1. Introduction
Modern buildings consume one-third of the energy generated worldwide [
1]. Space conditioning, i.e., Heating, Ventilation, and Air Conditioning (HVAC) in these buildings, utilizes up to 50% of the consumed energy [
2]. Use of thermal insulation in exterior walls is one of the energy conservation strategies [
3]. However, determining the optimal thickness for insulating materials has emerged as a challenge. Available research has shown that other than cost-effective insulation thickness for exterior walls, the wall material, geographic location, and even the orientation of the building significantly influence energy savings [
4,
5,
6,
7]. Enhancing building energy efficiency could yield estimated savings of 20–40% [
8].
Existing research in the field of insulation applications emphasizes the importance of a tailored approach that considers multiple factors in a given environment. Ramin et al. [
9] highlighted the complexities in determining the optimum insulation thickness for conventional walls, specifically focusing on materials like aerated brick and concrete. Their research findings demonstrate that selecting insulation materials and their positioning can lead to significant variation in yearly transmission. Rosti et al. [
10] undertook an extensive analysis across different climate zones in Iran, employing a Life Cycle Cost Analysis approach. They pointed out the variation in optimal thickness depending on wall type, orientation, and environmental factors. Hou et al. [
11] investigated the optimal insulation thickness in the northeastern Sichuan hills of China and corroborated the notion of a nuanced approach to insulation application. Their study showed the impact of a heating degree day on the optimum thickness and found that reducing the heating degree per day could increase energy-saving rates. Ziapour et al. [
12] investigated Composite Prefabricated Wall Block (CPWB) design and showed that optimum thickness varies with atmospheric conditions.
The energy conservation solutions use optimal thermal insulation and structural layer thickness using experimental and computational methods [
3,
4]. The studies aimed to reduce the overall cost, increase energy savings, and reduce the payback period of insulation. This study uses previous research by Huang et al. [
13] as a basis for model formulation and result comparison. This study assumed that the results of Huang et al. could be further improved by using the nonlinear Least Squares Error (LSE) technique instead of the Life Cycle Cost Analysis technique. The considered technique has not been applied previously to optimize CIEW.
The rest of the paper has been arranged as follows:
Section 2 discusses the methodology and case study used to optimize the performance of CIEW.
Section 3 discusses the results.
Section 4 and
Section 5 discuss the results and provide conclusions, respectively.
2. Methodology
2.1. Case Study of a Masonry Wall Construction in Beijing
This study considered a case study of a CIEW in a cold area of Beijing by Huang et al. [
13]. Three heat sources were employed, namely, Coal-Fired Boiler (CFB), Gas-Fired Boiler (GFB), and Air-Source Heat Pump (ASHP), and three insulation materials were examined: Extruded Polystyrene (XPS), Rock Wool (RW), and Glass Wool (GW) [
13]. The authors determined the optimal structural layer thickness (
X), optimal insulation layer thickness (
Y), and payback period (
N) of a CIEW. Mathematical formulations, given in
Section 2.2, were also taken by Huang et al. They minimized NCC and maximized NCS by optimizing
,
and
, considering
300 mm. The parameters related to different heat sources and insulation materials which have been used as input to the formulated optimization problem, adopted from Huang et al. [
13], are given below.
The parameter is the tower heating value of energy consumed per unit of heating, which has been taken as 29,307 kJ/kg for CFB, 35,600 kJ/m3 for GFB, and 3600 kJ/kWh for ASHP heat source. is the efficiency of the heating source taken as 0.8 for CFB and 0.9 for GFB. 0.92 is the efficiency of the network. 4 is the coefficient of performance of the cooling system, which depends on the operating parameters. represents the fuel cost for heating per unit kilogram, volume, or kilowatt hour. It has a value of USD 158.44/m·t for CFB, USD 365.72/km3 for GFB, and USD 74.68/MWh for ASHP heat source. is the price of the insulation material per unit volume. This value is USD 125.69/m3, USD 132.11/m3, and USD 79.86/m3 for XPS, RW, and GW, respectively. represents the labor and insulation installation cost per unit area, given as USD 4.81/m2 for XPS, USD 9.75/m2 for RW, and USD 9.15/m2 for GW. Material and labor cost associated with mortar plaster on a unit surface area is given as . This value has been taken as 8.75 for XPS and 9.3 for RW and GW. USD 55.23/m3 denotes the cost of structural layer material and construction of unit volume. denotes the thermal conductivity of the structure and insulation material layer, which has been given as 0.03 W/mK for XPS and 0.04 for RW and GW. 0.22 W/mK represents the thermal conductivity of the structure material layer (W/m·K). 2699 °C·d and 94 °C·d are heating and cooling degree days based on 18 °C and 26 °C, respectively. 0.177 m2k/W is the overall thermal resistance of walls, excluding that offered by structural and insulation layers. USD 0.08/kW is the price of electricity per unit kilowatt hour. 20 years is the service life of the insulating material. Inflation () and market discount rate () are taken as 4.9% and 2.06%, respectively.
