https://hdl.handle.net/10259/4393 Fri, 01 May 2026 12:25:17 GMT 2026-05-01T12:25:17Z https://hdl.handle.net/10259/11528 Identification of Simultaneous Soft Faults in Analog Circuits Using a Hybrid PSO-Machine Learning Approach Dieste Velasco, Mª Isabel Analog circuits are fundamental to a wide range of industrial systems, where their evaluation is essential for ensuring operational reliability and preventing system failures. However, diagnostic methodologies for analog circuits are markedly less developed than those for their digital counterparts, primarily due to the inherent difficulty of detecting soft faults within analog environments. One particularly challenging category of faults involves simultaneous degradations across multiple components that do not result in a hard failure of the circuit. Indeed, there is a notable lack of studies addressing the detection of simultaneous soft faults in analog circuits. This study proposes a method for identifying this type of soft fault occurrence in analog circuits by combining Machine Learning (ML) techniques, specifically Random Forests and Artificial Neural Networks, with an Evolutionary Algorithm (EA) based on Particle Swarm Optimization (PSO). The proposed approach is validated on a second-order Sallen-Key band-pass filter, a circuit in which soft fault classification is particularly challenging. Furthermore, the study highlights the performance improvements achieved through the proposed combined method in detecting and classifying simultaneous soft faults. This study demonstrates that an iterative process combining ML and EA techniques enables accurate fault prediction in electronic circuits. Moreover, the integration of these strategies can enhance the performance of classification problems that are traditionally addressed using either ML or EA in isolation. The effectiveness of the proposed method is evaluated using several statistical metrics, including the Matthews Correlation Coefficient (MCC), F1-score, and others. Sun, 01 Mar 2026 00:00:00 GMT https://hdl.handle.net/10259/11528 2026-03-01T00:00:00Z https://hdl.handle.net/10259/11193 Enhanced Monte Carlo-Based Uncertainty Quantification in Electronic Circuits Dieste Velasco, Mª Isabel The determination of measurement uncertainty is of paramount importance in defining the range of values assigned to a given response variable. In electronic circuits, this is a challenging task that still needs further development. Difficulties associated with repeated measurements and the inherent variability of components make it difficult to determine uncertainty unless a large number of measurements are taken, which increases costs. Furthermore, due to the strong nonlinear characteristics of most electronic circuits and the difficulty in determining a differentiable expression, it is not feasible to apply uncertainty propagation. An alternative for modeling uncertainty is the Monte Carlo method, as proposed by the Guide for the Expression of Measurement Uncertainty (GUM and GUM-S1). However, when the probability distribution of the response variables significantly deviates from normal, the GUM-S1 method may lead to inaccurate results. This study proposes a new Monte Carlo-based method to quantify measurement uncertainty in electronic circuits and includes a comparative study of GUM, GUM-S1, and the proposed method. It is shown that uncertainty can be determined more accurately with the proposed method when normality cannot be assumed. The proposed method is applied to two analog circuits: a Sallen-Key filter and a low-signal amplifier. Wed, 01 Oct 2025 00:00:00 GMT https://hdl.handle.net/10259/11193 2025-10-01T00:00:00Z https://hdl.handle.net/10259/11067 The effects of sky diffuse light on indoor illuminance through radiosity models: a case study in Burgos Granados López, Diego; García Fuente, Manuel; González Peña, David; García, Ignacio; Alonso Tristán, Cristina Diffuse radiation can play a critical role in the design of sustainable urban environments, in so far as it can transmit natural light to areas that direct sunlight cannot reach because of buildings and other structures. This characteristic of sky luminance is crucial for radiosity-based methods where luminance is used to determine energy transfer between surfaces. Consequently, the accuracy of a radiosity-based model will depend upon how well it can capture the subtle variations of sky luminance. In this study, both the accuracy and the performance of three luminance models are evaluated: the All-Weather model, the All-Sky model, and the CIE Standard General Sky model, focusing on their capability to replicate luminance at any point in the sky and at any given time. The results showed that while the CIE Standard Sky model offered the highest accuracy, it required more complex input data. The All-Weather and the All-Sky models rely on radiometric measurements. Both produced reliable results, with the All-Weather model standing out, because of its efficiency and minimal data requirements. Despite those strong points, all the models demonstrated higher error rates near the horizon, due to the challenges of accurately modeling luminance in this region. In this study, two radiosity methods were compared for calculating indoor illuminance: the Simplified Radiosity Algorithm (SRA), which considers spatial luminance variations across the openings, and the DeLight method, which assumes a uniform luminance distribution throughout the window view. The analysis of the results showed that the error rates produced in the luminance pattern estimations were reflected in the Radiosity model. Taking that effect into account, the combination of the All-Sky model with the SRA algorithm demonstrated a strong balance between accuracy and resource efficiency, offering a practical approach for sustainable urban lighting design. Thu, 01 Jan 2026 00:00:00 GMT https://hdl.handle.net/10259/11067 2026-01-01T00:00:00Z https://hdl.handle.net/10259/11066 Ultraviolet Erythemal Irradiance (UVER) under Different Sky Conditions in Burgos, Spain: Multilinear Regression and Artificial Neural Network Models García Rodríguez, Sol; García Rodríguez, Ana; Granados López, Diego; García, Ignacio; Alonso Tristán, Cristina Different strategies for modeling Global Horizontal UltraViolet Erythemal irradiance (G⁡H⁡U⁢V⁢E) based on meteorological parameters measured in Burgos (Spain) have been developed. The experimental campaign ran from September 2020 to June 2022. The selection of relevant variables for modeling was based on Pearson’s correlation coefficient. Multilinear Regression Model (M⁢L⁢R) and artificial neural network (A⁢N⁢N) techniques were employed to model G⁡H⁡U⁢V⁢E under different sky conditions (all skies, overcast, intermediate, and clear skies), classified according to the C⁢I⁢E standard on a 10 min basis. A⁢N⁢N models of G⁡H⁡U⁢V⁢E outperform those based on MLR according to the traditional statistical indices used in this study (R2, M⁢B⁢E, and n⁢R⁢M⁢S⁢E). Moreover, the work proposes a simple all-sky A⁢N⁢N model of G⁡H⁡U⁢V⁢E based on usually recorded variables at ground meteorological stations. Sun, 01 Oct 2023 00:00:00 GMT https://hdl.handle.net/10259/11066 2023-10-01T00:00:00Z