A frame circuit breaker housing and manufacturing process thereof

Abstract

The invention discloses a frame circuit breaker shell and a manufacturing process thereof, comprising: a first shell, a second shell being arranged on the side of the first shell, and a third shell being arranged on the side of the second shell; in the invention, a clamping protrusion, a clamping top block and a third supporting spring are used to realize firm fixation between the shells, so that the disassembly and assembly of the shells are more convenient and quick, and the fixing effect is reliable; by loosening and tightening the fixing screw, the position of the extrusion plate and the extrusion cushion can be easily adjusted, so that the firm fixation of circuit breakers of different sizes can be easily realized; the design of the first limiting slide bar and the second supporting spring ensures that the extrusion plate applies force evenly, and prevents the circuit breaker from being damaged; through the heat dissipation fan, the heat dissipation hole and the supporting baffle, the air can flow smoothly in the space at the rear side of the first shell, the second shell and the third shell, so that the heat dissipation efficiency is greatly improved and the service life of the equipment is extended.

Description

Technical Field

[0001] The invention relates to the technical field of frame circuit breaker housing equipment, in particular to a frame circuit breaker housing and a manufacturing process thereof. Background Art

[0002] Traditional circuit breaker frame shells are usually of fixed size and cannot be adjusted according to actual needs to adapt to different numbers of circuit breakers, which limits its scope of application. If the number of circuit breakers used is adjusted, the corresponding frame shell needs to be purchased again. At the same time, if the size of the replaced circuit breaker changes slightly, the entire frame shell needs to be replaced, which wastes resources and increases the user's maintenance cost. Therefore, a frame circuit breaker shell and its manufacturing process are provided to solve the above problems. Summary of the invention

[0003] The purpose of the present invention is to provide a frame circuit breaker housing and a manufacturing process thereof in order to solve the problems raised by the above-mentioned background technology, which has the advantages that the housing replacement and disassembly are very convenient and quick, the fixing effect is reliable, the assembly length is adjustable, and the circuit breakers of different sizes can be firmly fixed easily, thus solving the problems raised by the above-mentioned background technology.

[0004] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a frame circuit breaker housing, comprising: a first housing, a second housing is arranged on the side of the first housing, a third housing is arranged on the side of the second housing, circuit breakers are arranged inside the first housing, the second housing and the third housing, fixed rotating bars are fixedly connected to the sides of the first housing, the fixed rotating bars are rotatably connected to a transparent protective window, an extrusion top block is fixedly connected to the side of the transparent protective window, a first card slot is opened on the side of the first housing, the second housing and the third housing, a first support spring is fixedly connected inside the first card slot, an arc-shaped top block is fixedly connected to the side of the first support spring, two groups of splicing protrusions are fixedly connected to the sides of the first housing and the second housing, and the first housing The first shell and the second shell are fixedly connected with four groups of clamping protrusions on their sides, the first shell and the third shell are each provided with two groups of side wiring holes on their sides, the first shell, the second shell and the third shell are each provided with heat dissipation holes and rear wiring holes on their sides, the first shell, the second shell and the third shell are each fixedly connected with two groups of support baffles on their inner sides, the first shell and the third shell are each fixedly connected with two groups of cooling fans on their sides, the first shell, the second shell and the third shell are each provided with two groups of fixing plate slide grooves on their inner sides, a sliding fixing mechanism is provided on the fixing plate slide grooves, the second shell and the third shell are each provided with two groups of splicing grooves on their sides, four groups of clamping grooves are each provided on their sides, a sliding bar groove is provided below the clamping grooves, and a clamping fixing mechanism is provided below the clamping grooves;

[0005] The clamping and fixing mechanism includes a lifting handle, two groups of second limiting sliding bars are fixedly connected above the lifting handle, a clamping top block is fixedly connected above the second limiting sliding bar, and a third supporting spring is sleeved on the outer side of the second limiting sliding bar.

[0006] As a further solution of the present invention: the sliding and fixing mechanism includes a push plate, a fixed block is fixedly connected to the side of the push plate, a threaded hole is opened above the fixed block and extends to the bottom thereof, a fixing screw is threadedly connected to the inside of the fixed block, a soft gasket is fixedly connected to the bottom of the fixing screw, an extrusion plate is provided on the side of the push plate, an extrusion pad is fixedly connected to the side of the extrusion plate, two groups of limiting sliding blocks are fixedly connected to the bottom of the extrusion plate, four groups of first limiting sliding rods are fixedly connected to the side of the extrusion plate, and a second supporting spring is sleeved on the outer side of the first limiting sliding rod.

[0007] As a further solution of the present invention: the number of the second shells can be selected as needed, the number of the arc-shaped top blocks is set in two groups, and they are symmetrically arranged inside the first card slot, the splicing groove is adapted to the splicing protrusion, and the card groove is adapted to the position of the card protrusion.

[0008] As a further solution of the present invention: the position of the extrusion top block corresponds to the first card slot. When the transparent protection window is rotated downward to a suitable position, the extrusion top block can be inserted into the first card slot. Two groups of arc-shaped top blocks support the arc-shaped notches on both sides of the extrusion top block, thereby fixing the position of the transparent protection window. When the fixation needs to be released, the transparent protection window only needs to be rotated upward by applying force.

[0009] As a further solution of the present invention: the space at the rear side of the first shell, the second shell and the third shell is separated by the support baffle to facilitate air circulation, thereby enhancing the heat dissipation effect; the wind directions of the four groups of cooling fans are consistent, so that air can circulate in the space at the rear side of the first shell, the second shell and the third shell.

[0010] As a further solution of the present invention: the slide bar groove correspondingly penetrates to the bottom of the first shell, the second shell or the third shell.

[0011] As a further solution of the present invention: the upper end of the third support spring is fixedly connected to the clamping top block, the lower end of the third support spring is fixedly connected to the clamping groove, the clamping top block is adapted to the clamping protrusion, and when the clamping protrusion is pushed into the clamping groove, the clamping protrusion squeezes the clamping top block, and the third support spring is compressed. When it is pushed into the deepest part, the third support spring is reset, and the clamping top block clamps the clamping protrusion to complete the fixation.

[0012] As a further solution of the present invention: the limit sliding block is adapted to the sliding groove of the fixed plate, a through hole adapted to the first limit sliding rod is opened on the side of the push plate, one end of the first limit sliding rod is fixed to the extrusion plate, and the other end is fixed to a limit block to prevent it from sliding out of the through hole opened on the side of the push plate, one end of the second support spring is fixed to the side of the extrusion plate, and the other end is fixed to the side of the push plate.

