A heat supply system fault diagnosis method based on artificial intelligence technology and internet of things technology

Abstract

The application relates to a heating system fault diagnosis method based on artificial intelligence technology and Internet of Things technology and relates to the field of heating system fault diagnosis. The application is aimed at solving the problem of low fault detection accuracy of the existing heating system fault detection method. The application comprises the following steps: inputting a to-be-diagnosed heating equipment image into a heating equipment recognition model to obtain a to-be-heated equipment target detection frame position and a heating equipment type; obtaining a to-be-diagnosed heating equipment thermal image; mapping the heating equipment position in the to-be-diagnosed heating equipment image to the to-be-diagnosed heating equipment thermal image to obtain a target detection frame center coordinate in the thermal image, a ratio of the target detection frame width in the thermal image to the overall width of the thermal image and a ratio of the target detection frame height in the thermal image to the overall height of the thermal image; then, the average temperature in the range of the target detection frame in the thermal image is used to judge whether the to-be-diagnosed heating equipment is abnormal; if the to-be-diagnosed heating equipment is abnormal, an alarm is given; otherwise, the detection is ended. The application is used for heating system fault detection.

Description

Technical Field

[0001] This invention relates to the field of heating system fault diagnosis, and in particular to a heating system fault diagnosis method based on artificial intelligence and Internet of Things technologies. Background Technology

[0002] Global climate change and the energy crisis have urged countries to accelerate their energy transition, reduce dependence on fossil fuels, and promote the application of sustainable energy. To achieve these goals, countries have implemented energy-saving measures in various sectors, including industry, construction, and transportation. Heating systems, as a crucial component of energy consumption, are particularly important in cold northern regions. Monitoring their safe and normal operation and preventing energy waste has become a critical energy-saving task. Heating systems consist of various devices, each playing a vital role in the heating process. Core equipment such as boilers, radiators, heat exchangers, and circulating pumps ensure the generation and transfer of heat, while control equipment such as thermostats and pressure controllers ensure efficient and stable operation under different conditions. Heating systems are highly interconnected; any change in any equipment can cause fluctuations in the overall system, especially temperature changes. Both excessively high and low temperatures negatively impact system operation, safety, energy efficiency, and user comfort. Excessively high temperatures can lead to equipment damage, energy waste, and safety hazards, while excessively low temperatures can cause insufficient heating, system icing, and corrosion. Therefore, maintaining the heating system temperature within a reasonable range not only ensures the comfort of the living environment but also improves system efficiency, prevents energy waste, and extends equipment life.

[0003] Currently, fault detection in heating systems mainly relies on real-time monitoring by sensors. However, these sensors require regular maintenance and calibration, and changes in the environment can also cause them to become inaccurate, resulting in low accuracy in fault detection for heating systems. Summary of the Invention

[0004] The purpose of this invention is to address the problem of low accuracy in existing fault detection methods for heating systems, and to propose a fault diagnosis method for heating systems based on artificial intelligence and Internet of Things technologies.

[0005] The specific process of a fault diagnosis method for heating systems based on artificial intelligence and Internet of Things technologies is as follows:

[0006] S1. Obtain the image of the heating equipment to be diagnosed, input the image of the heating equipment to be diagnosed into the heating equipment recognition model, and obtain the position of the target detection box of the heating equipment and the type of the heating equipment in the image of the heating equipment to be diagnosed.

[0007] S2. Use an infrared detector to scan the heating equipment to be diagnosed and obtain a thermal image of the heating equipment to be diagnosed.

[0008] S3. Map the position of the target detection box of the heating equipment in the image of the heating equipment to be diagnosed obtained in S1 to the thermal image of the heating equipment to be diagnosed obtained in S3, and obtain the center coordinates of the target detection box in the thermal image, the ratio of the width of the target detection box in the thermal image to the overall width of the thermal image, and the ratio of the height of the target detection box in the thermal image to the overall height of the thermal image, thereby obtaining the range of the target detection box in the thermal image.

[0009] S4. Use the average temperature within the target detection box in the thermal image to determine whether the heating equipment to be diagnosed is abnormal. If an abnormality is found, an alarm will be triggered; otherwise, the detection will end directly.

[0010] Furthermore, the heating equipment identification model in S1 is obtained in the following way:

[0011] Step 1: Obtain raw images of the heating system equipment and preprocess them. Use the preprocessed raw images to form a dataset, and then divide the dataset into training, validation, and test sets.

[0012] Step 11: Obtain original images of the heating system equipment;

[0013] Step 1 and Step 2: Preprocess the original images of the heating system equipment to obtain preprocessed original images of the heating system equipment;

[0014] Step 13: Obtain the target detection box position in the preprocessed original image of the heating system equipment, use the heating system equipment type as the label of the preprocessed original image of the heating system equipment, and combine the preprocessed original image of the heating system equipment, the preprocessed original image of the heating system equipment with the target detection box position, and the label to form a dataset, and divide the dataset into training set, validation set and test set.

[0015] The types of equipment in the heating system include: boiler images, radiator images, heat exchanger images, and circulating pump images;

[0016] Step 2: Train the device recognition network using the training set, and verify and test the trained device recognition network using the test set and validation set to finally obtain the heating equipment recognition model.

[0017] Furthermore, the preprocessing of the original images of the heating system equipment in steps one and two specifically includes:

[0018] First, data augmentation is performed on the original images of the heating system equipment;

[0019] Then, bilinear interpolation is used to scale the original image of the heating system equipment after data augmentation to obtain the preprocessed original image of the heating system equipment.

[0020] Furthermore, the device identification network includes: a first convolutional module, a second convolutional module, a first C2f_add module, a second C2f_add module, a third convolutional module, a third C2f_add module, a fourth C2f_add module, a fifth C2f_add module, a sixth C2f_add module, a first SCDown module, a seventh C2f_add module, an eighth C2f_add module, a ninth C2f_add module, a tenth C2f_add module, a second SCDown module, a third SCDown module, a fourth SCDown module, a fifth SCDown module, and a sixth SCDown module. Modules: 7th SCDown module, 8th SCDown module, 1st C2fCIB_add module, SPPF module, PSA module, 1st Upsample module, 1st splicing module, 1st C2f module, 2nd Upsample module, 2nd splicing module, 2nd C2f module, 3rd C2f module, 4th convolution module, 3rd splicing module, 2nd C2fCIB_add module, 3rd C2fCIB_add module, 9th SCDown module, 4th splicing module, 3rd C2fCIB_add module, 4th C2fCIB_add module, Header module, Result output module;

[0021] The first convolutional module includes: a first convolutional layer, a first batch of normalization layers, and a first activation function layer; the input to the first convolutional module is the preprocessed original image of the heating system equipment;

[0022] The first convolutional layer is a 3×3 kernel layer; the input to the first convolutional layer is the preprocessed original image of the heating system equipment.

[0023] The input to the first batch of normalization layers is the output of the first convolutional layer;

[0024] The first activation function layer is the Silu activation function, and the input of the first activation function layer is the output of the first batch of normalization layers;

[0025] The second convolutional module includes: a second convolutional layer, a second batch normalization layer, and a second activation function layer; the input of the second convolutional module is the output of the first convolutional module.

[0026] The second convolutional layer is a convolutional layer with a kernel size of 3×3;

[0027] The input to the second batch of normalized layers is the output of the second convolutional layer;

[0028] The second activation function layer is the Silu activation function, and the input of the second activation function layer is the output of the second batch normalization layer;

[0029] The input to the first C2f_add module is the output of the second convolutional module;

[0030] The input to the second C2f_add module is the output of the first C2f_add module;

[0031] The third convolutional module includes: a third convolutional layer, a third batch normalization layer, and a third activation function layer; the input of the third convolutional module is the output of the second C2f_add module;

[0032] The third convolutional layer is a 3×3 kernel layer; the input to the third convolutional layer is the output of the second C2f_add module.

