An edge computing box with built-in machine learning algorithms

Abstract

This utility model discloses an edge computing box with a built-in machine learning algorithm, belonging to the field of edge computing box technology. It aims to solve the technical problems of low efficiency and easy local overheating caused by traditional natural air cooling. The box includes a fixed frame and a PCB board mounted on top of it. A heat dissipation mechanism is provided inside the fixed frame, and an air guide mechanism is provided on the front of the fixed frame to direct airflow to the heat dissipation mechanism. This utility model increases the heat dissipation area by using corrugated protrusions on the outside of the heat dissipation fins, which helps to better exchange heat with the air. Its streamlined, spindle-shaped contour creates a suction effect on the base of the fins near the heat source. A spiral guide plate located at the air inlet of the heat dissipation duct accelerates the airflow speed, increasing the air volume passing through the heat dissipation duct per unit time, thus solving the problems of low efficiency and easy local overheating caused by traditional natural air cooling.

Description

Technical Field

[0001] This utility model relates to the field of edge computing box technology, and more specifically, to an edge computing box with built-in machine learning algorithms. Background Technology

[0002] In today's rapidly evolving technological landscape, edge computing, with its unique advantages, has demonstrated extremely broad application prospects in numerous fields. Whether it's intelligent monitoring and real-time data analysis in industrial production, traffic flow management and environmental monitoring in smart city construction, or remote diagnosis and medical image analysis in the medical field, edge computing plays an indispensable role and has become a key force driving the intelligent upgrading of various industries.

[0003] Edge computing boxes, as core devices that support these complex and critical computing tasks, integrate numerous high-performance electronic components, such as CPUs, GPUs, and power chips. During continuous operation, these components act like small heat sources, constantly generating significant amounts of heat. This heat accumulates as the device runs for longer periods.

[0004] Traditional heat dissipation methods typically rely on simple natural air cooling, depending solely on the heat exchange between the device's own heat sink fins and the outside air. This method is extremely slow and struggles to keep up with the rate at which electronic components generate heat. Furthermore, the heat sink's relatively simple and flat structure limits its surface area, making it difficult to dissipate heat quickly and prone to localized overheating. This prevents the edge computing box from maintaining stable and reliable operation over extended periods, frequently leading to various malfunctions due to overheating issues. Therefore, we propose an edge computing box with built-in machine learning algorithms. Utility Model Content

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology, adapt to the needs of reality, and provide an edge computing box with built-in machine learning algorithms to solve the technical problems of low heat dissipation efficiency and easy local overheating caused by traditional natural air cooling.

[0006] To solve the above-mentioned technical problems, this utility model provides the following technical solution: an edge computing box with built-in machine learning algorithm, including a fixed frame and a PCB board mounted on its top, a heat dissipation mechanism is provided inside the fixed frame, and an air guide mechanism for guiding airflow to the heat dissipation mechanism is provided on the front of the fixed frame, the heat dissipation mechanism includes a plurality of parallel heat dissipation fins, the surface of the heat dissipation fins in contact with the airflow is provided with corrugated protrusions, the heat dissipation fins are disposed inside the fixed frame, and a heat dissipation air channel is formed between adjacent heat dissipation fins, the cross-section of the heat dissipation fins has a streamlined spindle-shaped profile, so that the cross-section of the heat dissipation air channel forms a Venturi tube structure, and the air guide mechanism includes a spiral guide plate, the spiral guide plate is disposed at the air inlet position of the heat dissipation air channel.

[0007] Preferably, the heat dissipation mechanism further includes a mesh back plate, which is fixedly installed on the top of the fixed frame. The mesh back plate has several threaded mounting holes on its surface. At least one heat-conducting column is threadedly installed inside the threaded mounting holes. A heat-conducting plate is fixedly installed on the top of the heat-conducting column. The heat dissipation fins are fixedly installed on the bottom of the mesh back plate.

[0008] Preferably, the air guiding mechanism further includes several air inlet pipes, which are fixedly installed on the front of the fixed frame, the spiral guide plate is fixedly installed inside the air inlet pipes, and a conical head is fixedly installed on the front of the air inlet pipes.

[0009] Preferably, the heat-conducting column has a cylindrical structure with a threaded section in its middle, which mates with the internal thread of the threaded mounting hole.

[0010] Preferably, the mesh back plate is a plate-like structure with mesh-like ribs, and the threaded mounting holes are located at the intersection nodes of the ribs.