[0013] A manufacturing process for a frame circuit breaker housing comprises the following steps:

[0014] Step 1: The user first selects the number of second shells according to the number of circuit breakers required, and splices them with other shells, so as to realize flexible fixation of three or more groups of circuit breakers. The first shell, the second shell and the third shell are preliminarily positioned by splicing protrusions and splicing grooves. By using a clamping fixing mechanism, when the clamping protrusion is pushed into the clamping groove, the clamping protrusion will squeeze the clamping top block, causing the third support spring to be compressed. When the clamping protrusion is pushed into the deepest position, the third support spring is reset, and the clamping top block clamps the clamping protrusion, thereby realizing a firm fixation between the shells;

[0015] Step 2: The transparent protection window is connected to the housing through a fixed rotation bar and can be rotated downward. When the transparent protection window is rotated downward to the bottom, the extrusion top block will be stuck in the first slot, and the two groups of arc-shaped top blocks will support the arc-shaped notches on both sides of the extrusion top block, thereby fixing the position of the transparent protection window. When it is necessary to release the fixation, it is only necessary to apply force to rotate the transparent protection window upward. At this time, the extrusion top block squeezes the arc-shaped top block so that the first support spring is compressed, allowing the arc-shaped top block to release the squeezing of the arc-shaped notch of the extrusion top block, so that the transparent protection window can rotate freely.

[0016] Step 3: The sliding fixing mechanism is used to fix the circuit breaker. First, by loosening the fixing screw, the push plate can be pushed to slide in the fixing plate slide groove. Then, the movement of the push plate will drive the extrusion plate and the extrusion cushion to move toward the circuit breaker, thereby fixing it in a proper position. The first limit slide bar and the second support spring ensure that the extrusion plate can move smoothly and exert proper pressure to prevent the circuit breaker from being damaged during the fixing process. When the circuit breaker is fixed to a suitable position, tighten the fixing screw to fix the position of the sliding fixing mechanism and complete the fixing of the circuit breaker.

[0017] Step 4: The cooling fans are installed on the sides of the first shell and the third shell. Their wind directions are consistent, so that air can circulate in the space on the rear side of the first shell, the second shell and the third shell, thereby enhancing the heat dissipation effect. The support baffle separates the internal space to ensure that the air can circulate smoothly, further improving the heat dissipation efficiency. The heat dissipation holes allow air circulation to help reduce the temperature of the circuit breaker and other internal components.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] 1. The present invention adopts a clamping fixing mechanism, through the interaction between the clamping protrusion and the clamping top block, combined with the reset function of the third support spring, to achieve a firm fixation between the shells. This design is not only very convenient and quick to replace and disassemble, but also has a reliable fixing effect.

[0020] 2. In the present invention, the positions of the extrusion plate and the extrusion cushion can be easily adjusted by loosening and tightening the fixing screw, thereby easily achieving stable fixation of circuit breakers of different sizes. The design of the first limiting slide bar and the second supporting spring ensures the stability of the extrusion plate during movement and the uniformity of applied pressure, effectively preventing the circuit breaker from being damaged during the fixing process.

[0021] 3. In the present invention, the cooling fan and the cooling holes are used in coordination, and the supporting baffles are reasonably arranged, so that the air can flow smoothly in the space at the rear side of the first shell, the second shell and the third shell, thereby greatly improving the heat dissipation efficiency, reducing the temperature of the circuit breaker and other internal components, and extending the service life of the equipment.

[0022] 4. Introduce convolutional neural network to analyze the temperature distribution data inside the shell, optimize the working strategy and wind direction adjustment of the cooling fan to achieve the best heat dissipation effect.

[0023] 5. Through the deep learning capabilities of convolutional neural networks, intelligent optimization of the heat dissipation system is achieved, significantly improving heat dissipation efficiency and ensuring that circuit breakers and other components operate within an appropriate temperature range. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the overall structure of the present invention;

[0025] Figure 2 This is a schematic diagram of the opening structure of the integral transparent protective window in the present invention;

[0026] Figure 3 It is a schematic diagram of the structure inside the first shell in the present invention;

[0027] Figure 4 It is a schematic diagram of the structure inside the second shell in the present invention;

[0028] Figure 5 is a schematic diagram of the structure of the interior of the second shell from a second viewing angle in the present invention;

[0029] Figure 6 is a schematic diagram of the structure of the interior of the second shell in the present invention from a third viewing angle;

[0030] Figure 7 It is a schematic diagram of the structure inside the third shell in the present invention;

[0031] Figure 8 is a schematic diagram of the structure of the interior of the third shell from a second viewing angle in the present invention;

[0032] Fig. 9 It is a structural schematic diagram of the clamping and fixing mechanism in the present invention;

[0033] Fig.10 It is a structural schematic diagram of the sliding fixing mechanism in the present invention;

[0034] Fig.11 It is a structural schematic diagram of the arc-shaped top block in the present invention;

[0035] Fig.12 It is a flow chart of the intelligent algorithm in the present invention.

[0036] In the figure: 1, first housing; 2, second housing; 3, third housing; 4, circuit breaker; 5, fixed rotating bar; 6, transparent protection window; 7, first card slot; 8, first support spring; 9, arc top block; 10, extrusion top block; 11, splicing protrusion; 12, card protrusion; 13, side wiring hole; 14, heat dissipation hole; 15, rear wiring hole; 16, support baffle; 17, heat dissipation fan; 18, fixed plate slide; 19, sliding fixing mechanism; 1 91. Push plate; 192. Fixed block; 193. Fixed screw; 194. Soft gasket; 195. Extrusion plate; 196. Extrusion cushion; 197. Limit slider; 198. First limit slide bar; 199. Second support spring; 20. Splicing groove; 21. Snap-in groove; 22. Slide bar groove; 23. Snap-in fixing mechanism; 231. Lifting handle; 232. Second limit slide bar; 233. Snap-in top block; 234. Third support spring. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, which are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", and "set" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or it can be the internal connection of two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. The following is an explanation of an embodiment of the present invention based on the overall structure of the present invention.

