Thermal Performance Analysis and Structural Optimization of Main Functional Components of Computers

Featured Application

This study aims to investigate the regularity and mechanism on the influence of positioning of the main functional components in a computer to its thermal performance.

Abstract

In today’s data-driven age, the thermal properties of computer transistors play an important role. In this research, finite element simulation is employed to construct the structural model of the primary components within a computer chassis, and the thermal performance is evaluated based on ambient temperature, thermal conductivity, and heat dissipation rate. By combining the particle swarm optimization algorithm with numerical simulation for joint simulation and structural optimization, the component layout was optimized to reduce the working temperature. The results show that when the background temperature, that is, the ambient temperature, rises from −20 °C to 60 °C, the maximum operating temperature of the computer is approximately 88 °C. The maximum temperature is mainly in the transistor core and the minimum temperature is in the intake grille, and the operating temperature of the optimized structure decreases by approximately 10 °C. The research shows that the operating temperature is most sensitive to the change of background temperature, and the transistor core is the main heating source. The maximum temperature can be reduced by rationally adjusting the position of the components. This study provides a reference for analyzing the thermal performance of computers and optimizing structures.

1. Introduction

With the rapid development of computer technology, there has been a significant increase in performance demands for computing systems, particularly in the domains of high-performance computing (HPC) and big data processing [1,2,3,4,5,6,7,8,9,10]. This escalating demand is exemplified by projects such as the Mochi framework, which enables efficient utilization of cutting-edge HPC hardware for customized data services [1], evolutionary optimization techniques applied to industrial processes [3], and advanced analytical models for healthcare data processing [5]. Consequently, there is growing pressure on critical components like central processing units (CPUs), necessitating enhanced thermal performance and structural optimization [1,2,3,4,5,6,7,8].

The escalating heat flux densities in high-performance components have rendered traditional thermal designs inadequate for efficient cooling, presenting substantial challenges to thermal management technologies [11,12,13,14,15,16,17,18,19,20,21,22,23,24]. In response, researchers have developed innovative solutions including nanofluid-cooled heat sinks demonstrating superior heat removal compared to conventional water cooling [11]; optimized fin structures achieving peak thermal performance at specific operational parameters (e.g., 12.5 m/s wind speed, 110 W power) [14]; and advanced thermal interface materials enabling compact aerospace computing systems [18]. Such thermal structure optimization not only enhances computational performance and reliability but also improves energy efficiency while ensuring stability across diverse environmental conditions [11,12,13,14,15,16,17,18,19].

Thermal management and structural optimization are particularly crucial for computer performance, and the field has already made some progress. He et al. [25] designed a finned water-cooled heat sink to find out the best structure by optimizing its thermal resistance and pressure drop. Results showed that the optimal fin height, fin thickness and fin spacing were 5 mm, 0.5 mm and 1 mm, respectively, and the cooling performance was significantly improved. Shuang et al. [26] constructed a thermal resistance network model of a thermal module for a laptop computer and optimized the thermal performance and cost using the non–dominated sorting genetic algorithm II (NSGA–II). It was shown that the cooling performance of the laptop was improved by 24% and the cost was reduced by 9%. Li et al. [27] proposed a 7th-order transfer function model for the thermal dissipation of devices and tested it on a computer. It is shown that the proposed model achieves a time delay of about 1 h and a maximum heat dissipation reduction of 33.89%, which is helpful for calculating the dynamic cooling demand. Jin et al. [28] investigated the impact of heat-generating components on tower servers and performed structural optimization. It was shown that narrowing the air outlet free area ratio (FAR) and repositioning the graphics processing unit (GPU) can significantly reduce the temperature inside the server, and the best optimized structure can reduce airflow by 35%. Antonio et al. [29] designed a silicon nanocrystal that enables localized heat dissipation in ultra-large scale devices.

The above studies have achieved a certain impact in various aspects, providing a reference direction for thermal performance analysis and structural optimization of computers. However, little has been done to focus on the influence of the thermal conductivity of materials, the heat dissipation rate of electronic components and the location of components on computer heat dissipation, all of which have a weighty role in the thermal performance of computers [30]. Therefore, in this study, after constructing a rough structural model of the mainframe box in finite element simulation, the background temperature of the computer, the thermal conductivity of materials per thermal analysis, and the heat dissipation rate of electronic components were taken as starting points. The particle swarm algorithm was used to optimize and adjust the positions of electronic components to derive the best thermal structure, keeping the operating temperatures of electronic components within a reasonable range and extending the service life of the computer.

