In this paper, electrical discharge images are classified using AI with quantum machine learning techniques. These discharges were originated in dielectric mineral oils and were detected by a high-resolution optical sensor. The captured images were processed in a Scikit-image environment to obtain a reduced number of features or qubits for later training of quantum circuits. Two quantum binary classification models were developed and compared in the Qiskit environment for four discharge binary combinations. The first was a quantum variational model (QVM), and the second was a conventional support vector machine (SVM) with a quantum kernel model (QKM). The execution of these two models was realized on three fault-tolerant physical quantum IBM computers. The novelty of this article lies in its application to a real problem, unlike other studies that focus on simulated or theoretical data sets. In addition, a study is carried out on the impact of the number of qubits in QKM, and it is shown that increasing the number of qubits in this model significantly improves the accuracy in the classification of the four binary combinations studied. In the QVM, with two qubits, an accuracy of 92% was observed in the first discharge combination in the three quantum computers used, with a margin of error of 1% compared to the simulation obtained on classical computers.

This study presents an innovative bimodal approach for laboratory partial discharge (PD) analysis using a YOLOv8-based convolutional neural network (CNN). The main contribution consists, first, in the transformation of a conventional DDX-type electrical detector into a smart and autonomous data source. By training the CNN, a system capable of automatically reading and interpreting the data from the detector display—discharge magnitude and applied voltage—is developed, achieving an average training accuracy of 0.91 and converting a passive instrument into a digitalized and structured data source. Second, and simultaneously, an optical visualization system captures direct images of the PDs with a high-resolution camera, allowing for their morphological characterization and spatial distribution. For electrical voltages of 10, 13, and 16 kV, PDs were detected with a confidence level of up to 0.92. The fusion of quantitative information intelligently extracted from the electrical detector with qualitative characterization from optical analysis offers a more complete and robust automated diagnosis of the origin and severity of PDs.

This work establishes a large language model (LLM) specialized in the domain of thermoelectric generators (TEGs), for deployment on local hardware. Starting with the generalist JanV1-4B model, an efficient fine-tuning (FT) methodology (QLoRA) was employed, modifying only 3.18% of the total parameters of this base model. The key to the process is the use of a custom-designed dataset, which merges deep theoretical knowledge with rigorous instruction tuning to refine behavior and mitigate catastrophic forgetting. Performance evaluation, conducted using a questionnaire of increasing complexity, revealed that the FT JanV1-4B-expert-TEG modelled in this work achieves an overall accuracy of 81%, demonstrating capabilities ranging from the correct formulation of equations to critical design reasoning. This study validates the specialization of LLMs using QLoRA as an effective and accessible strategy for developing highly competent engineering support tools, eliminating dependence on large-scale computing infrastructures.

In this paper, electrical discharge images are classified using AI with quantum machine learning techniques. These discharges are originated in dielectric mineral oils and were detected by a high-resolution optical sensor. The captured images were processed in a Scikit-image environment to obtain a reduced number of features or qubits for later training of quantum circuits. Two quantum binary classification models were developed and compared in the Qiskit environment for four discharge binary combinations. The first was a quantum variational model (QVM), and the second a conventional support vector machine (SVM) with a quantum kernel model (QKM). The execution of these two models was realized on three fault-tolerant physical quantum IBM computers. In the QVM, with two qubits, an accuracy of 92% was observed in the first discharge combination in the three quantum computers used, with a margin of error of 1% compared to the simulation obtained on classical computers. For the other three discharge combinations, the test set accuracy was around 88%. In the SVM-QKM, with two qubits, the test set accuracy for the first discharge combination reached 97%, while for the other three combinations the average accuracy was 89%. In addition, in this model, increasing the number of qubits to eight showed an improvement in accuracy to around 99% for the first combination and to an average of 97% for the other three combinations, demonstrating that the model improves using more features.

This work establishes a large language model (LLM) specialized in the domain of thermoelectric generators (TEGs), for deployment on local hardware. Starting with the generalist JanV1-4B model and Qwen3-4B-Thinking-2507 models, an efficient fine-tuning (FT) methodology using quantized low-rank adaptation (QLoRA) was employed, modifying only 3.18% of the total parameters of thee base models. The key to the process is the use of a custom-designed dataset, which merges deep theoretical knowledge with rigorous instruction tuning to refine behavior and mitigate catastrophic forgetting. The dataset employed for FT contains 202 curated questions and answers (QAs), strategically balanced between domain-specific knowledge (48.5%) and instruction-tuning for response behavior (51.5%). Performance of the models was evaluated using two complementary benchmarks: a 16-question multilevel cognitive benchmark (94% accuracy) and a specialized 42-question TEG benchmark (81% accuracy), scoring responses as excellent, correct with difficulties, or incorrect, based on technical accuracy and reasoning quality. The model’s utility is demonstrated through experimental TEG design guidance, providing expert-level reasoning on thermal management strategies. This study validates the specialization of LLMs using QLoRA as an effective and accessible strategy for developing highly competent engineering support tools, eliminating dependence on large-scale computing infrastructures, achieving specialization on a consumer-grade NVIDIA RTX 2070 SUPER GPU (8 GB VRAM) in 263 s.

Partial discharges (PDs) are critical phenomena in diagnosing the insulation condition of electrical equipment using dielectric oils. Their early detection and characterization are fundamental for the prevention of catastrophic failures. This work presents an innovative bimodal approach for laboratory PD analysis through the training of a convolutional neural network (CNN) based on YOLOv8. Firstly, a conventional DDX-type PD electrical detector is enhanced by endowing it with smart capabilities. Through the CNN training, a system is then developed capable of automatically reading and interpreting data dis-played on an electrical detector screen, such as discharge magnitude, pulse count, and applied voltage. This transforms a conventional instrument into an autonomous source of digitized and structured data. The mean precision in the training was 0.91. Concurrently, an optical visualization system using a high-resolution camera is employed to capture direct images of PDs occurring in the dielectric oil. These images provide complementary qualitative and quantitative information, enabling the classification of discharge types based on their visual characteristics. For electrical voltages of 10, 13 and 16 kV, PDs were detected with confidence scores of up to 0.92. This study therefore combines quantitative information intelligently extracted from an electrical detector, with qualitative and mor-phological characterization obtained through optical analysis.