[0039] Reference Figures 1 to 11In an embodiment of the present invention, a frame circuit breaker housing comprises: a first housing 1, a second housing 2 is arranged on the side of the first housing 1, a third housing 3 is arranged on the side of the second housing 2, a circuit breaker 4 is arranged inside the first housing 1, the second housing 2 and the third housing 3, and the number of the second housings 2 can be selected as needed. The first housing 1, the second housing 2 and the third housing 3 are all fixedly connected to the sides with fixed rotating bars 5, the fixed rotating bars 5 are rotatably connected to a transparent protective window 6, and the side of the transparent protective window 6 is fixedly connected to an extrusion top block 10, the first housing 1, the second housing 2 and the third housing 3 are all opened with first card slots 7, and the first card slots 7 are fixedly connected to the inside of the first supporting spring 8, The first supporting spring 8 is fixedly connected with an arc-shaped top block 9 on the side, and the arc-shaped top block 9 is provided with two groups, which are symmetrically arranged inside the first card slot 7. The position of the extrusion top block 10 corresponds to the first card slot 7. When the transparent protection window 6 is rotated downward to a suitable position, the extrusion top block 10 can be inserted into the first card slot 7. The two groups of arc-shaped top blocks 9 support the arc-shaped notches on both sides of the extrusion top block 10, so that the position of the transparent protection window 6 can be fixed. When the fixation needs to be released, the transparent protection window 6 only needs to be rotated upward by applying force. The first shell 1 and the second shell 2 are both fixedly connected with two groups of splicing protrusions 11 on the sides, and the first shell 1 and the second shell 2 are both fixedly connected with four groups of snap-on protrusions 12 on the sides. The first shell 1 and the third shell 3 are each provided with two groups of side wiring holes 13 on their sides, the first shell 1, the second shell 2 and the third shell 3 are each provided with heat dissipation holes 14 and rear wiring holes 15 on their sides, the first shell 1, the second shell 2 and the third shell 3 are each fixedly connected with two groups of support baffles 16 on their inner sides, the first shell 1 and the third shell 3 are each fixedly connected with two groups of heat dissipation fans 17 on their sides, the support baffles 16 are used to separate the space at the rear side of the first shell 1, the second shell 2 and the third shell 3 for air circulation, and the heat dissipation effect is enhanced, the wind directions of the four groups of heat dissipation fans 17 are consistent, so that the air can circulate in the space at the rear side of the first shell 1, the second shell 2 and the third shell 3, the first shell 1, the second shell 2 and the third shell 3 are Two groups of fixing plate slide grooves 18 are provided on the inner side of the second shell 2 and the third shell 3, and a sliding fixing mechanism 19 is provided on the fixing plate slide grooves 18. Two groups of splicing grooves 20 are provided on the side of the second shell 2 and the third shell 3, and the splicing grooves 20 are adapted to the splicing protrusions 11 to assist the connection between the first shell 1, the second shell 2 and the third shell 3. Four groups of clamping grooves 21 are provided on the side of the second shell 2 and the third shell 3, and a slide bar groove 22 is provided below the clamping grooves 21. The slide bar grooves 22 pass through the bottom of the first shell 1, the second shell 2 or the third shell 3 accordingly. A clamping fixing mechanism 23 is provided below the clamping grooves 21, and the clamping grooves 21 are adapted to the position of the clamping protrusions 12;

[0040] The clamping and fixing mechanism 23 includes a lifting handle 231, two groups of second limiting sliding bars 232 are fixedly connected above the lifting handle 231, and a clamping top block 233 is fixedly connected above the second limiting sliding bar 232. A third supporting spring 234 is sleeved on the outer side of the second limiting sliding bar 232, and the upper end of the third supporting spring 234 is fixedly connected to the clamping top block 233, and the lower end of the third supporting spring 234 is fixedly connected to the clamping groove 21. The clamping top block 233 is adapted to the clamping protrusion 12. When the clamping protrusion 12 is pushed into the clamping groove 21, the clamping protrusion 12 squeezes the clamping top block 233, and the third supporting spring 234 is compressed. When it is pushed into the deepest part, the third supporting spring 234 is reset, and the clamping top block 233 clamps the clamping protrusion 12 to complete the fixation. Such a design can be used for the connection and fixation between the first shell 1 and the second shell 2, the second shell 2 and the second shell 2, and the second shell 2 and the third shell 3.

[0041] The above scheme is adopted: the first shell 1, the second shell 2 and the third shell 3 are preliminarily positioned by the splicing protrusion 11 and the splicing groove 20, and a clamping fixing mechanism 23 is used to further fix these shells. When the clamping protrusion 12 is pushed into the clamping groove 21, the clamping protrusion 12 will squeeze the clamping top block 233, so that the third support spring 234 is compressed. When the clamping protrusion 12 is pushed into the deepest position, the third support spring 234 is reset, and the clamping top block 233 clamps the clamping protrusion 12, thereby achieving a firm fixation between the shells. Through the splicing design, the number of second shells 2 can be selected, and they can be spliced ​​with the first shell 1 and the third shell 3, so as to achieve flexible fixation of three or more groups of circuit breakers 4;

[0042] The transparent protective window 6 is connected to the shell through a fixed rotating bar 5. When the transparent protective window 6 is rotated downward to the bottom, the extrusion top block 10 will be stuck in the first slot 7, and the two groups of arc-shaped top blocks 9 will support the arc-shaped notches on both sides of the extrusion top block 10, thereby fixing the position of the transparent protective window 6. When it is necessary to release the fixation, it is only necessary to apply force to rotate the transparent protective window 6 upward.

[0043] Reference Fig.10The sliding fixing mechanism 19 includes a push plate 191, a fixed block 192 is fixedly connected to the side of the push plate 191, a threaded hole is opened on the top of the fixed block 192 and penetrates to the bottom thereof, a fixing screw 193 is threadedly connected to the inside of the fixed block 192, a soft gasket 194 is fixedly connected to the bottom of the fixing screw 193, an extrusion plate 195 is arranged on the side of the push plate 191, an extrusion soft pad 196 is fixedly connected to the side of the extrusion plate 195, and two groups of limit sliders 197 are fixedly connected to the bottom of the extrusion plate 195, and the limit sliders 197 slide with the fixed plate The first limiting slide bar 198 is adapted to the groove 18, and four groups of first limiting slide bars 198 are fixedly connected to the side of the extrusion plate 195. The side of the push plate 191 is provided with a through hole adapted to the first limiting slide bar 198. One end of the first limiting slide bar 198 is fixed to the extrusion plate 195, and the other end is fixed with a limiting block to prevent it from sliding out of the through hole opened on the side of the push plate 191. A second supporting spring 199 is sleeved on the outer side of the first limiting slide bar 198, and one end of the second supporting spring 199 is fixed to the side of the extrusion plate 195, and the other end is fixed to the side of the push plate 191.

[0044] The above scheme is adopted: the sliding fixing mechanism 19 is used to fix the circuit breaker 4. By loosening the fixing screw 193, the push plate 191 can be pushed to move, and then the circuit breaker 4 is fixed in an appropriate position through the extrusion plate 195 and the extrusion cushion 196. The first limiting slide bar 198 and the second supporting spring 199 ensure that the extrusion plate 195 can move smoothly and apply appropriate pressure. When fixed to the appropriate position, tightening the fixing screw 193 can fix the position of the sliding fixing mechanism 19 and complete the fixation of the circuit breaker 4.