[0033] The input to the third batch of normalization layers is the output of the third convolutional layer;

[0034] The third activation function layer is the Silu activation function, and the input of the third activation function layer is the output of the third batch normalization layer;

[0035] The input to the third C2f_add module is the output of the 3 convolutional module;

[0036] The input to the fourth C2f_add module is the output of the third C2f_add module;

[0037] The input to the fifth C2f_add module is the output of the fourth C2f_add module;

[0038] The input to the sixth C2f_add module is the output of the fifth C2f_add module;

[0039] The input to the first SCDown module is the output of the sixth C2f_add module;

[0040] The input of the seventh C2f_add module is the output of the first SCDown module;

[0041] The input of the eighth C2f_add module is the output of the seventh C2f_add module;

[0042] The input of the ninth C2f_add module is the output of the eighth C2f_add module;

[0043] The input of the tenth C2f_add module is the output of the ninth C2f_add module;

[0044] The input to the second SCDown module is the output of the tenth C2f_add module;

[0045] The input to the third SCDown module is the output of the second SCDown module;

[0046] The input to the fourth SCDown module is the output of the third SCDown module;

[0047] The input to the fifth SCDown module is the output of the fourth SCDown module;

[0048] The input of the sixth SCDown module is the output of the fifth SCDown module;

[0049] The input of the seventh SCDown module is the output of the sixth SCDown module;

[0050] The input of the eighth SCDown module is the output of the seventh SCDown module;

[0051] The input to the first C2fCIB_add module is the output of the eighth SCDown module;

[0052] The input to the SPPF module is the output of the first C2fCIB_add module;

[0053] The input of the PSA module is the output of the SPPF module;

[0054] The input to the first Upsample module is the output of the PSA module;

[0055] The first splicing module is used to splice the output of the first Upsample module and the output of the tenth C2f_add module to obtain the first splicing result;

[0056] The input to the first C2f module is the first splicing result;

[0057] The input to the second Upsample module is the output of the first C2f module;

[0058] The second splicing module is used to splice the output of the second Upsample module and the output of the sixth C2f_add module to obtain the second splicing result;

[0059] The input to the second C2f module is the second splicing result;

[0060] The input of the third C2f module is the output of the second C2f module; the output of the third C2f module is feature map X0;

[0061] The fourth convolutional module includes: a fourth convolutional layer, a fourth batch normalization layer, and a fourth activation function layer; the input of the fourth convolutional module is the output of the third C2f module.

[0062] The fourth convolutional layer is a 3×3 kernel layer; the input of the fourth convolutional layer is the output of the third C2f module.

[0063] The input to the fourth batch of normalized layers is the output of the fourth convolutional layer;

[0064] The fourth activation function layer is the Silu activation function, and the input of the fourth activation function layer is the output of the fourth batch normalization layer;

[0065] The third splicing module is used to splice the output of the first C2f module and the output of the fourth convolution module to obtain the third splicing result;

[0066] The input to the second C2fCIB_add module is the output of the third concatenation module;

[0067] The input of the third C2fCIB_add module is the output of the second C2fCIB_add module; the third C2fCIB_add module outputs feature map X1;

[0068] The input to the ninth SCDown module is the output of the third C2fCIB_add module;

[0069] The fourth splicing module is used to splice the output of the PSA module and the output of the ninth SCDown module to obtain the fourth splicing result;

[0070] The input to the third C2fCIB_add module is the fourth concatenation result;

[0071] The fourth C2fCIB_add module takes as input the output of the third C2fCIB_add module; the fourth C2fCIB_add module outputs feature map X2.

[0072] The input to the head module is feature map X0, feature map X1, and feature map X2 to obtain feature map X'0, feature map X'1, and feature map X'2;

[0073] The result output module is used to process feature map X0, processed feature map X1, and processed feature map X2 to obtain images of heating system equipment with target detection box positions and heating equipment types.

[0074] Furthermore, the head module includes: a feature map X0 processing unit, a feature map X1 processing unit, and a feature map X2 processing unit;

[0075] The feature map X0 processing unit includes: the 5th convolution sub-unit, the 6th convolution sub-unit, the 7th convolution sub-unit, the 8th convolution sub-unit, the 9th convolution sub-unit, the 10th convolution sub-unit, the 11th convolution sub-unit, the 12th convolution sub-unit, and the 5th splicing sub-unit; the input of the feature map X0 processing unit is the feature map X0;

[0076] The fifth convolutional subunit includes: a fifth convolutional layer, a fifth batch normalization layer, and a fifth activation function layer;

[0077] The fifth convolutional layer is a convolutional layer with a 3×3 kernel; the input of the fifth convolutional layer is the feature map X0;

[0078] The input to the fifth batch of normalization layers is the output of the fifth convolutional layer;

[0079] The fifth activation function layer is a silu activation function, and the input of the fifth activation function layer is the output of the fifth batch normalization layer;

[0080] The sixth convolutional subunit is the same as the fifth convolutional subunit; the input of the sixth convolutional subunit is the output of the fifth convolutional subunit.

[0081] The 7th convolutional sub-unit is a convolutional layer with a 1×1 kernel; the input of the 7th convolutional sub-unit is the output of the 6th convolutional sub-unit.

[0082] The 8th convolutional subunit is the same as the 5th convolutional subunit; the input of the 8th convolutional subunit is the feature map X1;

[0083] The 9th convolutional subunit includes: a 9th convolutional layer, a 9th batch normalization layer, and a 9th activation function layer; the input of the 9th convolutional subunit is the output of the 8th convolutional subunit.

[0084] The ninth convolutional layer is a convolutional layer with a 1×1 kernel; the input of the ninth convolutional layer is the output of the 8th convolutional sub-unit.

[0085] The input to the ninth batch of normalized layers is the output of the ninth convolutional layer;

[0086] The ninth activation function layer is a silu activation function, and the input of the ninth activation function layer is the output of the ninth batch normalization layer.

[0087] The 10th convolutional subunit is the same as the 5th convolutional subunit; the input of the 10th convolutional subunit is the output of the 9th convolutional subunit;

[0088] The 11th convolutional subunit is the same as the 9th convolutional subunit; the input of the 11th convolutional subunit is the output of the 10th convolutional subunit;

[0089] The 12th convolutional subunit is a convolutional layer with a 1×1 kernel; the input of the 12th convolutional subunit is the output of the 11th convolutional subunit.

[0090] The fifth splicing subunit is used to splice the output of the seventh convolutional subunit and the output of the 12th convolutional subunit to obtain the feature map X'0;

[0091] The structure of the feature map X1 processing unit is the same as that of the feature map X0 processing unit; the input of the feature map X1 processing unit is feature map X1, and the output is feature map X'1.

[0092] The structure of the feature map X2 processing unit is the same as that of the feature map X0 processing unit; the input of the feature map X2 processing unit is feature map X2, and the output is feature map X'2.

[0093] Furthermore, the result output module includes: an image mapping unit and a non-maximum suppression unit;

[0094] The image mapping unit is used to scale feature map X'0, feature map X'1 and feature map X'2 to the same size as the original image of the heating system equipment and then map them onto the original image of the heating system equipment to obtain the original image of the heating system equipment after mapping feature map X'0, feature map X'1 and feature map X'2;

[0095] The non-maximum suppression unit is used to perform non-maximum suppression processing on the original images of the heating system equipment after mapping feature map X'0, feature map X'1 and feature map X'2, to obtain images of the heating system equipment with target detection box positions and heating equipment types.

[0096] Furthermore, the center coordinates of the target detection box in the thermal image in S3 are obtained in the following way:

[0097]

[0098] Among them, (x heat ,y heat (x) represents the center coordinates of the target detection box in the thermal image. noraml ,y normal ) represents the center coordinates of the target detection box of the heating equipment in the image of the heating equipment to be diagnosed, W heat It is the width of the thermal image, W heat It is the height of the thermal image, W normal It is the image width of the heating equipment to be diagnosed, H normal It is the image height of the heating equipment to be diagnosed.

[0099] Furthermore, the ratio of the target detection box width to the overall width of the thermal image is obtained in the following way:

[0100]

[0101] Among them, w heat It is the ratio of the width of the target detection bounding box in the thermal image to the overall width of the thermal image, w normal It is the ratio of the width of the target detection box of the heating equipment in the image of the heating equipment to be diagnosed to the width of the image of the heating equipment to be diagnosed.

