Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Study of energy building efficiency based on inframed inspection by unmanned aerial vehicles

Zarkada Savvoula

Full record

---

Summary

Thermography is a valuable method of diagnostic examination in the field of energy efficiency of buildings. By using infrared cameras, it is possible to detect and quantify heat loss through the structural elements of buildings. This helps us identify areas where problems occur, such as thermal bridges or poor insulation, and implement appropriate interventions to address them. The present research focuses on the analysis of the energy performance of buildings using passive thermography and in particular infrared inspection, a technique that allows the measurement of surface temperature by detecting the infrared radiation emitted by objects. Thermography provides a picture of temperature profiles and can reveal areas of unusual heat or cold on surfaces allowing us to identify deviations, deficiencies or even problems related to a building's energy performance. In our case, we apply two special research methods in combination: "Automated fly-past surveys" and "Time-lapse surveys". In particular, we received 104 photos over two days, at 15-minute intervals, allowing us to monitor the fluctuations in the building's temperature and identify potential anomalies.The research study being carried out uses a code that analyzes images from thermal cameras and detects unusual areas of excessively hot or cold temperature. This is achieved through the supervised learning method, where a Convolutional Neural Network (CNN) model is trained to recognize these unusual regions in the images. It therefore combines both qualitative and quantitative analysis. This code also uses oversampling and undersampling methods to deal with the imbalance of classes in the dataset, labels that describe regions in images, and a combination of various enhancement techniques that affect the performance of CNNs. These methods are Data Augmentation, Dropout, Stratified_split and callbacks. During training, the CNN model is trained to learn the features of unusual regions in order to recognize them effectively. In addition, the code includes a procedure to evaluate the performance of the model on the control set. Based on this performance, the best model is saved for future use. This allows for continuous improvement of the accuracy and efficiency of the model. Finally, the prediction models are evaluated and the decisions that must be made for the smooth operation of the buildings are analyzed.Overall, this research is an important contribution to the field of building energy performance. The combined approach of thermography and machine learning can lead to improved methods of energy management, promoting the achievement of more sustainable and efficient building systems. This approach can help identify energy-related problems, anomalies and deficiencies in buildings, promoting the health and efficiency of building systems.

Available Files

Services