[0045] Example 1

[0046] A manufacturing process for a frame circuit breaker housing comprises the following steps:

[0047] Step 1: The user first selects the number of second housings 2 according to the number of circuit breakers 4, and splices them with other housings, so as to realize flexible fixation of three or more groups of circuit breakers 4. The first housing 1, the second housing 2 and the third housing 3 are preliminarily positioned by the splicing protrusions 11 and the splicing grooves 20. By using the clamping fixing mechanism 23, when the clamping protrusion 12 is pushed into the clamping groove 21, the clamping protrusion 12 will squeeze the clamping top block 233, causing the third support spring 234 to be compressed. When the clamping protrusion 12 is pushed into the deepest position, the third support spring 234 is reset, and the clamping top block 233 clamps the clamping protrusion 12, thereby realizing the firm fixation between the housings;

[0048] Step 2: The transparent protection window 6 is connected to the housing through the fixed rotation bar 5 and can be rotated downward. When the transparent protection window 6 is rotated downward to the bottom, the extrusion top block 10 will be stuck in the first card slot 7, and the two groups of arc-shaped top blocks 9 will support the arc-shaped notches on both sides of the extrusion top block 10, thereby fixing the position of the transparent protection window 6. When it is necessary to release the fixation, it is only necessary to apply force to rotate the transparent protection window 6 upward. At this time, the extrusion of the arc-shaped top block 9 by the extrusion top block 10 compresses the first support spring 8, allowing the arc-shaped top block 9 to release the extrusion of the arc-shaped notch of the extrusion top block 10, so that the transparent protection window 6 can rotate freely;

[0049] Step 3: The sliding fixing mechanism 19 is used to fix the circuit breaker 4. First, by loosening the fixing screw 193, the push plate 191 can be pushed to slide in the fixing plate slide groove 18. Then, the movement of the push plate 191 will drive the extrusion plate 195 and the extrusion cushion 196 to move toward the circuit breaker 4, thereby fixing it in a proper position. The first limit slide bar 198 and the second support spring 199 ensure that the extrusion plate 195 can move smoothly and apply proper pressure to prevent the circuit breaker 4 from being damaged during the fixing process. When the circuit breaker 4 is fixed to a proper position, tightening the fixing screw 193 can fix the position of the sliding fixing mechanism 19, thereby completing the fixing of the circuit breaker 4.

[0050] Step 4: The cooling fan 17 is installed on the side of the first shell 1 and the third shell 3, and their wind directions are consistent, so that air can circulate in the space on the rear side of the first shell 1, the second shell 2 and the third shell 3, thereby enhancing the heat dissipation effect. The support baffle 16 separates the internal space to ensure that the air can circulate smoothly, further improving the heat dissipation efficiency. The heat dissipation holes 14 allow air to circulate, helping to reduce the temperature of the circuit breaker 4 and other internal components. The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field within the technical scope disclosed by the present invention, according to the technical solution and its inventive concept of the present invention, shall be equivalently replaced or changed, and shall be covered within the protection scope of the present invention.

[0051] Example 2

[0052] Convolutional neural network is used to analyze the temperature distribution data inside the shell, optimize the working strategy and wind direction adjustment of the cooling fan to achieve the best cooling effect;

[0053] Through the deep learning capabilities of convolutional neural networks, intelligent optimization of the heat dissipation system is achieved, significantly improving heat dissipation efficiency and ensuring that circuit breakers and other components operate within an appropriate temperature range;

[0054] Step 1: Data collection and preprocessing

[0055] A temperature sensor is installed inside the shell to collect temperature data in real time; the data includes temperature readings at different time points and different locations; the collected temperature data is standardized to eliminate the differences between different sensors and time points; the processed data is converted into a format suitable for CNN input;

[0056] 1.1 Data Collection

[0057] Install temperature sensors: Install temperature sensors at several key locations inside the frame circuit breaker housing; key locations include: the top, bottom, left and right sides of the circuit breaker, the area near the cooling fan, and other components that generate higher temperatures;

[0058] Sensor parameters include sensor model and sampling frequency;

[0059] The data format has a time stamp T for each collected data. i ; Each data point needs to clearly identify the collection location P j ; The collected temperature data T ij It means that at time T i When the position P j Temperature;

[0060] 1.2 Data Preprocessing

[0061] In order to eliminate the differences between different sensors and time points and make the data more suitable for processing by convolutional neural networks (CNNs), the temperature data is converted into dimensionless data using a standardized formula.

[0062]

[0063] Where: T′ ij is the standardized temperature data, T ij is the original temperature data, μ j For position P j The mean temperature, σ j For position P j The standard deviation of temperature;

[0064] Convert the standardized data into a format suitable for CNN input; CNN requires two-dimensional or three-dimensional data input, represented here as matrices and tensors;

[0065] For a two-dimensional matrix: combine the temperature data at different time points and locations into a matrix; there are n locations and m time points, forming an m×n matrix;

[0066] For three-dimensional tensors: For data at multiple moments, multiple m×n matrices are superimposed to form a three-dimensional tensor m×n×t, where t is the length of the time dimension;

[0067] Step 2: Design and training of convolutional neural networks

[0068] Select a suitable CNN architecture, including ResNet and VGG; the CNN architecture includes an input layer, several convolutional layers, a pooling layer, a fully connected layer, and an output layer; divide the preprocessed temperature data into a training set and a test set; use the training set to train the CNN, and use the test set to evaluate the model performance; use the back propagation algorithm to optimize the network parameters by minimizing the loss function; use data enhancement technology during training to improve the generalization ability of the model;

[0069] 2.1 Choosing a suitable CNN architecture

[0070] In order to improve the performance of convolutional neural network (CNN) in analyzing the temperature distribution data inside the shell, two more complex CNN architectures were selected: ResNet (residual network) and VGG, and they were used in combination to optimize the working strategy and wind direction adjustment of the cooling fan;

[0071] VGG Architecture

[0072] Input layer: accepts the standardized temperature data matrix with a shape of m×n, where m is the number of time points and n is the number of sensor locations;

[0073] Convolutional layer: Several convolutional layers, each with a fixed-size convolution kernel (3×3), and each layer contains multiple filters to extract features at different levels;

[0074] Pooling layer: Add a maximum pooling layer (MaxPooling) after the convolution layer, with a pooling window size of 2×2, for downsampling and reducing data dimensions;

[0075] Fully connected layer: After the convolution and pooling layers, it is expanded into a one-dimensional vector and optimized through multiple fully connected layers to adjust the working strategy and wind direction of the cooling fan;

[0076] ResNet architecture

[0077] Input layer: accepts data output from the VGG network, with a shape of p×q, where p and q are the height and width of the feature map after VGG processing;

[0078] Residual modules: Multiple residual modules, each module contains two 3×3 convolutional layers and adds a skip connection at the output. The skip connection skips multiple convolutional layers and directly adds the input to the output;

[0079] Output layer: fully connected layer, used for final temperature prediction and strategy optimization;

[0080] 2.2 Training Dataset

[0081] Dataset division: training set, used for training the model, accounts for 70% of the total dataset; test set, used for evaluating model performance, accounts for 30% of the total dataset;

[0082] Data augmentation: Randomly rotate the temperature data matrix to enhance the robustness of the model; randomly translate the temperature data matrix to improve the generalization ability of the model;

[0083] 3. Training process

[0084] The loss function is Mean Squared Error (MSE);

[0085]

[0086] Where: N is the number of samples, y i is the true value, is the predicted value;

[0087] Use gradient descent to update network parameters;

[0088]

[0089] Among them: θ is the network parameter, η is the learning rate, is the loss function; Represents the loss function The gradient of the network parameter θ; it measures the rate of change of L relative to each parameter and is used to guide the direction and step of parameter update; through the gradient descent method, the network parameters are optimized to minimize the loss function and improve the prediction accuracy of the model.