[0102] Furthermore, the ratio of the target detection box height to the overall height of the thermal image is specifically as follows:

[0103]

[0104] Among them, h heat h is the ratio of the height of the target detection bounding box in the thermal image to the overall height of the thermal image. normal It is the ratio of the height of the target detection box of the heating equipment in the image of the heating equipment to be diagnosed to the height of the image of the heating equipment to be diagnosed.

[0105] Furthermore, in step S4, the average temperature within the target detection box range of the thermal image is used to determine whether the heating equipment to be diagnosed is malfunctioning. If an malfunction is detected, an alarm is triggered; otherwise, the detection process ends directly. Specifically:

[0106] S401. Obtain the average temperature within the range of the target detection box in the thermal image, specifically:

[0107]

[0108] Among them, T aver It is the average temperature (T) within the bounding box of the target in the thermal image. j is the temperature value of the j-th pixel within the bounding box of the target in the thermal image, M is the total number of pixels within the bounding box of the target in the thermal image, and j is the pixel number within the bounding box of the target in the thermal image;

[0109] S402. Based on the type of heating equipment and the average temperature within the target detection box range in the thermal image, determine whether the heating equipment to be diagnosed is abnormal. If an abnormality is found, issue an alarm; otherwise, end the detection. Specifically:

[0110] If the heating equipment is a boiler, then when 70℃≤T aver ≤90℃ indicates that the boiler has no heating failure, and the test ends; if T aver <70℃ indicates the boiler temperature is too low, triggering a boiler temperature low alarm; if T aver If the temperature is >90℃, it indicates that the boiler temperature is too high, and a boiler overheating alarm will be issued.

[0111] If the heating equipment type is a radiator, then when 60℃≤T aver ≤80℃ indicates that the radiator has no heating failure, and the test ends; if T aver <60℃ indicates the radiator temperature is too low, triggering a low radiator temperature alarm; if T aver If the temperature is >80℃, it indicates that the radiator temperature is too high, and an overheating alarm for the radiator will be issued.

[0112] If the heating equipment is a heat exchanger, then when 50℃≤T aver ≤90℃ indicates that the heat exchanger has no heating failure, and the test ends; if T aver <50℃ indicates that the heat exchanger temperature is too low, triggering a low temperature alarm; if T aver If the temperature is >90℃, it indicates that the heat exchanger temperature is too high, and a heat exchanger over-temperature alarm will be issued.

[0113] If the heating equipment is a circulating pump, then when 30℃≤T aver ≤60℃ indicates that the circulating pump has no heating failure, and the test ends; if T aver <30℃ indicates that the circulating pump temperature is too low, and a low circulating pump temperature alarm will be issued; if T aver If the temperature is greater than 60℃, it indicates that the circulating pump temperature is too high, and an alarm for excessive circulating pump temperature will be issued.

[0114] The beneficial effects of this invention are as follows:

[0115] This invention employs a heating equipment identification model to detect heating systems, then combines this with thermal imaging to obtain the current temperature of the heating system equipment, thereby determining whether any heating equipment in the system is malfunctioning. This improves the speed and accuracy of fault detection in heating systems. This invention provides different fault warnings for four types of heating equipment in the system: boilers, radiators, heat exchangers, and circulating pumps, enabling targeted heating adjustments based on the specific faults of each equipment. After detecting a target heating equipment, this invention uses thermal imaging technology to analyze the equipment temperature in real time and provide corresponding warnings, reducing energy waste and extending equipment lifespan. This invention detects abnormalities in heating equipment, enabling timely handling or repair of these abnormalities, avoiding problems such as heating outages due to sudden equipment failures. Attached Figure Description

[0116] Figure 1 This is the overall flowchart;

[0117] Figure 2(a) shows an image of the boiler;

[0118] Figure 2(b) shows an image of the radiator;

[0119] Figure 2(c) shows an image of the heat exchanger;

[0120] Figure 2(d) shows an image of the circulating pump;

[0121] Figure 3(a) shows the structure of the C2f_add module;

[0122] Figure 3(b) is a structural diagram of the SCDown module;

[0123] Figure 3(c) shows the structure of the C2fCIB_add module;

[0124] Figure 3(d) is a structural diagram of the CIB module;

[0125] Figure 3(e) is a diagram of the SPPF module structure;

[0126] Figure 3(f) is a structural diagram of the PSA module;

[0127] Figure 4(a) shows the overall network structure.

[0128] Figure 4(b) shows the structure of the header module. Detailed Implementation

[0129] Specific implementation method one: as follows Figure 1 As shown, the specific process of a heating system fault diagnosis method based on artificial intelligence and Internet of Things technologies in this embodiment is as follows:

[0130] S1. Obtain the image of the heating equipment to be diagnosed, input the image of the heating equipment to be diagnosed into the heating equipment recognition model, and obtain the position of the target detection box of the heating equipment and the type of the heating equipment in the image of the heating equipment to be diagnosed.

[0131] The heating equipment identification model was obtained through the following methods:

[0132] Step 1: Obtain raw images of the heating system equipment and preprocess them. Use the preprocessed raw images to form a dataset, and then divide the dataset into training, validation, and test sets.

[0133] Step 11: Obtain original images of the heating system equipment:

[0134] The categories of the original images of the heating equipment include: boiler images, radiator images, heat exchanger images, and circulating pump images;

[0135] The original image resolution of the heating equipment is 1080×810. The image data comes from the School of Civil Engineering, Heilongjiang University, totaling 4000 images, with 1000 images for each type of equipment. Examples of the equipment are shown below. Figures 2(a)-2(d) As shown.

[0136] Steps 1 and 2: Preprocess the original images of the heating system equipment;

[0137] First, data augmentation is performed on the original images of the heating system equipment, specifically including: 1) random horizontal or vertical flipping, 2) random cropping, and 3) Mosaic data augmentation.

[0138] The data augmentation method used in this step reduces the risk of overfitting in the subsequent network.

[0139] Then, bilinear interpolation is used to scale the original image of the heating system equipment after data augmentation to obtain the preprocessed original image of the heating system equipment.

[0140] The original image size of the preprocessed heating system equipment is 640×480×3;

[0141] Step 13: Obtain the target detection box positions in the preprocessed original images of the heating system equipment, set labels for the preprocessed original images of the heating system equipment, and combine the preprocessed original images of the heating system equipment, the preprocessed original images of the heating system equipment with target detection box positions, and the labels into a dataset. Divide the dataset into training set, validation set, and test set in a 6:2:2 ratio.

[0142] The tags are categories of the original images of the heating system equipment;

[0143] Step 2: Train the device recognition network using the training set to obtain the trained device recognition network;

[0144] As shown in Figure 4(a), the fault detection network includes: a first convolutional module, a second convolutional module, a first C2f_add module, a second C2f_add module, a third convolutional module, a third C2f_add module, a fourth C2f_add module, a fifth C2f_add module, a sixth C2f_add module, a first SCDown module, a seventh C2f_add module, an eighth C2f_add module, a ninth C2f_add module, a tenth C2f_add module, a second SCDown module, a third SCDown module, a fourth SCDown module, a fifth SCDown module, and a sixth SCDown module. The module consists of the following modules: wn module, seventh SCDown module, eighth SCDown module, first C2fCIB_add module, SPPF module, PSA module, first Upsample module, first concatenation module, first C2f module, second Upsample module, second concatenation module, second C2f module, third C2f module, fourth convolution module, third concatenation module, second C2fCIB_add module, third C2fCIB_add module, ninth SCDown module, fourth concatenation module, third C2fCIB_add module, fourth C2fCIB_add module, header module, and result output module.

[0145] The first convolutional module includes: a first convolutional layer, a first batch of normalization layers, and a first activation function layer; the output of the first convolutional module is a preprocessed original image of the heating system equipment with a size of 320×240×48.

[0146] The first convolutional layer is a convolutional layer with a kernel size of 3×3, a stride of 2, and a padding of 1. The input to the first convolutional layer is the preprocessed original image of the heating system equipment.

[0147] The input to the first batch of normalization layers is the output of the first convolutional layer;

[0148] The first activation function layer is the Silu activation function, and the input of the first activation function layer is the output of the first batch of normalization layers;

[0149] The second convolutional module includes: a second convolutional layer, a second batch normalization layer, and a second activation function layer; the output of the second convolutional module is a feature map of size 160×120×96; the input of the second convolutional module is the output of the first convolutional module.

