No 1 (2024)

Introduction. When receiving telemetry data from a mobile high-speed object (MHSO), numerous challenges arise due to the dynamic trajectory of the MHSO, a wide range of flight altitudes, spatial movement of transmitting antenna radiation patterns, and the multi-beam nature of radio wave propagation. These factors result in periodic deep and prolonged signal fading received by recording equipment. In such conditions, diversity reception using multiple reception points (RP) becomes necessary. Analytically assessing the state of such a communication channel is nearly impossible, given the dynamic nature of the MHSO and the various positioning options for MHSO and RP antennas. The goal of this study is to develop and implement a software-based model for transmitting telemetric data from the MHSO board, enabling the calculation of data reception quality characteristics such as bit error probability, throughput, and channel reliability.

The telemetry data transmission and reception system model comprises a signal propagation environment model and signal processing algorithms in the transmitter and receivers. The propagation environment model is primarily based on well-established analytical expressions describing radio wave propagation considering the influence of underlying surfaces, supplemented partly by simulation techniques to calculate the power of the signal component scattered over a large area of the underlying surface. Simulation modeling is also employed to implement signal processing algorithms.

Results. A simulation model of a telemetric communication system has been developed, capturing the main physical processes in the "transmitter – propagation medium – receivers – joint digital signal processing" system. Test simulation results obtained in a specific model scenario align well with the physical processes of radio wave propagation, transmission, and reception by antenna systems, as well as diversity reception theory.

Conclusion. The practical utility of the developed model lies in its ability to predict signal reception quality under given external conditions and to preliminarily determine the required configuration of transmitting antennas, installation locations of receiving points, and other telemetry system parameters ensuring the desired level of reception quality and reliability. The scientific significance of the study includes obtaining a new tool for analyzing complex dynamic radio channels in telemetry systems and demonstrating the possibility of continuous high-quality reception of telemetric data from the MHSO board during flight along a descending trajectory with constant spatial position changes of the two transmitting antennas.

Introduction. In modern digital control systems, ensuring the reliability of ADCs is topical. Self-diagnosis algorithms are commonly employed to detect and address failures, thereby enhancing reliability. This research aims at developing a novel approach by harnessing the capabilities of a local fragmented control device (LFCD) to identify failures in the main measuring neuron (MMN) system, followed by the exclusion of the failed MMN from the neural network.

Materials and Methods. The study applied self-diagnosis algorithms to identify failed MMNs for two neural network structures: the "Internal Feedback Structure" and the "Redundant Link Structure." Graphical interpretations of the operation sequence are provided for cases of complete uncertainty, where the state of all neurons from the base group is unknown, and for cases of the unknown state of one neuron. The concept of the base group is introduced as the minimum number of neurons required for self-diagnosis.

Results and Conclusion. In the MatLab Simulink environment, we developed a model to compare neural network structures and self-diagnosis algorithms. We utilized this model to investigate the algorithm complexity and total time required for neural network analysis based on the number of tested neurons. Our findings demonstrated that for the "Internal Feedback Structure," the base group consists of 2m MMNs, where m represents the ADC resolution during self-diagnosis, while for the "Redundant Link Structure," it is 2m+1. The analysis highlighted that the "Redundant Link Structure" and selecting parameter m=3 represent the most optimal solution, offering shorter verification time and requiring less hardware resources while maintaining other characteristics.

Practical Significance. Research findings will enable subsequent self-diagnosis of control system elements and developing a diagnostic algorithm that ensures parallel checking in different areas of neural networks and the process of analog-to-digital conversion on the free part.

Introduction. The main problem of neural network models in the form of embedded systems is the limited computing resources, which does not allow obtaining the output result in the shortest possible time. Minimizing the time spent on model training also remains a priority. Therefore, when comparing the accuracy of output data and the learning speed of neural network models, the main attention should be paid to different image sizes. The goal of the work is to identify a model that gives the most accurate results after training, has high performance and allows you to produce the most extensive output data with minimal input. The novelty of the study lies in the fact that it was carried out using a comparative analysis of five neural network models on one framework with input data sizes of 32 × 32,48 × 48 and 64 × 64, the combination of which has not been studied.