2.2. Mathematical Model
A quasi-steady state approach has been used to calculate the heat and cooling losses through the CIEW. Insulation thermal conductivity, material properties, and thickness are considered constant as these do not vary with temperature. It has been further assumed that insulation thickness and material requirements are as per the structural requirements and energy efficiency goals.
Composite Insulation External Walls (CIEW) comprise mortar plaster, a structural layer, and an insulation layer, as shown in
Figure 1. The thickness of each constituent layer has been kept constant.
2.2.1. Initial Investment Cost of a CIEW
Let
and
be the thickness of the structural and insulation layers, respectively. The initial investment cost of the wall per unit area
is computed using Equation (1) [
13].
2.2.2. Heating and Cooling Loss through CIEW
Considering all the constituents of the CIEW, the overall heat transfer coefficient (
) is computed using Equation (2) [
13].
The overall heat transfer coefficient (
) of the CIEW, without considering the insulation and structural layer, is given as Equation (3) [
13]. Accordingly, the overall heat transfer coefficient offered by the structural and insulation layer is computed using Equation (4) [
13].
The heating (
) and cooling (
) losses through the wall, considering structure and insulation layers, are computed using Equations (5) and (6), respectively [
13].
The difference in heating (
) and cooling (
) losses due to structure and insulation layers is computed using Equations (7) and (8), respectively [
13].
2.2.3. Energy Consumption Cost
The annual energy consumption cost through CIEW is given as Equation (9) [
13].
Accordingly, the difference between heating and cooling energy costs saved with and without considering structure and insulation layers is given by Equation (10) [
13].
2.2.4. Life Cycle Cost
Inflation and market discount rates are included using a present worth factor (
, Equation (11)) [
13].
and
are inflation and market discount rates, respectively.
The Life Cycle Cost (
LCC) per unit area is given in Equation (12) [
13].
The Life Cycle Savings (
LCS) per unit area is computed using Equation (13) [
13].
2.2.5. Payback Period
The structural and insulation cost will be recovered when Equation (13) is zero [
13]. Accordingly, the payback period (
) is computed using Equation (14).
2.2.6. Economic Insulation Layer Thickness and Overall U Value
Overall
U value (
) and economic insulation layer (
), with a given structural layer thickness (
), are given as Equations (15) and (16), respectively [
13].
The present work aims to minimize
and maximize
by optimizing
and
. Accordingly, the optimization problem is formulated in Equation (17).
2.3. Optimization Technique: Nonlinear Least Squares Error Method (LSE)
The LSE algorithm was used to formulate and solve a complex optimization problem, as given by Equation (17), to minimize Life Cycle Cost (LCC) and maximize Life Cycle Savings (LCS) by optimizing variables and . This application of the LSE represents a novel approach to determining the optimal construction form and overall U-value of Composite Insulation External Walls (CIEW), a subject of considerable interest in sustainable construction practices. The method involves an iterative process, where a linear estimation is initially made and then iteratively refined using a dataset of “” data points and a nonlinear fitting curve constructed using a vector of unknown parameters (). This method has not been applied previously to determine the optimal construction form and overall U-value of Composite Insulation External Walls (CIEW). The LSE is a regression technique specifically designed for nonlinear optimization problems where the model is nonlinear in several unknown parameters. Unlike linear regression techniques, the LSE requires an iterative process due to the nonlinearity of independent variables.
When a model is nonlinear in “
n” unknown parameters, nonlinear least squares analysis can be used to fit a collection of
m (>
n) observations. The method’s basic idea is to estimate the model using a linear model and then iteratively adjust the parameters. The dataset is given by “
” data points {(
), (
), (
), ……, (
)}. The nonlinear fitting curve is given by
, where
is a vector of unknown parameters. Finding the vector of parameters that allows the curve to best match the provided data in the least squares sense, the sum of squares is given as Equation (18).