[0150] The second convolutional layer is a convolutional layer with a kernel size of 3×3, a stride of 2, and a padding of 1;

[0151] The input to the second batch of normalized layers is the output of the second convolutional layer;

[0152] The second activation function layer is the Silu activation function, and the input of the second activation function layer is the output of the second batch normalization layer;

[0153] The input to the first C2f_add module is the output of the second convolutional module;

[0154] The input of the second C2f_add module is the output of the first C2f_add module, and the output is a feature map with a size of 160×120×96.

[0155] The third convolutional module includes: a third convolutional layer, a third batch normalization layer, and a third activation function layer; the third convolutional module outputs a feature map of size 80×60×192.

[0156] The third convolutional layer is a convolutional layer with a kernel size of 3×3, a stride of 2, and a padding of 1. The input is the output of the second C2f_add module.

[0157] The input to the third batch of normalization layers is the output of the third convolutional layer;

[0158] The third activation function layer is the Silu activation function, and the input of the third activation function layer is the output of the third batch normalization layer;

[0159] The third C2f_add module is the same as the first C2f_add module, and the input of the third C2f_add module is the output of the 3 convolution module;

[0160] The fourth C2f_add module is the same as the first C2f_add module, and the input of the fourth C2f_add module is the output of the third C2f_add module;

[0161] The fifth C2f_add module is the same as the first C2f_add module, and the input of the fifth C2f_add module is the output of the fourth C2f_add module;

[0162] The sixth C2f_add module is the same as the first C2f_add module, and the input of the sixth C2f_add module is the output of the fifth C2f_add module; the output of the sixth C2f_add module is a feature map with a specification of 80×60×192.

[0163] The input of the first SCDown module is the output of the sixth C2f_add module, and the output of the first SCDown module is a feature map with a specification of 40×30×384.

[0164] The seventh C2f_add module is the same as the first C2f_add module, and its input is the output of the first SCDown module;

[0165] The eighth C2f_add module is the same as the first C2f_add module, and its input is the output of the seventh C2f_add module;

[0166] The ninth C2f_add module is the same as the first C2f_add module, and its input is the output of the eighth C2f_add module;

[0167] The tenth C2f_add module is the same as the first C2f_add module, and its input is the output of the ninth C2f_add module. The output of the tenth C2f_add module is a feature map X1 with a specification of 40×30×384.

[0168] The input to the second SCDown module is the output of the tenth C2f_add module;

[0169] The input to the third SCDown module is the output of the second SCDown module;

[0170] The input to the fourth SCDown module is the output of the third SCDown module;

[0171] The input to the fifth SCDown module is the output of the fourth SCDown module;

[0172] The input of the sixth SCDown module is the output of the fifth SCDown module;

[0173] The input of the seventh SCDown module is the output of the sixth SCDown module;

[0174] The input of the eighth SCDown module is the output of the seventh SCDown module, and the output of the eighth SCDown module is a feature map with a specification of 20×15×576.

[0175] The input of the first C2fCIB_add module is the output of the eighth SCDown module, and the output is a feature map with a specification of 20×15×576.

[0176] The input of the SPPF module is the output of the first C2fCIB_add module, and the output is a feature map with a specification of 20×15×576.

[0177] The input of the PSA module is the output of the SPPF module, and the output is a feature map with a specification of 20×15×576.

[0178] The input of the first Upsample module is the output of the PSA module, and the output is a feature map with a specification of 40×30×576.

[0179] The first stitching module is used to stitch together the output of the first Upsample module and the output of the tenth C2f_add module to obtain the first stitching result; the first stitching result is a feature map with a size of 40×30×960.

[0180] The input to the first C2f module is the first concatenation result; the first C2f module does not have a residual connection layer, and the rest of the structure is the same as the first C2f_add module.

[0181] The input to the second Upsample module is the output of the first C2f module; the output of the second Upsample module is a feature map with a size of 80×60×384.

[0182] The second stitching module is used to stitch together the output of the second Upsample module and the output of the sixth C2f_add module to obtain the second stitching result; the second stitching result is a feature map with a specification of 80×60×576;

[0183] The input to the second C2f module is the second splicing result;

[0184] The input of the third C2f module is the output of the second C2f module, and the output of the third C2f module is a feature map X0 with a specification of 80×60×192.

[0185] The fourth convolutional module includes: a fourth convolutional layer, a fourth batch normalization layer, and a fourth activation function layer; the fourth convolutional module outputs a feature map with a size of 40×30×192.

[0186] The fourth convolutional layer is a convolutional layer with a kernel size of 3×3, a stride of 2, and a padding of 1. The input of the fourth convolutional module is the output of the third C2f module.

[0187] The input to the fourth batch of normalized layers is the output of the fourth convolutional layer;

[0188] The fourth activation function layer is the Silu activation function, and the input of the fourth activation function layer is the output of the fourth batch normalization layer;

[0189] The third stitching module is used to stitch together the output of the first C2f module and the output of the fourth convolution module to obtain the third stitching result; the third stitching module outputs a feature map with a specification of 40×30×576;

[0190] The input to the second C2fCIB_add module is the output of the third concatenation module;

[0191] The input of the third C2fCIB_add module is the output of the second C2fCIB_add module, and the output of the third C2fCIB_add module is a feature map X1 with a specification of 40×30×384.

[0192] The input of the ninth SCDown module is the output of the third C2fCIB_add module, and the output of the ninth SCDown module is a feature map with a specification of 20×15×384.

[0193] The fourth splicing module is used to splice the output of the PSA module and the output of the ninth SCDown module to obtain the fourth splicing result; the output of the fourth splicing module is a feature map with a specification of 20×15×960.

[0194] The input to the third C2fCIB_add module is the fourth concatenation result;

[0195] The input of the fourth C2fCIB_add module is the output of the third C2fCIB_add module; the output of the fourth C2fCIB_add module is a feature map X2 with a specification of 20×15×576;

[0196] As shown in Figure 4(b), the head module includes: a feature map X0 processing unit, a feature map X1 processing unit, and a feature map X2 processing unit;

[0197] The feature map X0 processing unit includes: the 5th convolution sub-unit, the 6th convolution sub-unit, the 7th convolution sub-unit, the 8th convolution sub-unit, the 9th convolution sub-unit, the 10th convolution sub-unit, the 11th convolution sub-unit, the 12th convolution sub-unit, and the 5th splicing sub-unit;

[0198] The fifth convolutional subunit includes: a fifth convolutional layer, a fifth batch normalization layer, and a fifth activation function layer;

[0199] The fifth convolutional layer is a convolutional layer with a kernel of 3×3, a stride of 1, and a padding of 1; the input of the fifth convolutional layer is the feature map X0.

[0200] The input to the fifth batch of normalization layers is the output of the fifth convolutional layer;

[0201] The fifth activation function layer is a silu activation function, and the input of the fifth activation function layer is the output of the fifth batch normalization layer;

[0202] The sixth convolutional subunit includes: a sixth convolutional layer, a sixth batch normalization layer, and a sixth activation function layer;

[0203] The sixth convolutional layer is a convolutional layer with a kernel of 3×3, a stride of 1, and a padding of 1; the input of the sixth convolutional layer is the output of the fifth convolutional sub-unit.

[0204] The input to the sixth batch of normalized layers is the output of the sixth convolutional layer;

[0205] The sixth activation function layer is a silu activation function, and the input of the sixth activation function layer is the output of the sixth batch normalization layer;

[0206] The 7th convolutional sub-unit is a convolutional layer with a kernel of 1×1 and a stride of 1. The input of the 7th convolutional sub-unit is the output of the 6th convolutional sub-unit. The output of the 7th convolutional sub-unit is a feature map with a size of 80×60×64.

[0207] The 8th convolutional subunit includes: the 8th convolutional layer, the 8th batch normalization layer, and the 8th activation function layer;

[0208] The eighth convolutional layer is a convolutional layer with a kernel of 3×3, a stride of 1, and a padding of 1; the input of the eighth convolutional layer is the feature map X1.

[0209] The input to the eighth batch of normalized layers is the output of the eighth convolutional layer;

[0210] The eighth activation function layer is a silu activation function, and the input of the eighth activation function layer is the output of the eighth batch normalization layer.

