Possibilities of regression analysis in processing thermal conductivity measurement data

When implementing energy saving measures, the key is the correct choice of thermal insulation materials, the main characteristic of which is the thermal conductivity coefficient. Missing part of the data, which may occur during investigation of materials in natural conditions, can lead to incorrect determination of the corresponding characteristic, which negatively affects the effectiveness of the implemented measures and energy saving. Therefore, reconstruction of the missing data at the stage of preliminary processing of measured signals to obtain complete and accurate data when determining the thermal conductivity of thermal insulation materials will avoid this situation. The article presents the results of regression analysis of data obtained during express control of thermal conductivity of thermal insulation materials based on the local thermal impact method. Regression models were built for signal reconstruction with 10%, 20% and 30% missing data, using which a relative error of determination the thermal conductivity coefficient of less than 8% was obtained. This is acceptable for express control of thermal conductivity and indicates the correctness of data restoration in this way. In addition, an algorithm is provided for determining signal stationarity, which allows to reasonably reduce the duration of each material with a given level of permissible error.

---

Opracowanie rekordu ze środków MNiSW, umowa nr POPUL/SP/0154/2024/02 w ramach programu "Społeczna odpowiedzialność nauki II" - moduł: Popularyzacja nauki (2025).