2. Models and Methods

Figure 1 shows the geometric structure of the power supply. It is composed of a vetilated casing with dimensions of 14 cm × 15 cm × 8.6 cm, made of aluminum 6063–T83. In-side the casing, only obstacles with characteristic lengths of at least 5 mm are displayed. The bottom of the casing represents the printed circuit board (PCB). The anisotropic thermal conductivities along the x-axis, y-axis, and z-axis are 10, 10, and 0.3 W/(m·K), respectively, with a density of 430 kg/m³ and a special heat capacity at constant pressure of 1100 J/(kg·K). Due to the relatively low thermal conductivity along the z–axis and the fact that the PCB and the sides of the casing are separated by thin layers of air, there is no need to model the bottom wall, nor is it necessary to consider the cooling of these sides.

The capacitors are approximated with aluminum parts. Heat sinks and casings are made of the same aluminum alloy. Inductors mainly consist of steel core and copper coils. The transformer is made of copper, steel, and plastic. Transistors are treated as dual-domain components with a silicon core and plastic casing, and the core contacts an aluminum heat sink for better heat transfer. Airflow is modeled as turbulent using the algebraic y + turbulence model. The max power of the simulated power supply is 230 W. Electronic components are grouped as heat sources in

Table 1. The parameters of the electronic components treated as heat sources.

Components	Dissipated Heat Rate (W)	Number (pcs)
Transistor cores	25	5
Large transformer coil	5	1
Small transformer coils	3	3
Inductors	2	3
Large capacitors	3	2
Medium capacitors	2	7
Small capacitors	1	4

. Total heat dissipation is 41 W with 82% efficiency.

The inlet air temperature is set to 20 °C, as the air is drawn from the computer case, which has already been used to cool other components. The inlet boundary is configured as a grille boundary condition, where the pressure must describe the head loss caused by the air entering the casing. The head loss coefficient k_grille is represented by the following 6th–order polynomial, as shown in Equation (1) [31].

$$k_{grille}=12084{\alpha }^6-42281{\alpha }^5+60989{\alpha }^4-46559{\alpha }^3+19963{\alpha }^2-4618.5\alpha +462.89$$

(1)

where α is the opening ratio of the grill.

Figure 2 shows the relationship between the pressure loss coefficient and the grating opening rate. OR is denoted as the grill opening ratio. Figure 2 indicates that the pressure loss coefficient decreases as the open area ratio increases, and it is almost zero when the open area ratio reaches 0.4. Considering the needs of the simulation, the open area ratio is taken as 0.4 for subsequent research.

Pressure head loss ΔP (Pa) is expressed by Equation (2).

$$\Delta P=k_{grille}{\displaystyle \frac{\rho {V_0}^2}{2}}$$

(2)

where ρ is the density (kg/m³) and V₀ is the flow rate (m³/s).

The chassis, printed circuit boards, inductor surfaces and heat sinks are set up as thin conductive layers.

2.1. Finite Element Simulation Model

Figure 3 is the solid–fluid finite element simulation model of the computer case. The model consists of the casing, power switch, power connector, fan, wire assembly, grille, circuit board, two large capacitors, seven medium capacitors, four small capacitors, one large transformer, three small transformers, three inductors, five transistor cores, two rows of fins, and two heat sinks.

Solid and fluid heat transfer simulations are performed for the transistors, large transformer, small transformer, inductor, large capacitor, medium capacitor and small capacitor with the initial value of ambient temperature (background temperature) and heat dissipation rates shown in Table 1. The inlet temperature is set based on the reference operating conditions. The actual value may increase due to the thermal load of the upstream components. The upstream temperature of the inflow is ambient, the thin-shell type is a multi-layered shell, the type of the layer is a thermally thin approximation, and thermal insulation is used for all the thin layers.