[0090] 2.3 Specific implementation process of the combined architecture (Note: FC is the fully connected layer, Conv is the convolutional layer, Pooling is the pooling layer, ReLU is a type of activation function, and Dropout is an overfitting method) Data input shape: temperature data matrix X, the shape is m×n, here it is 64×64;

[0091] For VGG architecture:

[0092] The convolutional layer and pooling layer are:

[0093] Conv1→ReLU→Conv2→ReLU→MaxPooling

[0094] Among them: the convolution layer 1Conv1 has a filter size of 3×3 and a number of 64, the convolution layer 2Conv2 has a filter size of 3×3 and a number of 128, the maximum pooling layer MaxPooling has a pooling window of 2×2, and ReLU is a type of activation function;

[0095] Fully connected layer: FC1→ReLU→Dropout→FC2

[0096] Among them: the number of nodes in the fully connected layer 1 is 256, the Dropout layer randomly discards some nodes to prevent overfitting, and ReLU is a type of activation function;

[0097] For ResNet architecture:

[0098] Residual module: Conv3→ReLU→Conv4→Add→ReLU

[0099] The residual module 1 contains two convolutional layers, each with a convolution kernel size of 3×3, a number of 256, and a skip connection; the residual module 2 contains two convolutional layers, each with a convolution kernel size of 3×3, a number of 512, and a skip connection. ReLU is a type of activation function, and Add is a residual connection.

[0100] Fully connected layer: the output node is 1, which is used for temperature prediction;

[0101] The mean square error (MSE) is used as the loss function, and the gradient descent method is used in the optimization process. The initial value of the learning rate is 0.001, which is adjusted according to the training effect. At the same time, the mean square error (MSE) of the test set evaluation model is used for model evaluation.

[0102] 2.4 Introduction of Bidirectional LSTM Module and Improvement of Attention Mechanism

[0103] Adding bidirectional LSTM modules and attention mechanisms to the convolutional neural network (CNN) architecture improves the model’s ability to capture the characteristics of temperature distribution over time and focuses on key temperature change areas, thereby further optimizing the cooling fan’s operating strategy and wind direction adjustment.

[0104] 2.4.1 Bidirectional LSTM Module

[0105] Input data shape: features from the output of the VGG network Figure X , the shape is t×p×q, where tt is the time step, p and q are the height and width of the feature map;

[0106] Input conversion: Expand the feature map into a two-dimensional matrix with a shape of t×(p×q) so that it can be input into the LSTM network;

[0107] Number of LSTM neurons: Each layer contains 128 neurons;

[0108] Bidirectional LSTM: Each time step has two LSTM networks, forward and backward, to capture the bidirectional dependencies in the time series; the forward LSTM is used to process the forward information flow of the time series, and the backward LSTM7 is used to process the reverse information flow of the time series;

[0109] LSTM formula:

[0110]

[0111] and are the hidden states of the forward and backward LSTM at time step t respectively; X t is the input data of time step t; LSTM fwd For the forward LSTM module, LSTM bwd Backward LSTM module;

[0112] Output shape: Concatenate the hidden states of the forward and backward LSTM to form a new feature matrix with a shape of t×(p×q×2×128);

[0113] 2.4.2 Attention Mechanism

[0114] Input data shape: feature matrix H from bidirectional LSTM output, shape is t×(p×q×2×128);

[0115] Calculate the attention score e ij =tanh(W h h ij +b h )

[0116] where h ij is the eigenvector of position (i, j) in the feature matrix, W h and b h is a trainable parameter;

[0117] Attention weights: where α ij is the attention weight, which is used to measure the importance of each feature; ij The eigenvector h at position (i, j) in the feature matrix H ij Importance score after linear transformation and activation function; e ik The eigenvector h at position (i, k) in the feature matrix HHH ik The importance score after the same processing is used to calculate the denominator of the attention weight;

[0118] The weighted feature map is H′ ij =α ij hij ; The output data shape is the weighted feature matrix H′, which has a shape of t×(p×q×2×128);

[0119] Step 3: Model application and optimization

[0120] Deploy the trained CNN model to the actual system to analyze the temperature distribution data inside the shell in real time and predict the temperature distribution trend. According to the temperature distribution prediction results output by CNN, dynamically adjust the working strategy of the cooling fan, including the fan speed and wind direction, to achieve the best cooling effect.

[0121] 3.1 Model Deployment

[0122] Select a suitable hardware platform for model deployment. Here, we use the embedded system NVIDIA Jetson Nano. Install the necessary deep learning framework. Here, we use PyTorch to run the trained CNN model. Configure the temperature sensor data interface to transmit the real-time collected temperature data to the computing platform to ensure the stability and real-time performance of data transmission.

[0123] 3.2 Temperature distribution prediction

[0124] Use the CNN model to infer the input temperature data and predict the temperature distribution trend at the next moment; the model outputs the predicted temperature distribution matrix The shape is m×t, where m is the time dimension and t is the sensor position dimension;

[0125] 3.3 Strategy Optimization

[0126] Use the fan policy adjustment algorithm, where the input parameters are:

[0127] Temperature distribution matrix Temperature distribution trend predicted by the CNN model,

[0128] Current fan status S f : Including the fan speed v f and wind direction f ,

[0129] Temperature threshold T th : The preset temperature threshold, if it exceeds this value, the heat dissipation needs to be strengthened.

[0130] The output parameters are:

[0131] Optimized fan speed v′ f , optimized fan wind direction d′ f ;

[0132] Temperature threshold detection: traverse the predicted temperature distribution matrix Check whether the temperature exceeds the threshold value T th location,

[0133] High temperature locations: Records all locations that exceed the temperature threshold

[0134] Fan speed adjustment: High temperature position ratio is the ratio of the calculated high temperature position The speed adjustment formula dynamically adjusts the fan speed according to the high temperature position ratio;

[0135] v′ f =v f ×(1+k v ·rhot)

[0136] where k v is the speed adjustment coefficient, which indicates the influence of high temperature ratio on fan speed;

[0137] Wind direction adjustment: The center of high temperature position is the center of gravity of the high temperature position.