[0211] The 9th convolutional subunit includes: a 9th convolutional layer, a 9th batch normalization layer, and a 9th activation function layer;

[0212] The ninth convolutional layer is a convolutional layer with a kernel of 1×1 and a stride of 1; the input of the ninth convolutional layer is the output of the 8th convolutional sub-unit.

[0213] The input to the ninth batch of normalized layers is the output of the ninth convolutional layer;

[0214] The ninth activation function layer is a silu activation function, and the input of the ninth activation function layer is the output of the ninth batch normalization layer.

[0215] The 10th convolutional subunit includes: a 10th convolutional layer, a 10th batch normalization layer, and a 10th activation function layer;

[0216] The tenth convolutional layer is a convolutional layer with a kernel of 3×3, a stride of 1, and a padding of 1; the input of the tenth convolutional layer is the output of the 9th convolutional sub-unit.

[0217] The input to the tenth batch of normalized layers is the output of the tenth convolutional layer;

[0218] The tenth activation function layer is a silu activation function, and the input of the tenth activation function layer is the output of the tenth batch normalization layer;

[0219] The 11th convolutional subunit includes: an eleventh convolutional layer, an eleventh batch normalization layer, and an eleventh activation function layer;

[0220] The eleventh convolutional layer is a convolutional layer with a kernel of 1×1 and a stride of 1; the input of the eleventh convolutional layer is the output of the 10th convolutional sub-unit.

[0221] The input to the eleventh batch of normalization layers is the output of the eleventh convolutional layer;

[0222] The eleventh activation function layer is a silu activation function, and the input of the eleventh activation function layer is the output of the eleventh batch normalization layer.

[0223] The 12th convolutional sub-unit is a convolutional layer with a kernel of 1×1 and a stride of 1. The input of the 12th convolutional sub-unit is the output of the 11th convolutional sub-unit. The output of the 12th convolutional sub-unit is a feature map with a size of 80×60×80.

[0224] The fifth splicing subunit is used to splice the output of the seventh convolutional subunit and the output of the twelfth convolutional subunit to obtain feature map X'0; feature map X'0 is a feature map with a size of 80×60×144.

[0225] The structure of the feature map X1 processing unit is the same as that of the feature map X0 processing unit; the input of the feature map X1 processing unit is feature map X1, and the output is feature map X'1; feature map X'1 is a feature map with a size of 40×30×144.

[0226] The structure of the feature map X2 processing unit is the same as that of the feature map X0 processing unit; the input of the feature map X2 processing unit is feature map X2, and the output is feature map X'2; feature map X'2 is a feature map with a size of 20×15×144.

[0227] The result output module includes: an image mapping unit and a non-maximum suppression unit;

[0228] The image mapping unit is used to scale feature map X'0, feature map X'1 and feature map X'2 to the same size as the original image of the heating system equipment and then map them onto the original image of the heating system equipment to obtain the original image of the heating system equipment after mapping feature map X'0, feature map X'1 and feature map X'2;

[0229] The non-maximum suppression unit is used to perform non-maximum suppression processing on the original images of the heating system equipment after mapping feature map X'0, feature map X'1 and feature map X'2, to obtain images of the heating system equipment with target detection box positions and heating equipment types.

[0230] The C2f_add module, SCDowm module, C2fCIB_add module, SPPF module, and PSA module mentioned in step two are existing modules, specifically:

[0231] As shown in Figure 3(a), the C2f_add module in this step includes: a first CBS unit, a split unit, n Bottleneck units, a channel fusion unit, and a second CBS unit;

[0232] The first CBS unit and the second CBS unit each include a 1*1 convolutional layer, a batch normalization layer, and an activation function layer;

[0233] The output of the first CBS unit is the input of the split unit. The split unit divides the output of the first CBS unit into two feature maps along the channel dimension. The input of the first Bottleneck unit is one of the feature maps output by the split unit. The output of the first Bottleneck unit is the input of the second Bottleneck unit, and so on, until the output of the (n-1)th Bottleneck unit is input to the nth Bottleneck unit. The outputs of the n Bottleneck units are all input to the channel fusion unit. The input of the channel fusion unit is the other feature map output by the split unit and the outputs of the n Bottleneck units. The output of the channel fusion unit is input to the second CBS unit.

[0234] As shown in Figure 3(b), the SCDowm module in this step includes a CBS unit and a CB unit;

[0235] The CB unit includes a 3*3 convolutional layer and a batch normalization layer;

[0236] As shown in Figure 3(c), the C2fCIB_add module includes: a first CBS unit, a Split unit, a first CIB unit, a second CIB unit, a feature connection unit, and a second CBS unit;

[0237] The output of the first CBS unit is the input of the Split unit. The Split unit divides the output of the first CBS unit into two feature maps, P0 and P1. P1 is input to the first CIB unit, and the output of the first CIB unit is input to the second CIB unit and the feature connection unit. The input of the second CIB unit is input to the feature connection unit. Feature map P0 is input to the feature connection unit. The feature connection unit concatenates the three feature maps and inputs them to the second CBS unit.

[0238] The first CBS unit and the second CIB unit each contain a 1*1 convolutional layer, a batch normalization layer, and an activation function layer;

[0239] As shown in Figure 3(d), the CIB unit includes: a first CBS subunit, a second CBS subunit, a third CBS subunit, a fourth CBS subunit, a fifth CBS subunit, and a feature addition subunit;

[0240] The first CBS subunit, the third CBS subunit, and the fifth CBS subunit each include a 3*3 convolutional layer, a batch normalization layer, and an activation function layer;

[0241] The second CBS subunit and the fourth CBS subunit each include a 1*1 convolutional layer, a batch normalization layer, and an activation function layer;

[0242] In the CIB unit, the output of the first CBS subunit is the input of the second CBS subunit, the output of the second CBS subunit is the input of the third CBS subunit, the input of the third CBS subunit is the input of the fourth CBS subunit, the output of the fourth CBS subunit is the input of the fifth CBS subunit, and the output of the fifth CBS subunit and the input of the first CBS subunit are the inputs of the feature addition subunit.

[0243] As shown in Figure 3(e), the SPPF module includes: a first CBS unit, a first max pooling layer unit, a second max pooling layer unit, a third max pooling layer unit, a feature connection unit, and a second CBS unit;

[0244] The first CBS unit and the second CBS unit each include a 1*1 convolutional layer, a batch normalization layer, and an activation function layer;

[0245] The output of the first CBS unit is the input of the first max pooling layer unit and the feature connection unit. The output of the first max pooling layer unit is the input of the second max pooling layer unit and the feature connection unit. The output of the second max pooling layer unit is the input of the third max pooling layer unit and the feature connection unit. The output of the second max pooling layer unit is the input of the feature connection unit. The feature connection unit concatenates the outputs of the first CBS unit, the first max pooling layer unit, the second max pooling layer unit, and the third max pooling layer unit according to the channel dimension, and inputs the concatenation result into the second CBS unit.

[0246] As shown in Figure 3(f), the PSA module includes: a first CBS unit, a Split unit, an attention unit, a first feature addition unit, a second CBS unit, a CB unit, a second feature addition unit, a feature connection unit, and a third CBS unit;

[0247] The first CBS unit, the second CBS unit, and the third CBS unit each include a 1*1 convolutional layer, a batch normalization layer, and an activation function layer;

[0248] The output of the first CBS unit is the input of the Split unit. The Split unit divides the output of the first CBS unit into feature maps M0 and M1. M1 is input to the attention unit and the first feature addition unit. The output of the attention unit is input to the first feature addition unit. The first feature addition unit adds feature map M1 and the output feature map of the attention unit element-wise. The output of the first feature addition unit is input to the second CBS unit and the second feature addition unit. The output of the second CBS unit is used as the input of the CB unit. The output of the CB unit is used as the input of the second feature addition unit. The second feature addition unit adds the output of the CB unit and the output of the first feature addition unit element-wise and sends the addition result to the feature connection unit. Feature map M0 is input to the feature connection unit. The feature connection unit concatenates the output of the second feature addition unit and feature map M0 according to the channel dimension and sends the concatenation result to the third CBS unit.