The air domain is turbulent, with algebraic y⁺ physics field simulation, and the reference pressure level is 1 Pa. The initial velocity field is $\overrightarrow{\mathrm{u}}$ = (−1, 0, 0), and the pressure is 0 Pa. All boundary wall conditions are no–slip, and the fin inner wall boundary conditions are set to no–slip conditions.

The grill is set as the inlet, with the flow direction being normal flow, the input pressure P_input = 0 Pa, the flow condition is characterized by secondary loss, and the secondary loss coefficient qlc (kg/m⁷) is shown in Equation (3).

$$qlc=k_{grille}(OR)\times nitf1.\rho /2$$

(3)

where k_grille is the head loss coefficient, OR is the grating opening ratio, and $nitf1.\rho$ is the density of the fluid.

The fan is set as the outlet, with the outlet pressure P_exit = 0 Pa, and the flow condition follows a static pressure curve. The static pressure curve data is shown in

Table 2. Data of static pressure curve.

Flux (m3/s)	Static Pressure (Pa)
0	12.3
8.33333 × 10−4	11.4
1.666667 × 10−3	9
2.5 × 10−3	6.3
2.916667 × 10−3	6
3.333333 × 10−3	5.8
3.75 × 10−3	4.3
4.166667 × 10−3	2.2
4.583333 × 10−3	0.7
4.833333 × 10−3	0

, and the interpolation function type for the static pressure curve is piecewise cubic.

In this study, the meshing parameters for the finite element simulation model are as follows. The maximum element size is 1.4 cm, the minimum element size is 0.25 cm, the maximum element growth rate is 1.6, the curvature factor is 0.7, and the narrow region resolution is 0.4. For the wire assembly surface mapping, the maximum element size is 0.5 cm, the minimum element size is 0.4 cm, and the maximum element growth rate, curvature factor, and narrow region resolution are the same as above; for the small wire assembly surface mapping, the maximum element size is 0.15 cm, and the minimum element size, maximum element growth rate, curvature factor, and narrow region resolution are the same as above. The heat exchange surface mesh type is a free triangular mesh, with a maximum element size of 0.5 cm, a minimum element size of 0.4 cm, a maximum element growth rate of 1.05, a curvature factor of 1, and a narrow region resolution of 1; the curved area mesh type is a free triangular mesh, with a maximum element size of 0.735 cm, a minimum element size of 0.22 cm, a maximum element growth rate of 1.15, a curvature factor of 0.6, and a narrow region resolution of 0.7; the remaining parts are free tetrahedral meshes. The boundary layer property parameters are the following: number of layers is 3, stretch factor is 1.2, thickness detail is automatic, and thickness adjustment factor is 1.

2.2. Parameters of the Study

In this study, the adjustable parameters are divided into three main categories: background temperature, thermal conductivity, and heat dissipation rate. The thermal conductivity is further divided into the thermal conductivity of acrylic plastic used for encapsulation, the thermal conductivity of steel components, the thermal conductivity of aluminum capacitors, the thermal conductivity of copper transformer coils, the thermal conductivity of transistor silicon chips, the thermal conductivity of heat sinks, the thermal conductivity of copper cladding, the thermal conductivity of the circuit board in the x (y) direction, and the thermal conductivity of the circuit board in the z direction. The initial parameters are shown in

Table 3. Initial values of thermal conductivity.

Name	Numerical Value
Thermal conductivity of acrylic plastic for encapsulation	0.2 W/(m·K)
Thermal conductivity of steel components	45 W/(m·K)
Thermal conductivity of aluminum capacitor	240 W/(m·K)
Thermal conductivity of copper transformer coils	400 W/(m·K)
Thermal conductivity of transistorized silicon chip	130 W/(m·K)
Thermal conductivity of heat sink	200 W/(m·K)
Thermal conductivity of copper skin	400 W/(m·K)
Thermal conductivity of the circuit board in the x (y) direction	10
Thermal conductivity in the z direction of the circuit board	0.3

. The classification and initial parameters of the heat dissipation rate are shown in Table 1, and the initial value of the ambient temperature is 20 °C.

The setting of initial values takes into account practical situations, thus closely relating the research to real-life situations. In terms of the range for parametric scanning in this study, considering the rigor and reliability of the data, all adjustable parameters except for the ambient temperature are scanned with an interval of 1/20 of their initial values, taking five parameter values on each side of the initial value, totaling 11 points as the range for parameter scanning. The ambient temperature, due to its significant impact, needs to be considered separately, and its range is set from −20 °C to 60 °C, with an interval of 5 °C, totaling 17 points as the range for parametric scanning.