[0138]

[0139] For the wind direction adjustment formula, there is d′ f =adjust_direction(i c ,j c ) that is, adjusting the fan wind direction according to the center of gravity of the high temperature position, wherein the function adjust_direction is used to calculate the optimal wind direction according to the center of gravity of the high temperature position; i c is the average position of the high temperature position in the row direction, that is, the center of gravity of the high temperature area in the row direction. c is the average position of the high temperature position in the column direction, that is, the center of gravity of the high temperature area in the column direction

[0140] 3.3 Strategy Execution and Feedback

[0141] The optimized fan speed v′ f and wind direction d′ f Applied to the actual system, adjust the working status of the cooling fan; monitor the temperature changes inside the shell in real time to ensure that the cooling effect reaches the expected goal; feed back the new temperature data to the CNN model to continuously optimize the temperature distribution prediction and fan strategy adjustment;

[0142] 3.4 Specific implementation cases

[0143] Eight temperature sensors are installed in the housing of a frame circuit breaker to collect temperature data in real time. The CNN model is used to predict the temperature distribution trend and dynamically adjust the working strategy of the cooling fan.

[0144] Use NVIDIA Jetson Nano to deploy the model; install the PyTorch framework; configure the temperature sensor to connect to Jetson Nano through the I2C interface; set the sampling frequency to 1 time per second, standardize the temperature data and convert it into an 8×8 temperature matrix to input into the CNN model;

[0145] The current temperature distribution matrix is Temperature threshold T th =50°C, current fan speed v f =1000RPM, wind direction d f =0; traverse 3 locations were found to exceed the temperature threshold T th , record the high temperature position P hot ; Therefore, the proportion of high temperature position Speed ​​adjustment factor k v =2; the adjusted fan speed can be obtained

[0146] v′ f =1000×(1+2×0.047)=1000×1.094=1094RPM; then calculate the center of gravity of the high temperature position (i c ,j c )=(4,4), adjust the wind direction d′ f =adjust_direction(4,4)=45°; therefore, the fan speed is adjusted to 1094RPM and the wind direction is adjusted to 45 degrees; the temperature data is continuously monitored to ensure the heat dissipation effect, and the data is fed back to the CNN model to optimize the prediction effect.

Claims

1. A frame circuit breaker housing, characterized in that: include: A first shell (1), a second shell (2) is arranged on the side of the first shell (1), a third shell (3) is arranged on the side of the second shell (2), a circuit breaker (4) is arranged inside the first shell (1), the second shell (2) and the third shell (3), a fixed rotating bar (5) is fixedly connected to the side of the first shell (1), the second shell (2) and the third shell (3), a transparent protective window (6) is rotatably connected to the fixed rotating bar (5), an extrusion top block (10) is fixedly connected to the side of the transparent protective window (6), a first clamping groove (7) is opened on the side of the first shell (1), the second shell (2) and the third shell (3), a first support spring (8) is fixedly connected inside the first clamping groove (7), an arc-shaped top block (9) is fixedly connected to the side of the first support spring (8), two groups of splicing protrusions (11) are fixedly connected to the side of the first shell (1) and the second shell (2), four groups of clamping protrusions (12) are fixedly connected to the side of the first shell (1) and the second shell (2), The first shell (1) and the third shell (3) are each provided with two groups of side wiring holes (13) on their sides; the first shell (1), the second shell (2) and the third shell (3) are each provided with heat dissipation holes (14) and rear wiring holes (15) on their sides; the first shell (1), the second shell (2) and the third shell (3) are each provided with two groups of support baffles (16) fixedly connected to their inner sides; the first shell (1) and the third shell (3) are each provided with two groups of heat dissipation fans (17) fixedly connected to their inner sides; Two groups of fixed plate slide grooves (18) are provided on the inner sides of the second shell (2) and the third shell (3), and a sliding fixing mechanism (19) is provided on the fixed plate slide grooves (18); two groups of splicing grooves (20) are provided on the side surfaces of the second shell (2) and the third shell (3); four groups of clamping grooves (21) are provided on the side surfaces of the second shell (2) and the third shell (3), and a slide bar groove (22) is provided below the clamping groove (21); and a clamping fixing mechanism (23) is provided below the clamping groove (21); The clamping and fixing mechanism (23) comprises a lifting handle (231), two groups of second limiting sliding bars (232) are fixedly connected above the lifting handle (231), a clamping top block (233) is fixedly connected above the second limiting sliding bars (232), and a third supporting spring (234) is sleeved on the outer side of the second limiting sliding bars (232); The sliding fixing mechanism (19) comprises a push plate (191), a fixed block (192) is fixedly connected to the side of the push plate (191), a threaded hole is opened on the top of the fixed block (192) and penetrates to the bottom thereof, a fixing screw (193) is threadedly connected to the inside of the fixed block, a soft gasket (194) is fixedly connected to the bottom of the fixing screw (193), an extrusion plate (195) is arranged on the side of the push plate (191), an extrusion soft pad (196) is fixedly connected to the side of the extrusion plate (195), two groups of limiting sliding blocks (197) are fixedly connected to the bottom of the extrusion plate (195), four groups of first limiting sliding bars (198) are fixedly connected to the side of the extrusion plate (195), and a second supporting spring (199) is sleeved on the outer side of the first limiting sliding bar (198); The upper end of the third support spring (234) is fixedly connected to the clamping top block (233), and the lower end of the third support spring (234) is fixedly connected to the clamping groove (21). The clamping top block (233) is adapted to the clamping protrusion (12). When the clamping protrusion (12) is pushed into the clamping groove (21), the clamping protrusion (12) presses the clamping top block (233), and the third support spring (234) is compressed. When the third support spring (234) is pushed into the deepest position, the third support spring (234) is reset, and the clamping top block (233) clamps the clamping protrusion (12), thereby completing the fixation. The limiting slider (197) is compatible with the fixed plate slide groove (18), and a through hole compatible with the first limiting slider (198) is opened on the side of the push plate (191). One end of the first limiting slider (198) is fixed to the extrusion plate (195), and the other end is fixed to a limiting block to prevent it from sliding out of the through hole opened on the side of the push plate (191). One end of the second support spring (199) is fixed to the side of the extrusion plate (195), and the other end is fixed to the side of the push plate (191).

2. A frame circuit breaker housing according to claim 1, characterized in that: The number of the second shells (2) can be selected as required, the number of the arc-shaped top blocks (9) is provided in two groups, and they are symmetrically arranged inside the first clamping groove (7), the splicing groove (20) is adapted to the splicing protrusion (11), and the clamping groove (21) is adapted to the position of the clamping protrusion (12).

3. A frame circuit breaker housing according to claim 1, characterized in that: The position of the extrusion top block (10) corresponds to the first card slot (7). When the transparent protection window (6) is rotated downward to a suitable position, the extrusion top block (10) can be inserted into the first card slot (7). The two groups of arc-shaped top blocks (9) support the arc-shaped notches on both sides of the extrusion top block (10), thereby fixing the position of the transparent protection window (6). When the fixation needs to be released, the transparent protection window (6) only needs to be rotated upward by applying force.