[0011] Preferably, the number of air inlet pipes is the same as the number of heat dissipation ducts, and the positions of the air inlet pipes correspond to the positions of the heat dissipation ducts.

[0012] Preferably, it also includes an edge computing box, the fixing frame is disposed inside the edge computing box, and a dustproof net is provided on the side wall of the edge computing box at the position corresponding to the air guide mechanism.

[0013] Compared with the prior art, the beneficial effects of this utility model are: 1. This utility model increases the heat dissipation area by using corrugated protrusions on the outside of the heat dissipation fins, which helps to better exchange heat with the air. Its streamlined spindle-shaped contour makes the cross-section of the heat dissipation airflow channel formed between adjacent heat dissipation fins resemble a Venturi tube structure. When the airflow passes through the narrow area, the flow velocity increases, forming a low-pressure zone, which creates a suction effect on the base of the fins near the heat source. The spiral guide plate set at the air inlet of the heat dissipation airflow channel can accelerate the airflow speed and increase the air volume passing through the heat dissipation airflow channel per unit time, solving the problems of low heat dissipation efficiency and easy local overheating caused by traditional natural air cooling.

[0014] 2. This utility model also features pre-set threaded mounting holes on the grid back plate, allowing the heat-conducting pillars to be screwed into the corresponding positions as independent heat dissipation modules as needed. This enables flexible allocation of heat dissipation resources. The heat-conducting plate can make close contact with the heat-concentrated areas of the PCB board for heat conduction. Subsequently, the heat is conducted sequentially through the heat-conducting plate and the heat-conducting pillars to the grid back plate, achieving rapid heat conduction in both the longitudinal and transverse directions. This allows the heat in the heat-concentrated areas to be effectively dispersed to the surrounding low-temperature areas, improving the uniformity and efficiency of overall heat dissipation and preventing local heat accumulation. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the external structure of the present utility model; Figure 2 This is a cross-sectional structural diagram of the present invention; Figure 3 This is an exploded structural diagram of the PCB board and its heat dissipation mechanism of this utility model. Figure 4 This is an exploded view of the heat dissipation mechanism of this utility model; Figure 5 This is a schematic diagram of the heat dissipation fins and related structures of the present invention; Figure 6 This is a schematic diagram of the air guide mechanism of this utility model.

[0016] The following are the labels in the diagram: 1. Edge computing box; 11. Dustproof mesh; 2. Fixing frame; 3. PCB board; 4. Heat dissipation mechanism; 41. Mesh backplate; 411. Threaded mounting hole; 42. Heat conduction plate; 43. Heat conduction column; 431. Threaded section; 44. Heat dissipation fins; 441. Corrugated bump; 45. Heat dissipation duct; 5. Air guide mechanism; 51. Air inlet pipe; 52. Spiral guide plate; 53. Conical head. Detailed Implementation

[0017] Example: Figures 1 to 6As shown, this utility model relates to an edge computing box with a built-in machine learning algorithm, including a fixed frame 2 and a PCB board 3 mounted on top of it. A heat dissipation mechanism 4 is provided inside the fixed frame 2, and an air guide mechanism 5 for directing airflow to the heat dissipation mechanism 4 is provided on the front of the fixed frame 2. It also includes an edge computing box body 1, with the fixed frame 2 disposed inside the edge computing box body 1. A dustproof mesh 11 is provided on the side wall of the edge computing box body 1 at the position corresponding to the air guide mechanism 5. The heat dissipation mechanism 4 includes multiple parallelly arranged heat dissipation fins 44. The surface in contact with the airflow is provided with corrugated protrusions 441. The heat dissipation fins 44 are disposed inside the fixed frame 2, and a heat dissipation airflow channel 45 is formed between adjacent heat dissipation fins 44. The cross-section of the heat dissipation fins 44 has a streamlined spindle-shaped profile, which makes the cross-section of the heat dissipation airflow channel 45 form a Venturi tube structure. The air guiding mechanism 5 includes a spiral guide plate 52, which is disposed at the air inlet position of the heat dissipation airflow channel 45. This utility model increases the heat dissipation area by providing corrugated protrusions 441 on the outside of the heat dissipation fins 44, which helps to better exchange heat with the air. Its streamlined spindle-shaped profile makes the cross-section of the heat dissipation airflow channel 45 formed between adjacent heat dissipation fins 44 form a Venturi tube structure. When the air passes through the narrow area, the flow velocity increases and a low-pressure area is formed, which produces a suction effect on the root of the fins near the heat source. The spiral guide plate 52 disposed at the air inlet position of the heat dissipation airflow channel 45 can accelerate the airflow speed and increase the air volume passing through the heat dissipation airflow channel 45 per unit time, solving the problems of low heat dissipation efficiency and easy local overheating of traditional natural air cooling.