3. Results and Discussions

Based on the constructed solid–fluid heat transfer and turbulent, algebraic y⁺ finite element simulation, the thermal performance of the computer case was simulated, thereby verifying the significance and practicality of this study.

3.1. Thermal Performance Mechanism

The thermal performance mechanism of the computer case is shown in Figure 4. This performance was achieved through the finite element simulation calculations using the finite element simulation model shown in Figure 3, with an ambient temperature of 20 °C and parameters corresponding to the initial values in Table 1 and Table 3.

From Figure 4a, it can be seen that high temperatures are mainly concentrated on the heat sinks, transformers, and inductors, while low temperatures are found on the case shell. Figure 4b indicates that the air velocity is the fastest along the horizontal distance from the inlet grille to the outlet fan. Figure 4c shows that except for the outlet fan having pressure, the rest of the areas are essentially at 0. Figure 4d describes that the wall resolution of the surrounding walls is higher than that of the internal areas. Figure 4e can be considered a combination of Figure 4a,b.

Figure 4 demonstrates that the temperature distribution has the greatest influence on the thermal performance mechanism, so this study mainly focuses on the temperature distribution to analyze the thermal performance of the computer case under various changing parameters.

3.2. Background Temperature

Before analyzing the background temperature, it is essential to ensure that other thermal conductivity coefficients and heat dissipation rates remain unchanged, taking their initial values from Table 1 and Table 3 to ensure the accuracy of the study. The impact of background temperature changes on the working temperature distribution is shown in Figure 5, with the trend depicted in Figure 6. The range of background temperature variation is from −20 °C to 60 °C, with intervals of 10 °C. It can be observed that as the background temperature increases, the minimum and maximum working temperatures also rise, with a significant range of change. When the background temperature rises by 10 °C, the highest and lowest working temperatures increase by approximately 10 °C, essentially following the linear growth relationship shown in Figure 6, and they always maintain a temperature difference of about 50 °C. This indicates that the impact of background temperature changes on working temperature is substantial, consistent with actual conditions, and the two are essentially directly proportional. Looking at the locations of the highest and lowest temperatures, the highest is always at the transistor core, while the lowest is at the grille and air intake, which aligns with the concept of the transistor being the primary heat source.

3.3. Heat Consumption Rate

3.3.1. Transistor Heat Dissipation Rate

At a background temperature of 20 °C, with only the transistor’s heat dissipation rate changing, the temperature distribution and temperature trend change diagrams for the minimum value of 18.75 W and the maximum value of 31.25 W are shown in Figure 7. As the transistor’s heat dissipation rate increases, the highest temperature steadily rises to 80.0938 °C, essentially increasing linearly; the lowest temperature remains essentially unchanged, fluctuating within a small range of 0.4 °C. The highest and lowest temperatures are still located at the transistor core and the intake grille, respectively. The study indicates that although the transistor core, as the main heat source, does not have as significant an impact on working temperature as the background temperature, it is still not negligible and must be considered in optimization.

3.3.2. Heat Consumption Rate of Large Transformers

At a background temperature of 20 °C, with only the heat dissipation rate of the large transformer changing, the temperature distribution and temperature trend change diagrams for the maximum value of 6.25 W and the minimum value of 3.75 W are shown in Figure 8. Similarly to the previous case, as the heat dissipation rate of the large transformer increases, the highest temperature also rises, essentially maintaining a direct proportional relationship, while the lowest temperature fluctuates within a small range of 0.2 °C. However, when the heat dissipation rate is 6.25 W, the highest temperature is no longer at the transistor core but appears on the large transformer. This is because when the large transformer is at 6.25 W, its heat generation capability exceeds that of the transistor core, making it the primary heat source. The study indicates that the heat dissipation rate of the large transformer still has a certain impact on the working temperature, but not as significant as the background temperature and the heat dissipation rate of the transistor.