4. A frame circuit breaker housing according to claim 1, characterized in that: The support baffle (16) is used to separate the spaces at the rear sides of the first shell (1), the second shell (2) and the third shell (3) to facilitate air circulation and enhance the heat dissipation effect. The wind directions of the four groups of heat dissipation fans (17) are consistent so that air can circulate in the spaces at the rear sides of the first shell (1), the second shell (2) and the third shell (3).

5. A frame circuit breaker housing according to claim 1, characterized in that: The slide bar groove (22) correspondingly penetrates to the bottom of the first shell (1), the second shell (2) or the third shell (3).

6. A manufacturing process for a frame circuit breaker housing according to claim 1, characterized in that: The following steps are involved: Step 1: The user first selects the number of second housings (2) according to the number of circuit breakers (4) required, and splices them with other housings, thereby realizing flexible fixation of three or more groups of circuit breakers (4). The first housing (1), the second housing (2) and the third housing (3) are preliminarily positioned by means of splicing protrusions (11) and splicing grooves (20). By using a snap-fit ​​fixing mechanism (23), when the snap-fit ​​protrusion (12) is pushed into the snap-fit ​​groove (21), the snap-fit ​​protrusion (12) will squeeze the snap-fit ​​top block (233), causing the third support spring (234) to be compressed. When the snap-fit ​​protrusion (12) is pushed into the deepest position, the third support spring (234) is reset, and the snap-fit ​​top block (233) clamps the snap-fit ​​protrusion (12), thereby realizing firm fixation between the housings. Step 2: The transparent protection window (6) is connected to the shell body through the fixed rotation bar (5) and can be rotated downward. When the transparent protection window (6) is rotated downward to the bottom, the extrusion top block (10) will be inserted into the first slot (7), and the two groups of arc-shaped top blocks (9) will support the arc-shaped notches on both sides of the extrusion top block (10), thereby fixing the position of the transparent protection window (6). When it is necessary to release the fixation, it is only necessary to apply force to rotate the transparent protection window (6) upward. At this time, the extrusion top block (10) squeezes the arc-shaped top block (9) so that the first support spring (8) is compressed, allowing the arc-shaped top block (9) to release the squeezing of the arc-shaped notch of the extrusion top block (10), so that the transparent protection window (6) can rotate freely. Step 3: The sliding fixing mechanism (19) is used to fix the circuit breaker (4). First, by loosening the fixing screw (193), the push plate (191) can be pushed to slide in the fixing plate slide groove (18). Then, the movement of the push plate (191) will drive the extrusion plate (195) and the extrusion cushion (196) to move in the direction of the circuit breaker (4), thereby fixing it in a suitable position. The first limit slide bar (198) and the second support spring (199) ensure that the extrusion plate (195) can move smoothly and apply suitable pressure to prevent the circuit breaker (4) from being damaged during the fixing process. When the circuit breaker (4) is fixed to a suitable position, the fixing screw (193) can be tightened to fix the position of the sliding fixing mechanism (19), thereby completing the fixing of the circuit breaker (4); Step 4: The cooling fans (17) are installed on the sides of the first shell (1) and the third shell (3). Their wind directions are consistent, so that air can circulate in the space on the rear side of the first shell (1), the second shell (2) and the third shell (3), thereby enhancing the heat dissipation effect. The support baffle (16) separates the internal space to ensure that the air can circulate smoothly, further improving the heat dissipation efficiency. The heat dissipation holes (14) allow air to circulate, helping to reduce the temperature of the circuit breaker (4) and other internal components.

7. A manufacturing process for a frame circuit breaker housing according to claim 6, characterized in that: In step 4, a convolutional neural network is used to analyze the temperature distribution data inside the housing to optimize the working strategy and wind direction adjustment of the cooling fan. The specific steps are as follows: Step 1: Data collection and preprocessing A temperature sensor is installed inside the shell to collect temperature data in real time; the data includes temperature readings at different time points and different locations; the collected temperature data is standardized to eliminate the differences between different sensors and time points; the processed data is converted into a format suitable for CNN input; 1.1 Data Collection Install temperature sensors: Install temperature sensors at several key locations inside the frame circuit breaker housing; Key locations include: the top, bottom, left and right sides of the circuit breaker, the area near the cooling fan, and other components that generate higher temperatures; Sensor parameters include sensor model and sampling frequency; The data format has a time stamp T for each collected data. i ; Each data point needs to clearly identify the collection location P j ; The collected temperature data T ij It means that at time T i When the position P j Temperature; 1.2 Data Preprocessing In order to eliminate the differences between different sensors and time points and make the data more suitable for processing by the convolutional neural network CNN; the temperature data is converted into dimensionless data using a standardized formula; Where: T′ ij is the standardized temperature data, T ij is the original temperature data, μ j For position P j The mean temperature, σ j For position P j The standard deviation of temperature; Convert the standardized data into a format suitable for CNN input; CNN requires two-dimensional or three-dimensional data input, represented here as matrices and tensors; For a two-dimensional matrix: combine the temperature data at different time points and locations into a matrix; there are n locations and m time points, forming an m×n matrix; For three-dimensional tensors: For data at multiple moments, multiple m×n matrices are superimposed to form a three-dimensional tensor m×n×t, where t is the length of the time dimension; Step 2: Design and training of convolutional neural networks 2.1 Choosing a suitable CNN architecture In order to improve the performance of convolutional neural network (CNN) in analyzing the temperature distribution data inside the shell, two more complex CNN architectures are selected: residual network ResNet and VGG network, and they are used in combination to optimize the working strategy and wind direction adjustment of the cooling fan; VGG Architecture Input layer: accepts the standardized temperature data matrix with a shape of m×n, where m is the number of time points and n is the number of sensor locations; Convolutional layer: Several convolutional layers, each with a fixed-size convolution kernel of 3×3. Each layer contains multiple filters to extract features at different levels. Pooling layer: Add the maximum pooling layer MaxPool ing after the convolution layer, with a pooling window size of 2×2, which is used for downsampling and reducing the data dimension; Fully connected layer: After the convolution and pooling layers, it is expanded into a one-dimensional vector and optimized through multiple fully connected layers to adjust the working strategy and wind direction of the cooling fan; ResNet architecture Input layer: accepts data output from the VGG network, with a shape of p×q, where p and q are the height and width of the feature map after VGG processing; Residual modules: Multiple residual modules, each module contains two 3×3 convolutional layers and adds a skip connection at the output. The skip connection skips multiple convolutional layers and directly adds the input to the output; Output layer: fully connected layer, used for final temperature prediction and strategy optimization; 2.2 Training Dataset Dataset division: training set, used for training models, accounts for 70% of the total dataset; The test set is used to evaluate the model performance, accounting for 30% of the total dataset; Data augmentation: Randomly rotate the temperature data matrix to enhance the robustness of the model; randomly translate the temperature data matrix to improve the generalization ability of the model; 2.