Description of research objects. Five of the most common models were selected as objects of comparative analysis: DenseNet, GoogLeNet, LeNet, MobileNet and ResNet50. The study uses the available TSRD dataset, which includes 6164 images of road signs containing 58 categories. The images were combined into two sub-databases, one of which is training, and the second is test. There are 4170 image files in the training database, and 1994 in the testing database. Each image contains an annotation with information about the four coordinates of the sign and its category. Their size varies from 26 × 28 to 491 × 402 pixels. Tensor-Flow framework was also used to run all neural network models. The advantage of the models is the accuracy of the output data with a relatively small number of inputs.

Conclusion. According to the results of the study, the following leaders in output accuracy can be identified: ResNet50 and MobileNet. In terms of learning speed, MobileNet takes first place with an average of 169 seconds on the GPU, ResNet50 takes second place with a difference of 465 seconds in terms of learning speed on GPU. MobileNet data accuracy showed high results of 94.5% on CPU and 94.5% on GPU. In terms of accuracy and training time, the LeNet and GoogleNet models showed similar performance and accuracy. The DenseNet model showed the most negative results in its accuracy and performance. Changing the size of the input data had little effect on the output accuracy and scalability of the training environment. Due to the fact that the data sizes may not affect the accuracy of the result, when using the TSRD data set, it is proposed to use 32 × 32 dimensions as input data sizes that allow you to find road signs. This solution is more economical and does not affect the accuracy of the output data. The experiments conducted during the study showed that the accuracy of ResNet50 is similar to the accuracy of MobilNet. This proves that for the definition of road signs, deeper models with a longer learning time are not in every case preferable to small ones.

Introduction. Monitoring water levels at discrete points in reservoirs is crucial for various industries such as nuclear fuel and energy, chemical, and agriculture. Fiber-optic sensors (FOS) have become essential for remote water level sensing over the past twenty-five years. However, the drawback of quasi-distributed multiplexed FOS networks lies in the use of expensive multiplexing technologies, typically wave technology, as well as interrogation. These technologies require costly and temperature-sensitive components like wideband light sources for the entire C-band and ordered waveguide gratings, for instance, fiber Bragg gratings with different central wavelengths. This study aims at addressing these challenges by designing a passive multiplexed quasi-distributed fiber-optic sensor network for water level monitoring in reservoirs. This network offers improved metrological and functional characteristics while enabling inexpensive address multiplexing and interrogation of FOSs based on addressable fiber Bragg structures (AFBS) using radio-photonic technologies.

Methods. Fresnel-type water pumps generate reflected radiation whose power varies depending on its location in water or air. Each FOS channel contains an AFBS with a common central wavelength and a unique address frequency formed by two symmetrical transparency windows with different spacings, achieved by introducing two phase π-shifts. The output of the channel combiner connects to a photodetector via a fiber-optic backbone, and the information is sent to a radio-photonic interrogator. This interrogator allocates the power of the received signal at a specific address frequency, allowing determination of whether the FOS is in water or air. The address of the FOS is determined by the AFBS's address frequency in its measuring channel.

Conclusion. A simple, fully optical quasi-distributed sensor network for monitoring water levels at discrete points, based on the address multiplexing method, is proposed. This system was demonstrated for monitoring a supply tank for preparing reagent solutions for wastewater treatment plants. Simulation and experimental studies confirmed the feasibility of this system. Statistical analysis of data obtained during a twenty-day test cycle showed a relative error in level measurement of 0.3% and an absolute measurement error of 2 mm. These results are comparable to standard continuous electronic level gauges operating on pressure and microwave radars.