3.3.3. Residual Heat Consumption Rate

The residual heat dissipation rates include small transformer heat dissipation rate, inductor heat dissipation rate, large capacitor heat dissipation rate, medium capacitor heat dissipation rate, and small capacitor heat dissipation rate. The temperature distribution for the maximum and minimum values is shown in Figure 9, with the maximum and minimum values listed in

Table 4. Maximum and minimum values for the various residual heat dissipation rate.

Name	Max (W)	Min (W)
Heat consumption rate of small transformers	3.75	2.25
Inductor heat dissipation rate	2.5	1.5
Large capacitor heat consumption rate	3.75	2.25
Medium capacitor heat consumption rate	2.5	1.5
Heat dissipation rate of small capacitors	1.25	0.75

, and the temperature trend changes are shown in Figure 10. Research indicates that the impact of these heat dissipation rates on working temperature is minimal, with the highest temperature fluctuating within a range of about 1 °C; the lowest temperature varies within a range of 0.1 °C, or even remains unchanged, as seen in Figure 9i,j. This phenomenon is more evident in Figure 10, where the highest and lowest temperatures almost form a straight line. The main reason for this is that these electronic components are not primary heat sources, and the initial values taken are relatively small. Furthermore, taking one–twentieth of these values results in a limited scanning range, which collectively causes them to have little impact on working temperature. As with most of the previous cases, the highest temperature consistently appears at the transistor core, while the lowest temperature is at the intake grille, which aligns with the construction concept and the transistor being the main heat source.

3.4. Thermal Conductivity

3.4.1. Heat Sink Thermal Conductivity

At a background temperature of 20 °C, with only the thermal conductivity of the heat sink changing, the temperature distribution and temperature trend changes for the maximum value of 250 W/(m·K) and the minimum value of 150 W/(m·K) are shown in Figure 11. It can be observed that as the thermal conductivity of the heat sink increases, the highest working temperature shows a linear decrease, while the lowest temperature fluctuates within a small range of 0.3 °C. This is because the higher the thermal conductivity of the heat sink, the more heat it removes, leading to a decrease in the overall temperature of the casing. Similarly, the highest temperature remains at the transistor core, and the lowest temperature is at the intake grille, which aligns with the concept.

3.4.2. Remaining Thermal Conductivity

The remaining thermal conductivities include the thermal conductivity of the acrylic plastic used for encapsulation, the thermal conductivity of steel components, the thermal conductivity of aluminum capacitors, the thermal conductivity of copper transformer coils, the thermal conductivity of the transistor silicon chip, the thermal conductivity of the copper skin, the thermal conductivity of the circuit board in the x (y) direction, and the thermal conductivity of the circuit board in the z direction. The maximum and minimum values are shown in

Table 5. Maximum and minimum values for the various remaining thermal conductivity.

Name	Max(W/(m·K))	Min(W/(m·K))
Thermal conductivity of acrylic plastic for encapsulation	0.25	0.15
Thermal conductivity of steel components	33.75	56.25
Aluminum capacitor thermal conductivity	180	300
Thermal conductivity of copper transformer coils	300	500
Transistorized silicon chip thermal conductivity	97.5	162.5
Copper skin thermal conductivity	300	500
Thermal conductivity of the circuit board in the x (y) direction	7.5	12.5
Thermal conductivity of the circuit board in the z direction	0.225	0.375

, the temperature distribution is shown in Figure 12, and the temperature trend changes are shown in Figure 13. From Figure 13, it can be observed that the changes in these thermal conductivities have a negligible impact on the working temperature, with the highest temperature fluctuating only within a range of 1 °C, and the lowest temperature remaining almost unchanged. This is because the cooling effect is primarily achieved by the heat sink, and these electronic components with the mentioned thermal conductivities do not have good cooling effects, thus not removing much heat. However, the locations of the highest and lowest temperatures are consistent with the previous studies, being at the transistor core and the intake grille, respectively, which aligns with the research findings.

All the research findings indicate that background temperature is the most significant factor affecting working temperature, as it is the primary source of heat. Among the heat dissipation rates, the transistor heat dissipation rate has the greatest impact because the transistor core is the main heat source among the electronic components. In terms of thermal conductivity, the heat sink’s thermal conductivity is the most influential because the heat sink is the primary cooling device. Except for the case of the large transformer heat dissipation rate, the highest temperature is always at the transistor core, as it is the main heat source among the electronic components; the lowest temperature is always at the grille, as the grille is the air intake.