[0018] Furthermore, such as Figures 2 to 5 As shown, the heat dissipation mechanism 4 also includes a grid back plate 41, which is fixedly installed on the top of the fixed frame 2. The grid back plate 41 is a plate-shaped structure with grid-like ribs. Threaded mounting holes 411 are set at the intersection nodes of the ribs. Several threaded mounting holes 411 are opened on the surface of the grid back plate 41. At least one heat-conducting column 43 is threadedly installed inside the threaded mounting hole 411. The heat-conducting column 43 is a cylindrical structure with a threaded section 431 in the middle, which cooperates with the internal thread of the threaded mounting hole 411. A heat-conducting plate 42 is fixedly installed on the top of the heat-conducting column 43. The heat dissipation fins 44 are fixedly installed on the bottom of the grid back plate 41. Furthermore, by pre-setting threaded mounting holes 411 on the grid back plate 41, the heat-conducting column 43 can be screwed into the corresponding position as an independent heat dissipation module as needed, so as to realize the flexible allocation of heat dissipation resources. The heat-conducting plate 42 can make close contact with the heat-concentrated area of ​​the PCB board 3 to conduct heat. Then, the heat is conducted to the grid back plate 41 through the heat-conducting plate 42 and the heat-conducting column 43 in sequence, realizing the rapid conduction of heat in the longitudinal and lateral directions. This allows the heat in the heat-concentrated area to be effectively dispersed to the surrounding low-temperature area, improving the uniformity and efficiency of the overall heat dissipation and avoiding local heat accumulation.

[0019] Furthermore, such as Figure 6As shown, the air guiding mechanism 5 also includes several air inlet pipes 51. The air inlet pipes 51 are fixedly installed on the front of the fixed frame 2, and the spiral guide plate 52 is fixedly installed inside the air inlet pipes 51. A conical head 53 is fixedly installed on the front of the air inlet pipes 51. The number of air inlet pipes 51 is the same as the number of heat dissipation air ducts 45, and the position of the air inlet pipes 51 corresponds to the position of the heat dissipation air ducts 45. The conical head 53 is located at the front end of the air inlet pipe 51. It has a tapered structure, which can capture and accelerate the external wind, increase the air volume entering the air inlet pipe 51, and provide a sufficient source of cold air for subsequent heat dissipation.

[0020] It should be noted that the algorithm integrated into the edge computing box PCB board 3 uses the actual computing power of the CPU and GPU on the PCB board 3 as the core constraint. It reduces hardware resource consumption through two key optimization methods: model pruning and parameter quantization. Model pruning reduces the amount of computation by removing redundant convolutional layers, fully connected layers, and invalid neurons in the deep learning model. For example, in a CNN model used for device status recognition, removing network branches that contribute less than 5% to feature extraction reduces the model's computation by more than 40% while ensuring that the recognition accuracy decreases by no more than 3%. Parameter quantization compresses traditional 32-bit floating-point parameters into 16-bit integer parameters, reducing memory usage while improving the data reading and processing speed of the CPU and GPU, and avoiding hardware overload caused by excessively large parameter sizes. Through the above optimizations, the algorithm can maintain an image analysis speed of 15-20 frames per second or a processing efficiency of 300+ sets of device parameters per second under the limited hardware resources of the edge computing box, and the computing power utilization rate of CPU and GPU is stably controlled at 60%-70%, which will not cause abnormal heat generation of components due to excessive occupation of hardware resources, thus reducing the load pressure on the heat dissipation system from the source.

[0021] Working principle: This embodiment provides an edge computing box with built-in machine learning algorithm. In use, the PCB board 3 is installed on the top of the fixed frame 2. At this time, first observe the positions on the PCB board that are prone to heat generation, such as the CPU, GPU and power chip. The corresponding number and specifications of heat conduction pillars 43 are used as independent heat dissipation modules and screwed into the preset threaded mounting holes 411 on the mesh back plate 41 to realize the on-demand allocation of heat dissipation resources. The mesh back plate 41 is located at the bottom of the PCB board 3. At this time, the heat conduction plate 42 conducts heat to the heat concentration area of ​​the PCB board 3. The heat conduction plate 42 transfers heat to the heat conduction pillars 43, and then the heat conduction pillars 43 conduct heat to the mesh back plate 41, so that the heat is quickly conducted vertically and horizontally, so that the heat in the heat concentration area is transferred to the surrounding low temperature area. The heat dissipation fins 44 are used to dissipate the heat.