[0249] The labels of the CBS unit, CB unit, feature splicing unit, feature addition unit, and Split unit in each module introduced here are labels within their respective modules and do not represent the same processing unit.

[0250] Step 3: Test and validate the trained device identification network using the test set and validation set. If the accuracy of the trained device identification network is greater than the preset accuracy threshold, then use the currently trained fault detection network as the heating equipment identification model; otherwise, return to Step 1.

[0251] Thermal imaging technology is a detection method that uses infrared detectors to capture the infrared radiation energy (i.e., heat) emitted by objects or the human body and convert it into electrical signals. These electrical signals are used by an image processing system to generate a visualized thermal image, showing areas of different temperatures. This technology is based on a fundamental physical law: any object with a temperature above absolute zero emits infrared radiation, and the higher the temperature, the stronger the radiation. Thermal imaging devices capture this infrared radiation and convert it into an image, thereby measuring the surface temperature of an object. This means that thermal imaging technology can operate without external light sources, even in complete darkness or low-light environments, and has the advantages of being non-contact, highly sensitive, and capable of real-time detection. Therefore, using thermal imaging technology can significantly reduce labor costs and save energy. Any object with a temperature emits infrared radiation, the intensity of which can be described by Planck's Law. Therefore, this invention utilizes an infrared detector to scan the heating equipment to be diagnosed, obtaining a thermal image of the equipment.

[0252] S2. Use an infrared detector to scan the heating equipment to be diagnosed and obtain a thermal image of the equipment, specifically:

[0253] S201. Obtain the total radiation intensity corresponding to each pixel in the image of the heating equipment to be diagnosed, and based on the total radiation intensity corresponding to each pixel in the image of the heating equipment to be diagnosed, obtain the electrical signal corresponding to each pixel in the image of the heating equipment to be diagnosed output by the detector, specifically:

[0254] S201-1. Use an infrared detector to scan the heating equipment to be diagnosed and obtain the total radiation intensity E of each pixel in the image of the heating equipment to be diagnosed. I i R (T);

[0255] Infrared detectors scan the heating equipment to be diagnosed to obtain the total radiation intensity E of the heating equipment to be diagnosed. I i R (T) Specifically:

[0256] First, the radiation intensity of the i-th pixel in the image of the heating equipment to be diagnosed at different wavelengths is obtained as follows:

[0257]

[0258] Among them, E i (λ,T) is the total radiant intensity of the i-th pixel in the image of the heating equipment to be diagnosed at wavelength λ, and T iλ is the absolute temperature of the i-th pixel in the image of the heating equipment to be diagnosed, λ is the infrared wavelength (8μm≤λ≤14μm), and h is Planck's constant (6.626×10⁻⁶). -34 J·s, where c is the speed of light, and c is 3 × 10⁻⁶. 8 m / s, k is the Boltzmann constant, k = 1.381 × 10 - 23 J / K;

[0259] The total radiation intensity of the i-th pixel in the image of the heating equipment to be diagnosed is obtained by integrating the radiation intensity of the i-th pixel in the image of the heating equipment to be diagnosed at different wavelengths, specifically as follows:

[0260]

[0261] Where, λ max It is the smallest wavelength in the infrared band, λ min It is the maximum wavelength in the infrared band;

[0262] S201-2, Using the total radiation intensity E of the i-th pixel in the image of the heating equipment to be diagnosed IR (T i Acquire the detector's output electrical signal V i :

[0263] V i =k'·E IR (T i )

[0264] Where k' is the sensitivity coefficient of the infrared detector.

[0265] S202, the detector's output electrical signal V i The processed electrical signal is obtained by amplification, filtering, and correction. Using the processed electrical signal Obtain thermal images of the heating equipment to be diagnosed, specifically:

[0266] S202-1, Convert the output electrical signal V of the infrared detector i The processed electrical signal is obtained by amplification, filtering, and correction.

[0267] Signal amplification: The electrical signal V is amplified using a low-noise amplifier (LNA). i ;

[0268] Signal filtering: Use a low-pass filter to remove the amplified electrical signal V. i High-frequency noise in;

[0269] Signal correction: Blackbody calibration technique is used to correct the filtered electrical signal to obtain the processed electrical signal.

[0270] The response of the sensor is corrected using a blackbody source at a known temperature to obtain an accurate radiation intensity;

[0271] S202-2, Utilizing the processed electrical signal Obtain the temperature value of the heating equipment to be diagnosed corresponding to the i-th pixel in the image of the heating equipment to be diagnosed:

[0272]

[0273] Where k* is the scaling factor for converting electrical signals into temperature, and ε represents emissivity;

[0274] ∈=1: This indicates that the heating equipment to be diagnosed is a perfect blackbody, capable of emitting radiation with maximum efficiency at all wavelengths.

[0275] ∈<1: This indicates that the emissivity of the heating equipment to be diagnosed is lower than that of a perfect blackbody. The surface properties, smoothness, temperature, and wavelength of different materials are the main factors affecting the emissivity.

[0276] S203. Perform color mapping using the temperature values ​​of the heating equipment to be diagnosed corresponding to all pixels in the image of the heating equipment to be diagnosed, to obtain a thermal image of the heating equipment to be diagnosed:

[0277] The temperature value of the heating equipment to be diagnosed, corresponding to the i-th pixel in the image, is color-mapped as follows:

[0278]

[0279] in, yes Mapped color yes The color mapping function.

[0280] Each pixel of a thermal imager corresponds to a temperature value or radiation intensity value, meaning that the thermal imager collects temperature data from various points in the entire image.

[0281] In this step, color mapping typically corresponds to higher temperatures as red or yellow, and lower temperatures as blue or green; this enhances the visualization of temperature differences.

[0282] The specific processing procedure of the color mapping function is as follows:

[0283] To normalize the temperature to be mapped to the range [0,1], use the following formula:

[0284]

[0285] in, yes The normalized value, T max It is the maximum temperature value of the heating equipment to be diagnosed, corresponding to the pixel in the image of the heating equipment to be diagnosed, T. min It is the minimum temperature value of the heating equipment to be diagnosed, corresponding to the pixel in the image of the heating equipment to be diagnosed;

[0286] For each normalized value Determine its position within the color gradient. Use linear interpolation to achieve a smooth color transition, as shown in the following formula:

[0287]

[0288] Where (R0,G0,B0) and (R1,G1,B1) are two adjacent color nodes, and (R,G,B) is the linear interpolation between them.

[0289] The final color is generated by combining the (R, G, B) values ​​obtained through interpolation. The temperature value of each pixel is then converted to its corresponding color to produce the final thermal image. This thermal image can be displayed on a monitor or stored in a computer for further analysis.

[0290] S3. Map the location of the heating equipment in the image of the heating equipment to be diagnosed obtained in S1 to the thermal image of the heating equipment to be diagnosed obtained in S3, and obtain the center coordinates of the target detection box in the thermal image, the ratio of the width of the target detection box in the thermal image to the overall width of the thermal image, and the ratio of the height of the target detection box in the thermal image to the overall height of the thermal image, thereby obtaining the range of the target detection box in the thermal image.

[0291] The specific coordinates of the center of the target detection box in the thermal image are:

[0292]

[0293] Among them, (x heat ,y heat (x) represents the center coordinates of the target detection box in the thermal image. noraml ,y normal ) represents the center coordinates of the target detection box of the heating equipment in the image of the heating equipment to be diagnosed, W heat It is the width of the thermal image, W heat It is the height of the thermal image, W normal It is the image width of the heating equipment to be diagnosed, H normal It is the image height of the heating equipment to be diagnosed;

[0294] The ratio of the width of the target detection box in the thermal image to the overall width of the thermal image is obtained in the following way:

[0295]

[0296] Among them, w heat It is the ratio of the width of the target detection bounding box in the thermal image to the overall width of the thermal image, w normal W is the ratio of the width of the target detection box of the heating equipment in the image of the heating equipment to be diagnosed to the width of the image of the heating equipment to be diagnosed. heat It is the width of the thermal image, W normal It is the image width of the heating equipment to be diagnosed;

[0297] The ratio of the target detection box height to the overall height of the thermal image is obtained in the following way:

[0298]

[0299] Among them, h heat h is the ratio of the height of the target detection bounding box in the thermal image to the overall height of the thermal image. normal H is the ratio of the height of the target detection box of the heating equipment in the image of the heating equipment to be diagnosed to the height of the image of the heating equipment to be diagnosed. heat It is the height of the thermal image, H normal It is the height and width of the image of the heating equipment to be diagnosed.