3. Training process The loss function is mean square error MSE; Where: N is the number of samples, y i is the true value, is the predicted value; Use gradient descent method to update network parameters; Among them: θ is the network parameter, η is the learning rate, is the loss function; Represents the loss function The gradient of the network parameter θ; it measures The rate of change of each parameter is used to guide the direction and step of parameter update; the network parameters are optimized through the gradient descent method to minimize the loss function and improve the prediction accuracy of the model; 2.4 Specific implementation process of the combined architecture: FC is a fully connected layer, Conv is a convolutional layer, Pooling is a pooling layer, ReLU is a type of activation function, Dropout is an overfitting method, and Add is a residual connection; Data input shape: temperature data matrix X, shape is m×n, here it is 64×64; For VGG architecture: The convolutional layer and pooling layer are: Conv1→ReLU→Conv2→ReLU→MaxPooling Among them: the convolution layer 1Conv1 has a filter size of 3×3 and a number of 64, the convolution layer 2Conv2 has a filter size of 3×3 and a number of 128, the maximum pooling layer MaxPooling has a pooling window of 2×2, and ReLU is a type of activation function; Fully connected layer: FC1→ReLU→Dropout→FC2 Among them: there are two fully connected layers FC1 and FC2, with 256 nodes. The Dropout layer randomly discards some nodes to prevent overfitting, and ReLU is a type of activation function; For ResNet architecture: Residual module: Conv3→ReLU→Conv4→Add→ReLU Among them, the architecture has two residual modules. Residual module 1 contains two convolutional layers, each layer has a convolution kernel size of 3×3, a number of 256, and a jump connection; residual module 2 contains two convolutional layers, each layer has a convolution kernel size of 3×3, a number of 512, a jump connection, ReLU is a type of activation function, and Add is a residual connection; Fully connected layer: the output node is 1, which is used for temperature prediction; The mean square error (MSE) is used as the loss function, and the optimization process uses the gradient descent method. The initial value of the learning rate is 0.001, which is adjusted according to the training effect. At the same time, the mean square error (MSE) of the test set evaluation model is used for model evaluation. 2.5 Introduction of Bidirectional LSTM Module and Improvement of Attention Mechanism Add bidirectional LSTM modules and attention mechanisms to the convolutional neural network (CNN) architecture to improve the model's ability to capture the characteristics of temperature distribution over time and focus on key temperature change areas, thereby further optimizing the cooling fan's working strategy and wind direction adjustment. 2.5.1 Bidirectional LSTM Module Input data shape: feature map X output from the VGG network, with a shape of t×p×q, where t is the time step, p and q are the height and width of the feature map; Input conversion: Expand the feature map into a two-dimensional matrix with a shape of t×(p×q) so that it can be input into the LSTM network; Number of LSTM neurons: Each layer contains 128 neurons; Bidirectional LSTM: Each time step has two LSTM networks, forward and backward, to capture the bidirectional dependencies in the time series; the forward LSTM is used to process the forward information flow of the time series, and the backward LSTM7 is used to process the reverse information flow of the time series; LSTM formula: and are the hidden states of the forward and backward LSTM at time step t respectively; X t is the input data of time step t; LSTM fwd For the forward LSTM module, LSTM bwd Backward LSTM module; Output shape: Concatenate the hidden states of the forward and backward LSTM to form a new feature matrix with a shape of t×(p×q×2×128); 2.5.2 Attention Mechanism Input data shape: feature matrix H from bidirectional LSTM output, shape: t×(p×q×2×128); Calculate the attention score e ij =tanh(W h h ij +b h ) where h ij is the eigenvector of position (i, j) in the feature matrix, W h and b h is a trainable parameter; Attention weights: where α ij is the attention weight, which is used to measure the importance of each feature; ij The eigenvector h at position (i, j) in the feature matrix H ij Importance score after linear transformation and activation function; e ik The eigenvector h at position (i, k) in the feature matrix H ik The importance score after the same processing is used to calculate the denominator of the attention weight; The weighted feature map is H′ ij =α ij h ij ; The output data shape is the weighted feature matrix H′, which has a shape of t×(p×q×2×128); Step 3: Model application and optimization Deploy the trained CNN model to the actual system to analyze the temperature distribution data inside the shell in real time and predict the temperature distribution trend. According to the temperature distribution prediction results output by CNN, dynamically adjust the working strategy of the cooling fan, including the fan speed and wind direction, to achieve the best cooling effect. 3.1 Model Deployment Select a suitable hardware platform for model deployment. Here, we use the embedded system NVIDIA Jetson Nano. Install the necessary deep learning framework. Here, we use PyTorch to run the trained CNN model. Configure the temperature sensor data interface to transmit the real-time collected temperature data to the computing platform to ensure the stability and real-time performance of data transmission. 3.2 Temperature distribution prediction Use the CNN model to infer the input temperature data and predict the temperature distribution trend at the next moment; the model outputs the predicted temperature distribution matrix The shape is m×t, where m is the time dimension and t is the sensor position dimension; 3.3 Strategy Optimization Use the fan policy adjustment algorithm, where the input parameters are: Temperature distribution matrix Temperature distribution trend predicted by the CNN model, Current fan status S f : Including the fan speed v f and wind direction f , Temperature threshold T th : The preset temperature threshold, if it exceeds this value, the heat dissipation needs to be strengthened. The output parameters are: Optimized fan speed v′ f , optimized fan wind direction d′ f ; Temperature threshold detection: traverse the predicted temperature distribution matrix Check whether the temperature exceeds the threshold value T th location, High temperature locations: Records all locations that exceed the temperature threshold i and j are the horizontal and vertical coordinates respectively; Fan speed adjustment: The high temperature position ratio is the ratio of the calculated high temperature position. The speed adjustment formula dynamically adjusts the fan speed according to the high temperature position ratio; v′ f =v f ×(1+k v ·r hot ) where k v is the speed adjustment coefficient, which indicates the influence of high temperature ratio on fan speed; Wind direction adjustment: The center of high temperature position is the center of gravity of the high temperature position; For the wind direction adjustment formula, there is d′ f =adjust_direction(i c ,j c ) that is, adjusting the fan wind direction according to the center of gravity of the high temperature position, wherein the function adjust_direction is used to calculate the optimal wind direction according to the center of gravity of the high temperature position; i c is the average position of the high temperature position in the row direction, that is, the center of gravity of the high temperature area in the row direction, j c is the average position of the high temperature position in the column direction, that is, the center of gravity of the high temperature area in the column direction; 3.4 Strategy Execution and Feedback The optimized fan speed v′ f and wind direction d′ f Applied to the actual system, it adjusts the working status of the cooling fan; monitors the temperature changes inside the shell in real time to ensure that the cooling effect reaches the expected target; feeds the new temperature data back to the CNN model to continuously optimize the temperature distribution prediction and fan strategy adjustment.

Images