Since the derivatives in a nonlinear system () rely on both the independent variable and the parameters. These gradient equations typically lack a closed solution. Instead, the parameters must be given beginning values. The parameters are then repeatedly refined, i.e., values are derived by repeated approximation.
The iterative process of the nonlinear Least Squares Error (LSE) method uses specific constraints to ensure precision and convergence. The maximum number of iterations was 400, and the maximum function evaluations were limited to 100. After each iteration, the maximum change in independent and dependent variables was constrained to 10−6 to maintain accuracy. The derivatives within the system were approximated using Newton’s forward difference method, a technique that balances computational efficiency and accuracy. Additionally, the perturbations were bound within a range, with minimum and maximum perturbations set at 0 and infinity, respectively. These constraints were carefully chosen to align with this study’s objectives and provide a robust optimization framework.
As shown in
Figure 2, the optimization begins by defining the design variables’ error function, constraints, termination criteria, and upper and lower bounds. An initial guess for the design variables is established, and the iteration counter is set to zero. The target output variable is computed using the current design variable vector. Then, the error is calculated using a specified function. A decision is made to determine whether the computed error is less than or equal to a predefined threshold (
10
−6), or the number of iterations is less than 100, or the maximum change in any design variable is less than 10
−6 to process the next iteration. If the error is within acceptable bounds, the process stops; otherwise, the design variable vector is updated, and the iteration counter is incremented. The process then loops back, and the computation of the target output variable is repeated using the updated design variables. This iterative refinement continues until the error criterion is met. At this point, the optimization process concludes. The method described represents a systematic approach to finding optimal solutions in various scientific and engineering applications by continuously refining the design variables.
This meticulous approach, balancing computational efficiency and accuracy, offers a robust framework for exploring the intricate relationships between insulation and structural layers in building design. Doing so contributes valuable insights to the ongoing discourse on thermal efficiency, cost considerations, and environmental impact in mechanical engineering and sustainable construction, demonstrating the potential of optimization techniques in enhancing building performance.
3. Results
The optimized values of structural and insulation thickness for CFB heating systems when insulated with XPS, RW, or GW are given in
Table 1. The computed
LCC,
LCS, and payback period are also mentioned in
Table 1. The optimal values of
X and
Y and computed values of
LCC,
LCS, and
N for GFB and ASHP are given in
Table 2 and
Table 3, respectively.
Table 1 and
Figure 3 illustrate the variation in the thickness of
and
and corresponding computed values of
LCC,
LCS, and
, when
= constant and CFB is the heat source. For XPS, as the constant value of
varied from 200 mm to 400 mm with an interval of 50 mm,
increased, while
decreased. This led to a gradual increase in
LCC from USD 36.95 to USD 44.05, a decrease in
LCS from USD 115.83 to USD 108.70, and an increase in
N from 4.97 to 6.22 years. The trend for RW was similar: LCC increased from USD 43.21 to USD 48.80,
N increased from 6.07 to 7.08 years, and
LCS decreased from USD 109.54 to USD 103.92. For GW, the pattern was also similar:
LCC increased from USD 40.31 to USD 49.88,
N increased from 5.56 to 7.27 years, and
LCS decreased from USD 112.48 to USD 102.91. The
LCC and N increased, and LCS decreased with a decrease in
and an increase in
. These were more gradual in RW insulation material, followed by GW and XPS. The change was almost linear in RW and GW, whereas fluctuating but continuous change was seen in XPS.
Table 2 and
Figure 4 illustrate the variation in
and
layers concerning
LCC,
LCS, and
N when GFB is the heat source. For XPS, as the
increased from 200 mm to 400 mm,
LCC increased from USD 42.32 to USD 50.27, and
increased from 2.95 to 3.75 years, while
decreased from USD 214.52 to USD 206.57. The RW insulation followed a similar trend:
LCC increased from USD 48.17 to USD 53.76,
N from 3.53 to 4.10 years, and
LCS decreased from USD 208.65 to USD 203.03. GW response was also similar:
LCC increased from USD 45.27 to USD 54.84,
N from 3.24 to 4.21 years, and
LCS decreased from USD 211.60 to USD 202.02. The
LCC and
N increased, and
LCS decreased with a decrease in
and an increase in
. These changes were more gradual in RW insulation material, followed by GW and XPS. The change was almost linear with RW. A fluctuating but continuous change was seen with XPS. The change was two piecewise linear with GW which was less steep until
300 mm, and then the slope comparatively increased until
400 mm.