4. Optimization

The typical operating temperature range for common electronic components is −40 °C to 85 °C. According to Figure 5f,g, when the background temperature increases from 30 °C to 40 °C, the working temperature rises from 80.3918 °C to 91.3424 °C. There is a certain background temperature in between that causes the working temperature to exceed the maximum limit of 85 °C. By extracting the temperature distribution at a background temperature of 35 °C, as shown in Figure 14, it can be observed that the working temperature has already exceeded the range when the background temperature is 35 °C, rendering the computer inoperable. However, during summer, the background temperature can reach around 40 °C. Therefore, to meet operational requirements, it is necessary to optimize the model.

4.1. Optimization Methods

Particle Swarm Optimization (PSO), also known as the PSO algorithm or Bird Flocking Algorithm, is a new evolutionary algorithm developed in recent years by J. Kennedy, R. C. Eberhart, and others [32,33,34,35,36,37,38,39]. PSO is a type of evolutionary algorithm, similar to the simulated annealing algorithm, as it starts from a random solution and iteratively searches for the optimal solution [32]. It also evaluates the quality of solutions using fitness, but it is simpler than genetic algorithms, lacking the “crossover” and “mutation” operations [36]. Instead, it searches for the global optimum by following the best solution found so far [38]. This algorithm has gained attention in the academic community for its ease of implementation, high precision, and fast convergence, and has demonstrated its superiority in solving practical problems. PSO is a parallel algorithm [32,33,34,35,36,37,38,39].

This study optimizes the best positions of electronic components using the PSO algorithm to reduce working temperatures and keep them within the range of −40 °C to 85 °C. However, it is important to note that during the design optimization, electronic components should not interfere with each other. Ensuring the independence of their positions is crucial to guarantee the accuracy and rigor of the optimization results.

The specific flowchart of the PSO algorithm in this study is shown in Figure 15, where randomly generated particles (representing the combination of component positions) are initialized and the initial velocity is assigned. The fitness calculation is to calculate the maximum operating temperature (objective function value) of the current layout through finite element simulation. The update machine updates the individual/global optimal solution to compare the current temperature with the historical optimal value. Constraint processing checks whether components overlap (position interference), and if violated, the position is updated. Iteration termination occurs when the maximum number of iterations (50 times) is reached or the temperature drops to a safe range (≤85 °C).

The core elements of an algorithm consist of four components: the objective function, design variables, constraints, and convergence settings.

The definition of the objective function is to minimize the maximum operating temperature of electronic components, which is expressed by Formula (4).

$$M\mathrm{in}imizef(x)=T_{\max }(x)$$

(4)

Here, $x$ represents the design variable vector, and $T_{\max }$ is the maximum temperature value obtained through finite element simulation.

The design variables are the three-dimensional coordinate positions of the electronic components in

Table 6. Changes in position parameters.

Components	Initial Position	Optimized Position
Large transformers	(11.05, 8.5, 1.75)	(4.05, 8.5, 1.75)
Small transformers	(5.775, 8.75, 0.875)	(11.775, 10.25, 0.875)
Heat sinks and transistor cores	(8.75, 6.2, 30)	(8.9, 7.8, 34)
Inductors	(8, 7.3, 1)	(8, 6, 1)

. The specific variables are the coordinates of the large transformer, the small transformer, the heat sink/transistor core, and the inductor. The rotation angle of the large transformer also needs to be considered.

The constraint condition is that there is no positional interference, that is, the geometric spaces of any two components do not overlap. The boundary constraint is that the component positions must be inside the chassis.

$$dist(C_i,C_j)\ge r_i+r_j,\forall i\ne j$$

(5)

Here, $C_i$ represents the center coordinates of the component, and $r_i$ represents the envelope radius of the component.

The maximum number of iterations for the convergence setting is 50 times. The target value threshold is to terminate prematurely when T_max ≤ 85 °C. The tolerance is stopped if the improvement amount of the global optimal solution for k consecutive generations is less than $\epsilon$ (for example, $\epsilon$ = 0.1 °C), then it is terminated.

