the velocity field in one plane. Then, the vorticity in the flow field was evaluated based on Crocco’s theorem, and in the final step, the helicity density was calculated. The impact of the inlet flow angle on the velocity field as well as on the vortical structures was investigated. The effect of the value of the H threshold on the vortex identification is discussed in the last part of the paper.

Assessing the efficacy of flat-plate solar collectors using nanofluids in the climatic context of Kirkuk city, Iraq

2023-11-14T09:02:35+01:00

Solar energy is a key renewable energy source. Research and development have focused on enhancing the heat transfer coefficient, heat gain, and practical efficiency of solar systems. The aim of this study is to evaluate the performance of a flat solar panel collector using a nanofluid under conditions in the city of Kirkuk/Iraq, 35° latitude and 45° longitude, in terms of practical calculation of thermal efficiency. The study included making two solar collectors, one traditional and the other improved using a nanofluid (CuO). The CuO/Water nanofluid was prepared with a volumetric concentration of 0.25 % by mechanical mixing and then ultrasonic mixing to homogenise the particles and eliminate the agglomerations that form inside the fluid. Practical testing was conducted for the two solar collectors, one using distilled water and the other using the nanofluid, during four months (January, February, March, and April) of the year 2023. The experiments revealed that the efficiency progressively improves from 9:00 a.m. to 1:00 p.m. This increase is attributed to solar radiation’s decreasing intensity post 12:00 p.m., while thermal storage and minimised thermal losses continue to contribute. After 2:00 p.m., the efficiency dwindles due to the declining solar radiation intensity. The practical efficiency of a 0.25 % nanofluid (CuO) attains its zenith at a mass flow rate of 0.015 ls⁻¹. Higher mass flow rates enhance heat transfer within fluid-filled tubes. The collector efficiency at this flow rate ranges from 31.66 % in January to 44.44 % in April.

Problems with reinforced concrete industrial floors with regard to subsoil swelling

2023-11-28T11:15:02+01:00

Most of the problems associated with open cracks in reinforced concrete industrial floors do not arise from technological indiscipline in the execution or exceeding the permitted floor load, but from the geotechnical profile beneath the floor. In the presence of swelling soil in the subsoil, the floors can then be shifted upwards by centimeters and create open cracks. This article describes regression relationships for the prediction of swelling pressure and deformation of reinforced concrete industrial floors based on indirect measurements. These relationships were obtained by evaluating a large database of measurements carried out by the company GeoTec-GS and the Czech Technical University in Prague using neural networks, multiple correlation, regression analysis, and sensitivity analysis. The article also presents the actual classification of the risk of surface damage of reinforced concrete floors due to swelling of the subsoil and an example of its application is given.

Outdoor performance investigation of a thermoelectric cooler-integrated solar air heating collector

2023-11-28T11:53:03+01:00

Solar air heating systems are continuously being improved by combining them with other energy conversion technologies. In this paper, outdoor tests were carried out on a thermoelectric heat-pumping solar air heater (TE-SAH), with four (04) TECs (TEC1–12706) attached to the backside of the absorber plate, and powered by a 40Wp mono-Si PV module to pump heat from the absorber plate into the heated air. The thermal energy production, energy efficiency, heat loss coefficients, and heat removal factor were evaluated and compared with a reference system without TECs. The heat collection and energy efficiency of the solar air heater were improved by 7.14 % and 66.71 %, respectively, with the integration of TECs. Heat losses also decreased by 0.46 MJ. Furthermore, the estimated heat removal factor for the TE-SAH was 0.55, higher than 0.49 obtained for the reference SAH. These results showed that PV-TE heat-pumping is a viable means of improving the thermal performance of solar air heaters.

Reverse engineering of pump as turbine for CFD analysis

2023-11-28T11:00:45+01:00

Reverse engineering (RE) using 3D scanning is already a relatively technologically simple and cost-effective method. For water turbines, this is particularly true for RE of larger machines. With microturbines, there is a lot of pressure to minimise costs, even at the cost of reduced accuracy. Using an existing micro-PAT (Pump as Turbine) as an example, we showed the approach to assessing these microturbines, starting with scanning the entire internal flow profile of the turbine, reconstructing the surface into a 3D model, and numerically assessing it using CFD (Computational Fluid Dynamics). A different approach is necessary compared to standard large machines, whose dimensions allow troublefree scanning of the flow parts of the turbine. Using CFD, we assessed the reconstructed geometry of the PAT. Two significant findings were made: the importance of high-quality 3D scanning by combining several cheaper 3D scanners and the necessity for reliable in-situ measurements for a successful CFD validation. Our future focus involves optimising PAT runner geometry in turbine mode to enhance energy production at the site and, at the same time, eliminating existing cavitation.

Digital model of automated unmanned aerial vehicle with edge computing application for railway reconnaissance

2023-12-31T08:41:20+01:00

In the last decade, unmanned aerial vehicles have been widely used in various civil and military field operations. Most modern drones are manually controlled by an operator and require decision making, which requires experience gained through training, which is time consuming and expensive, and even with experienced operators, the human factor can never be excluded. In this paper, an algorithm for an automatic flight mode of unmanned aerial vehicles without an operator’s participation has been developed. The proposed algorithm has been considered in the military reconnaissance of railways using edge computing and computer vision approaches to recognise railroads and moving trains for simulated and real cases as well as send report data to the web platform. As a result, the head of the operation received information about the number of carriages and Global Positioning System coordinates of the recognised train to make the necessary decisions.