Application of Multilayer Perceptron Method on Heat Flow Meter Results for Reducing the Measurement Time

Abstract

To reduce the impact on climate change, many countries have developed strategies for the building sector with a goal to reduce the energy demands and carbon emission of buildings. As most buildings that exist today will very likely exist in foreseeable future, many buildings will need to undergo major renovations. One of the most important parameters in determining the transmission heat losses through the building envelope is the U-value, i.e., thermal transmittance, and it is simply the rate of heat transfer per unit temperature. Since the U-value is one of the most important parameters regarding building energy performance, envelope elements that do not perform well in terms of transmission heat losses must undergo a renovation processes. The in-situ U-value of building elements is usually determined by the Heat Flux Method (HFM). One of the issues of current application of the HFM is the measurement duration. This paper explores the possibilities of reducing the measurement time by predicting the heat flux rate using a multilayer perceptron (MLP), a class of artificial neural network. The MLP uses two input layers that are the interior and exterior air temperatures, and the output layer that is the predicted heat flux rate. The predicted value is trained by comparing the predicted heat flux rates with the measured values, and by rearranging the neural network parameters (weights and biases) in corresponding neurons by minimizing the mean squared error defined for trained values (measured versus predicted heat flux rates). The use of MLP shows promising results for predicting the heat flux rates for the analyzed cases (4 days HFM results) when the training is performed on 2/3 or 1/2 of the overall measurement time. The application of the MLP could be in reducing the in-situ measurement time when determining heat losses through building elements in shorter time periods.

AI Powered Obstacle Distance Estimation for Onboard Autonomous Train Operation

“…The training process can be considered as adjusting the model weights and biases with the goal of minimizing a cost function [36][37][38]. The most important part of network design and performance is the adequate selection of the inputs and the preparation of the training set.…”

Image Retrieval for Local Architectural Heritage Recommendation Based on Deep Hashing

“…Recently, the work by Tejedor et al [5] introduces the future perspectives of the DL method for the diagnosis of heritage buildings. The methods reviewed in the paper cover several important topics, e.g., object detection which can be integrated with UAVs and a GPS system [6], 3D models for the reconstruction of lost architectural heritage [7], and neural networks applied to heat flux meter (HFM) methods [8,9]. Some defects of the DL method are also given in the paper, especially the limitation of dataset and computation time cost.…”

Combining Deep Learning and the Heat Flux Method for In-Situ Thermal-Transmittance Measurement Improvement

“…We can measure more than one element at a time with multiple HFM devices, but this way is financially and operationally inefficient [34]. On the other hand, there are examples in the literature of shortening the measurement time by applying artificial-intelligence (AI) models to the results to predict the resulting U-value based on in-situ methods [35][36][37][38][39].…”

“…These predicted heat-flux time series are then used to determine the U-value. The models from [37,38] have two input parameters-outdoor and indoor air temperature-and the resulting heat flux is predicted. In this case, differences from 0.78% to 8.73% can be obtained with respect to HFM measurements.…”

“…In this case, differences from 0.78% to 8.73% can be obtained with respect to HFM measurements. However, the article [37] does not show how the typology and architecture of the neural network was determined, and the application was developed only for one wall.…”