4.2. Optimization Results and Discussion

After 50 iterations of the PSO algorithm, the optimal results for the model were determined, as shown in Figure 16, with results presented for background temperatures of 30 °C, 40 °C, and 45 °C. Research indicates that the working temperature of the optimized model has decreased by approximately 10 °C overall, which can be observed by comparing Figure 5f,g with Figure 16a,b. Furthermore, according to Figure 16b,c, when the background temperature is 40 °C, the highest temperature is 80.8977 °C; and when the background temperature is 45 °C, the highest temperature is 86.2335 °C. This is because the main heat-generating electronic components in the model optimized by the PSO algorithm are closer to the fan and the grille, resulting in better heat dissipation. Therefore, the operating temperature of the computer will drop. This demonstrates that the optimized model can still function normally between 40 °C and 45 °C, aligning with the optimization concept and allowing the computer to operate effectively during high summer temperatures, thereby enhancing its thermal performance and adaptability to high-temperature environments. The initial and optimized layout of the computer are shown in Figure 17, and the optimized changes in the positions of electronic components are listed in Table 6, with the midpoint of each electronic component used as the representative position. The electronic components circled in red in Figure 17b are those which have changed positions after optimization using the PSO algorithm.

Table 7. Temperature changes before and after optimization.

Background Temperature	Optimized Working Temperature Before Optimization (Max)	Optimized Working Temperature (Max)
30 °C	80.3918 °C	70.2208 °C
40 °C	91.3424 °C	80.8977 °C
50 °C	102.313 °C	91.5669 °C

shows the comparison of the maximum operating temperatures before and after optimization.

It should be noted that the large transformer underwent a 90° rotation around the z-axis passing through the point (11, 8.5). Since the midpoint does not change after rotation, there is no intuitive representation of this change. Research indicates that adjusting the positions of electronic components can effectively reduce working temperatures to adapt to high-temperature working environments. This approach differs from the previous studies, which often involved selecting higher-performance materials for optimization. By focusing on component placement, the cost of optimization is reduced, and the efficiency is increased, providing a valuable reference for future research in this area.

5. Conclusions

This study, through finite element simulation, established a model of the computer chassis and investigated its thermal performance, using the PSO algorithm to optimize the structure, leading to the following conclusions.

(1)

By the finite element simulation method, the computer mainframe box simulation model was established. Meanwhile, the influences of temperature, velocity field, pressure, wall resolution and temperature of the fluid flow on the thermal performance were investigated, which indicated that the temperature is the most important factor.

(2)

Three major factors, which include the background temperature, heat dissipation rate, and thermal conductivity, were selected as the primary influences on temperature distribution. A series of temperature distribution diagrams were obtained through the parametric scanning function of finite element simulation, and the phenomena observed in these diagrams were explained and discussed. The study found that background temperature is the most significant factor. At a background temperature of 35 °C, the operating temperature is around 86 °C, exceeding the normal operating temperature range for electronic components, which is the reason for subsequent optimization. Among the seven heat dissipation rates, the transistor heat dissipation rate has the greatest impact. Among the nine thermal conductivities, the heat sink’s thermal conductivity is the most influential. The highest temperature generally appears at the transistor core, while the lowest temperature is at the air intake grille.

(3)

Understanding that the operating temperature range for electronic components is −40 °C to 85 °C, the PSO algorithm was used in conjunction with finite element simulation software for joint simulation. Without affecting the positions of electronic components relative to each other, the positions of the components were adjusted to optimize the base model, resulting in an overall decrease in working temperature of approximately 10 °C. Originally, at a background temperature of 35 °C, the working temperature was 86 °C, but after optimization, it reached 86 °C only at a background temperature of 45 °C. This ensures that the computer can function normally during high summer temperatures, enhancing its practicality.

(4)

Surprisingly, under certain conditions, the large transformer’s temperature can exceed that of the transistor, making it the main heat source. This indicates that even though the transistor is the most sensitive to temperature, once it is not the main heat source, the highest temperature is not at the transistor but at the electronic component that replaces the transistor as the main heat source. However, in most cases, the highest temperature is always at the transistor.

In the future, specific engineering experiments can be conducted to assist simulation and make the results more convincing. The PSO algorithm can also be optimized to obtain more accurate results.