[0300] S4. Use the average temperature within the target detection box range in the thermal image to determine whether the heating equipment to be diagnosed is abnormal. If an abnormality is found, an alarm is triggered; otherwise, the process ends. Specifically:

[0301] S401. Obtain the average temperature within the range of the target detection box in the thermal image, specifically:

[0302]

[0303] Among them, T aver It is the average temperature (T) within the bounding box of the target in the thermal image. j is the temperature value of the j-th pixel within the bounding box of the target in the thermal image, M is the total number of pixels within the bounding box of the target in the thermal image, and j is the pixel number within the bounding box of the target in the thermal image.

[0304] S402. Based on the heating equipment type obtained in S1 and the average temperature within the target detection box range in the thermal image, determine whether the heating equipment to be diagnosed is abnormal. If an abnormality is found, issue an alarm; otherwise, end the detection. Specifically:

[0305] If the heating equipment is a boiler, then when 70℃≤T aver ≤90℃ indicates that the boiler has not experienced a heating failure. If T aver <70℃ indicates the boiler temperature is too low, triggering a boiler temperature low alarm; if Taver If the temperature is >90℃, it indicates that the boiler temperature is too high, and a boiler overheating alarm will be issued.

[0306] If the heating equipment type is a radiator, then when 60℃≤T aver ≤80℃ indicates that the radiator has no heating failure. If T aver <60℃ indicates the radiator temperature is too low, triggering a low radiator temperature alarm; if T aver If the temperature is >80℃, it indicates that the radiator temperature is too high, and an overheating alarm for the radiator will be issued.

[0307] If the heating equipment is a heat exchanger, then when 50℃≤T aver ≤90℃ indicates that the heat exchanger has not experienced a heating failure. If T aver <50℃ indicates that the heat exchanger temperature is too low, triggering a low temperature alarm; if T aver If the temperature is >90℃, it indicates that the heat exchanger temperature is too high, and a heat exchanger over-temperature alarm will be issued.

[0308] If the heating equipment is a circulating pump, then when 30℃≤T aver ≤60℃ indicates that the circulating pump has not experienced a heating failure. If T aver <30℃ indicates that the circulating pump temperature is too low, and a low circulating pump temperature alarm will be issued; if T aver If the temperature is >60℃, it indicates that the circulating pump temperature is too high, and an alarm for excessive circulating pump temperature will be issued.

[0309] For network-related early warnings, different handling measures need to be taken depending on the type of equipment. When the boiler temperature is too high, reduce fuel input to decrease heating capacity, promptly inject cooling water, and check the safety valve to ensure it is functioning properly and prevent explosion risks; when it is too low, increase fuel input to increase heating capacity. When the radiator temperature is too high, reduce heat source output (e.g., close valves), check thermostat settings to ensure temperature parameters are reasonable, and ensure there are no obstructions around the radiator hindering heat dissipation; when it is too low, increase heat source output (e.g., open valves), check for air blockages in the system, ensure smooth water flow, and check the thermostat to ensure it is functioning properly. When the heat exchanger temperature is too high, adjust the cooling water flow to ensure sufficient cooling, check for blockages, ensure smooth flow, check the cleanliness of the heat exchanger, and clean it if necessary; when it is too low, increase fluid flow or adjust the heating source output, and check insulation measures to ensure no heat loss. When the circulating pump temperature is too high, check the pump flow setting to ensure that the flow rate is appropriate, check the cooling system to ensure that the pump has good heat dissipation and that its operating frequency is within a reasonable range; if it is too low, increase the pump flow rate or adjust the operating frequency, check for blockages, and ensure that the flow is unobstructed.

Claims

1. A method for fault diagnosis of a heating system based on artificial intelligence and Internet of Things technologies, characterized in that... The specific process of the method is as follows: S1. Obtain the image of the heating equipment to be diagnosed, input the image of the heating equipment to be diagnosed into the heating equipment recognition model, and obtain the position of the target detection box of the heating equipment and the type of the heating equipment in the image of the heating equipment to be diagnosed. The heating equipment identification model is obtained by verifying and testing a trained equipment identification network. The device identification network includes: a first convolutional module, a second convolutional module, a first C2f_add module, a second C2f_add module, a third convolutional module, a third C2f_add module, a fourth C2f_add module, a fifth C2f_add module, a sixth C2f_add module, a first SCDown module, a seventh C2f_add module, an eighth C2f_add module, a ninth C2f_add module, a tenth C2f_add module, a second SCDown module, a third SCDown module, a fourth SCDown module, a fifth SCDown module, and a sixth SCDown module. The modules are: 7th SCDown module, 8th SCDown module, 1st C2fCIB_add module, SPPF module, PSA module, 1st Upsample module, 1st splicing module, 1st C2f module, 2nd Upsample module, 2nd splicing module, 2nd C2f module, 3rd C2f module, 4th convolution module, 3rd splicing module, 2nd C2fCIB_add module, 3rd C2fCIB_add module, 9th SCDown module, 4th splicing module, 3rd C2fCIB_add module, 4th C2fCIB_add module, header module, and result output module. The third C2f module outputs feature map X0; The third C2fCIB_add module outputs feature map X1; The fourth C2fCIB_add module outputs feature map X2; The head module obtains feature maps X'0, X'1, and X'2 by processing feature maps X0, X1, and X2. The result output module processes feature maps X'0, X'1, and X'2 to obtain images of heating system equipment with target detection box locations and heating equipment types. S2. Use an infrared detector to scan the heating equipment to be diagnosed and obtain a thermal image of the heating equipment to be diagnosed. S3. Map the position of the target detection box of the heating equipment in the image of the heating equipment to be diagnosed obtained in S1 to the thermal image of the heating equipment to be diagnosed obtained in S2, and obtain the center coordinates of the target detection box in the thermal image, the ratio of the width of the target detection box in the thermal image to the overall width of the thermal image, and the ratio of the height of the target detection box in the thermal image to the overall height of the thermal image, thereby obtaining the range of the target detection box in the thermal image. S4. Use the average temperature within the target detection box in the thermal image to determine whether the heating equipment to be diagnosed is abnormal. If an abnormality is found, an alarm will be triggered; otherwise, the detection will end directly.

2. The method for fault diagnosis of a heating system based on artificial intelligence and Internet of Things technology according to claim 1, characterized in that: The heating equipment identification model in S1 is obtained through the following method: Step 1: Obtain raw images of the heating system equipment and preprocess them. Use the preprocessed raw images to form a dataset, and then divide the dataset into training, validation, and test sets. Step 11: Obtain original images of the heating system equipment; Step 1 and Step 2: Preprocess the original images of the heating system equipment to obtain preprocessed original images of the heating system equipment; Step 13: Obtain the target detection box position in the preprocessed original image of the heating system equipment, use the heating system equipment type as the label of the preprocessed original image of the heating system equipment, and combine the preprocessed original image of the heating system equipment, the preprocessed original image of the heating system equipment with the target detection box position, and the label to form a dataset, and divide the dataset into training set, validation set and test set. The types of equipment in the heating system include: boiler images, radiator images, heat exchanger images, and circulating pump images; Step 2: Train the device recognition network using the training set, and verify and test the trained device recognition network using the test set and validation set to finally obtain the heating equipment recognition model.

3. The method for fault diagnosis of a heating system based on artificial intelligence and Internet of Things technology according to claim 2, characterized in that: The preprocessing of the original images of the heating system equipment in steps one and two specifically includes: First, data augmentation is performed on the original images of the heating system equipment; Then, bilinear interpolation is used to scale the original image of the heating system equipment after data augmentation to obtain the preprocessed original image of the heating system equipment.

