Combining Galerkin Methods and Neural Network Analysis to Inversely determine Thermal Conductivity of Living Green Roof Materials

Abstract

Use of living biological materials for roof greening is an effective strategy for Urban Heat Island (UHI) mitigation. Sunagoke moss mat is a living composite material with good potential for active roof greening. However, the determination of its thermal conductivity is difficult because it is fabricated from organic (live Sunagoke moss and cotton wool) and inorganic (polyvinyl chloride (PVC) netting) constituents. Thermal conductivity of this living composite material was determined using inverse finite-element modelling and neural network optimisation. The research hypothesis was that this property can be modelled as a weighted average of thermal conductivities of its constituents. Temperature regimes on either side of a Sunagoke moss mat (dimension 110 mm by 100 mm by 18·1 mm) in a thermally insulated transmission system were measured at different dry basis (db) water contents. Temperature regimes at 0%, 50% and 100% water contents (db) were input into a finite-element model and used to calculate nodal temperatures at six nodes across the thickness of the material. A weighted average model was used to compute effective conductivity for each element from the thermal conductivities of its constituents. Four multi-layer Perceptron (MLP) neural network models were developed and used to optimise effective thermal conductivity of the material at 0%, 50% and 100% water contents (db). The effective thermal conductivity of the material was 0·1 W m−1 K−1at 0%; 0·24 at 50% and 0·28 at 100% water contents (db). Nodal temperature profiles showed that the material exhibits steady-state heat transfer at 0% and 100% water contents (db) and transient-state heat transfer at water contents in between. It was concluded that inverse finite-element modelling is a feasible alternative technique for evaluating the thermal conductivity of living roof and wall greening materials and other biomaterials where use of conventional methods is difficult or highly erroneous.