Table 3 and
Figure 5 show the optimal
and
layer thickness and corresponding
,
, and
values when ASHP is the heat source. The algebraic sum of
and
is a constant, which varies from 200 mm to 400 mm, with an interval of 50 mm. When XPS was used as an insulating material,
increased from USD 37.40 to USD 46.75,
increased from 4.90 to 6.52 years, and
decreased from USD 118.89 to USD 109.54. A similar pattern was seen when using RW insulation.
LCC increased from USD 44.23 to USD 51.12,
increased from 6.08 to 7.29 years, and
decreased from USD 112.05 to USD 105.15. The GW pattern was similar, where
LCC increased from USD 40.95 to USD 50.22,
increased from 5.51 to 7.13 years, and
decreased from USD 115.37 to USD 106.08. The
LCC and
N increased, and
LCS decreased with a decrease in
and an increase in
. These changes were more gradual in XPS insulation material, followed by RW and GW. A bilinear curve was seen with XPS with continuity break point at
200 mm. The slope decreased with an increase in
after
200 mm. The change was almost linear with RW. Bilinear piecewise curve was seen with XPS at intersection point
200 mm. The slope increased with an increase in
after
200 mm.
Table 4 presents the coefficient of correlation between different parameters for three heat sources: CFB, GFB, and ASHP. Three insulation materials are considered for each heat source: XPS, RW, and GW. The correlation coefficients show a significant correlation between
and
with
LCC,
LCS, and
. For the CFB heat source, CFB-XPS and CFB-RW show strong positive correlations between
,
LCC, and
N and strong negative correlations with
LCS. The GFB and ASHP heat sources follow a similar pattern.
The results from nonlinear LSE were compared with those of Huang et al. [
13], as given in
Table 5. The LSE technique could minimize the
LCC of all heating systems (CFB, GFB, and ASHP) for XPS and RW but not for GW insulation material compared to Huang et al.’s Life Cycle Cost Analysis (LCCA). The nonlinear LSE technique maximized
LCS for all heating systems with three insulation materials. The negative sign showed a decrease in
LCC percentage when using the LSE technique compared to Huang et al.’s technique.
5. Conclusions
Heating, Ventilation, and Air Conditioning (HVAC) in modern buildings consumes around 50% of the total energy consumed. The energy consumed due to the space conditioning can be reduced by providing insulation in the outer walls of the building. This requires the selection of the optimal thickness of insulation material used in the construction. The performance of CIEW can be improved by optimizing the structural and insulation layer thickness. The use of the nonlinear Least Squares Estimation (LSE) optimization technique for optimizing the LCC and LCS of CIEW has shown promising results. This study used three insulation materials—Extruded Polystyrene (XPS), Rock Wool (RW), and Glass Wool (GW)—across three heat sources, namely, Circulating Fluidized Bed (CFB), Grate-Fired Boiler (GFB), and Air-Source Heat Pump (ASHP), to calculate the thickness of structural and insulation layers for optimum CIEW. This study compared the use of traditional methods using LCCA and nonlinear LSE for optimizing the LCC and LCS of CIEW. The key innovations of this study are described in the following paragraphs.
The use of the LSE technique resulted in improved performance over that of using LCCA while using XPS, RW, and GW as insulation materials with any of the three heat sources, namely, CFB, GFB, and ASHP. In every configuration of the insulation material and heat source, there is a strong positive correlation between optimal structural layer thickness (X) and LCC and payback period, while there is a strong negative correlation between X and LCS.
In every configuration of the insulation material and heat source, there is a strong negative correlation between optimal insulation layer thickness (Y) and LCC and payback period, while there is a strong positive correlation between Y and LCS. The LSE method improved the LCS for all configurations of insulation material and heat sources, as compared to the LCCA method. The highest improvement was for the CFB-RW configuration, with 9.12% more than the LCS value. The LSE method reduced the LCC for all configurations of insulation material and heat sources, except for the GW heat source configurations, for which it showed an increase in LCC. The highest reduction was for the CFB-RW configuration, with a 7.41% decrease in LCC value.
The results show that the LSE method has a greater economic efficiency than the LCCA method. However, it also has some limitations, for example, it reports higher LCC values for GW heat source configurations. Therefore, an integrated approach should be developed to use the strengths of both methods.