[0022] The heat-conducting column 43 is threaded inside the threaded mounting hole 411 and passes through the threaded mounting hole 411. The heat-conducting column 43 is located inside the heat dissipation air duct 45 between the two heat dissipation fins 44. When the outside wind enters the edge computing box 1 through the dustproof net 11, the conical head 53 captures the outside wind and increases the air volume entering the air inlet pipe 51. Because the conical head 53 is a shrinking type, it can accelerate the wind. After the wind enters the air inlet pipe 51, it passes through the spiral guide plate 52 to increase the air velocity, thereby cooling the heat dissipation fins 44.

[0023] Because the heat dissipation fins 44 have a streamlined spindle-shaped profile, and the heat dissipation airflow 45 becomes narrow in the central thickness area of ​​the spindle-shaped heat dissipation fins 44, the airflow speed increases after passing through this narrow area, forming a low-pressure area. This will create a certain suction effect on the area near the heat source at the root of the fins, ensuring that the cooling airflow can effectively cover the entire heat dissipation surface and avoid local overheating.

[0024] The embodiments disclosed herein are preferred embodiments, but are not limited thereto. Those skilled in the art can readily grasp the spirit of this utility model based on the above embodiments and make different extensions and variations. However, as long as they do not depart from the spirit of this utility model, they are all within the protection scope of this utility model.

Claims

1. An edge computing box with built-in machine learning algorithms, comprising a fixed frame (2) and a PCB board (3) mounted on top thereof, characterized in that, The fixed frame (2) is provided with a heat dissipation mechanism (4) inside, and the fixed frame (2) is provided with an air guide mechanism (5) on the front for guiding airflow to the heat dissipation mechanism (4). The heat dissipation mechanism (4) includes a plurality of parallel heat dissipation fins (44). The surface of the heat dissipation fins (44) that comes into contact with the airflow is provided with corrugated protrusions (441). The heat dissipation fins (44) are disposed inside the fixed frame (2). A heat dissipation air duct (45) is formed between adjacent heat dissipation fins (44). The cross section of the heat dissipation fins (44) has a streamlined spindle-shaped profile, so that the cross section of the heat dissipation air duct (45) forms a Venturi tube structure. The air guiding mechanism (5) includes a spiral guide plate (52), which is located at the air inlet of the heat dissipation duct (45).

2. The edge computing box with a built-in machine learning algorithm according to claim 1, characterized in that, The heat dissipation mechanism (4) also includes a mesh back plate (41), which is fixedly installed on the top of the fixed frame (2). The mesh back plate (41) has several threaded mounting holes (411) on its surface. At least one heat-conducting column (43) is threadedly installed inside the threaded mounting holes (411), and a heat-conducting plate (42) is fixedly installed on the top of the heat-conducting column (43). The heat dissipation fins (44) are fixedly installed at the bottom of the mesh backplate (41).

3. The edge computing box with a built-in machine learning algorithm according to claim 2, characterized in that, The air guide mechanism (5) also includes several air inlet pipes (51), which are fixedly installed on the front of the fixed frame (2). The spiral guide plate (52) is fixedly installed inside the air inlet pipe (51), and a conical head (53) is fixedly installed on the front of the air inlet pipe (51).

4. The edge computing box with a built-in machine learning algorithm according to claim 3, characterized in that, The heat-conducting column (43) is a cylindrical structure with a threaded section (431) in the middle, which mates with the internal thread of the threaded mounting hole (411).

5. An edge computing box with a built-in machine learning algorithm according to claim 3, characterized in that, The mesh back plate (41) is a plate-shaped structure with mesh-like ribs, and the threaded mounting holes (411) are located at the intersection nodes of the ribs.

6. An edge computing box with a built-in machine learning algorithm according to claim 3, characterized in that, The number of air inlet pipes (51) is the same as the number of heat dissipation ducts (45), and the position of the air inlet pipes (51) corresponds to the position of the heat dissipation ducts (45).

7. An edge computing box with a built-in machine learning algorithm according to claim 3, characterized in that, It also includes an edge computing box (1), the fixed frame (2) is set inside the edge computing box (1), and a dustproof net (11) is set on the side wall of the edge computing box (1) at the position corresponding to the air guide mechanism (5).

Images