4. The method for fault diagnosis of a heating system based on artificial intelligence and Internet of Things technology according to claim 3, characterized in that: The first convolutional module includes: a first convolutional layer, a first batch of normalization layers, and a first activation function layer; the input to the first convolutional module is the preprocessed original image of the heating system equipment; The first convolutional layer has a kernel size of 3. The first convolutional layer has 3 layers; the input to the first convolutional layer is the preprocessed original image of the heating system equipment. The input to the first batch of normalization layers is the output of the first convolutional layer; The first activation function layer is the Silu activation function, and the input of the first activation function layer is the output of the first batch of normalization layers; The second convolutional module includes: a second convolutional layer, a second batch normalization layer, and a second activation function layer; the input of the second convolutional module is the output of the first convolutional module. The second convolutional layer has a kernel size of 3. 3 convolutional layers; The input to the second batch of normalized layers is the output of the second convolutional layer; The second activation function layer is the Silu activation function, and the input of the second activation function layer is the output of the second batch normalization layer; The input to the first C2f_add module is the output of the second convolutional module; The input to the second C2f_add module is the output of the first C2f_add module; The third convolutional module includes: a third convolutional layer, a third batch normalization layer, and a third activation function layer; the input of the third convolutional module is the output of the second C2f_add module; The third convolutional layer has a kernel size of 3. The third convolutional layer has three layers; the input to the third convolutional layer is the output of the second C2f_add module. The input to the third batch of normalization layers is the output of the third convolutional layer; The third activation function layer is the Silu activation function, and the input of the third activation function layer is the output of the third batch normalization layer; The input to the third C2f_add module is the output of the 3 convolutional module; The input to the fourth C2f_add module is the output of the third C2f_add module; The input to the fifth C2f_add module is the output of the fourth C2f_add module; The input to the sixth C2f_add module is the output of the fifth C2f_add module; The input to the first SCDown module is the output of the sixth C2f_add module; The input of the seventh C2f_add module is the output of the first SCDown module; The input of the eighth C2f_add module is the output of the seventh C2f_add module; The input of the ninth C2f_add module is the output of the eighth C2f_add module; The input of the tenth C2f_add module is the output of the ninth C2f_add module; The input to the second SCDown module is the output of the tenth C2f_add module; The input to the third SCDown module is the output of the second SCDown module; The input to the fourth SCDown module is the output of the third SCDown module; The input to the fifth SCDown module is the output of the fourth SCDown module; The input of the sixth SCDown module is the output of the fifth SCDown module; The input of the seventh SCDown module is the output of the sixth SCDown module; The input of the eighth SCDown module is the output of the seventh SCDown module; The input to the first C2fCIB_add module is the output of the eighth SCDown module; The input to the SPPF module is the output of the first C2fCIB_add module; The input of the PSA module is the output of the SPPF module; The input to the first Upsample module is the output of the PSA module; The first splicing module is used to splice the output of the first Upsample module and the output of the tenth C2f_add module to obtain the first splicing result; The input to the first C2f module is the first splicing result; The input to the second Upsample module is the output of the first C2f module; The second splicing module is used to splice the output of the second Upsample module and the output of the sixth C2f_add module to obtain the second splicing result; The input to the second C2f module is the second splicing result; The input of the third C2f module is the output of the second C2f module; the output of the third C2f module is feature map X0; The fourth convolutional module includes: a fourth convolutional layer, a fourth batch normalization layer, and a fourth activation function layer; the input of the fourth convolutional module is the output of the third C2f module. The fourth convolutional layer has a kernel size of 3. The third convolutional layer; the input to the fourth convolutional layer is the output of the third C2f module; The input to the fourth batch of normalized layers is the output of the fourth convolutional layer; The fourth activation function layer is the Silu activation function, and the input of the fourth activation function layer is the output of the fourth batch normalization layer; The third splicing module is used to splice the output of the first C2f module and the output of the fourth convolution module to obtain the third splicing result; The input to the second C2fCIB_add module is the output of the third concatenation module; The input of the third C2fCIB_add module is the output of the second C2fCIB_add module; the third C2fCIB_add module outputs feature map X1; The input to the ninth SCDown module is the output of the third C2fCIB_add module; The fourth splicing module is used to splice the output of the PSA module and the output of the ninth SCDown module to obtain the fourth splicing result; The input to the third C2fCIB_add module is the fourth concatenation result; The fourth C2fCIB_add module takes as input the output of the third C2fCIB_add module; the fourth C2fCIB_add module outputs feature map X2. The head module obtains feature maps X'0, X'1, and X'2 by processing feature maps X0, X1, and X2. The result output module processes feature maps X'0, X'1, and X'2 to obtain images of heating system equipment with target detection box locations and heating equipment types.

5. The method for fault diagnosis of a heating system based on artificial intelligence and Internet of Things technology according to claim 4, characterized in that: The head module includes: a feature map X0 processing unit, a feature map X1 processing unit, and a feature map X2 processing unit; The feature map X0 processing unit includes: the 5th convolution sub-unit, the 6th convolution sub-unit, the 7th convolution sub-unit, the 8th convolution sub-unit, the 9th convolution sub-unit, the 10th convolution sub-unit, the 11th convolution sub-unit, the 12th convolution sub-unit, and the 5th splicing sub-unit; the input of the feature map X0 processing unit is the feature map X0; The fifth convolutional subunit includes: a fifth convolutional layer, a fifth batch normalization layer, and a fifth activation function layer; The fifth convolutional layer has a convolutional kernel of... The fifth convolutional layer is input to the feature map X0. The input to the fifth batch of normalization layers is the output of the fifth convolutional layer; The fifth activation function layer is a silu activation function, and the input of the fifth activation function layer is the output of the fifth batch normalization layer; The sixth convolutional subunit is the same as the fifth convolutional subunit; the input of the sixth convolutional subunit is the output of the fifth convolutional subunit. The 7th convolutional subunit has a convolutional kernel of 1. A convolutional layer of size 1; the input of the 7th convolutional sub-unit is the output of the 6th convolutional sub-unit. The 8th convolutional subunit is the same as the 5th convolutional subunit; the input of the 8th convolutional subunit is the feature map X1; The 9th convolutional subunit includes: a 9th convolutional layer, a 9th batch normalization layer, and a 9th activation function layer; the input of the 9th convolutional subunit is the output of the 8th convolutional subunit. The ninth convolutional layer has a convolutional kernel of 1. The ninth convolutional layer is the input of the eighth convolutional subunit; The input to the ninth batch of normalized layers is the output of the ninth convolutional layer; The ninth activation function layer is a silu activation function, and the input of the ninth activation function layer is the output of the ninth batch normalization layer. The 10th convolutional subunit is the same as the 5th convolutional subunit; the input of the 10th convolutional subunit is the output of the 9th convolutional subunit; The 11th convolutional subunit is the same as the 9th convolutional subunit; the input of the 11th convolutional subunit is the output of the 10th convolutional subunit; The 12th convolutional subunit has a convolutional kernel of 1. A convolutional layer of 1; the input of the 12th convolutional sub-unit is the output of the 11th convolutional sub-unit; The fifth splicing subunit is used to splice the output of the seventh convolutional subunit and the output of the 12th convolutional subunit to obtain the feature map X'0; The structure of the feature map X1 processing unit is the same as that of the feature map X0 processing unit; the input of the feature map X1 processing unit is feature map X1, and the output is feature map X'1. The structure of the feature map X2 processing unit is the same as that of the feature map X0 processing unit; the input of the feature map X2 processing unit is feature map X2, and the output is feature map X'2.

6. The method for fault diagnosis of a heating system based on artificial intelligence and Internet of Things technology according to claim 5, characterized in that: The result output module includes: an image mapping unit and a non-maximum suppression unit; The image mapping unit is used to scale feature map X'0, feature map X'1 and feature map X'2 to the same size as the original image of the heating system equipment and then map them onto the original image of the heating system equipment to obtain the original image of the heating system equipment after mapping feature map X'0, feature map X'1 and feature map X'2; The non-maximum suppression unit is used to perform non-maximum suppression processing on the original images of the heating system equipment after mapping feature map X'0, feature map X'1 and feature map X'2, to obtain images of the heating system equipment with target detection box positions and heating equipment types.

7. The method for fault diagnosis of a heating system based on artificial intelligence and Internet of Things technology according to claim 6, characterized in that: The coordinates of the center of the target detection box in the thermal image in S3 are obtained in the following way: in, These are the coordinates of the center of the target detection box in the thermal image. These are the coordinates of the center of the target detection box of the heating equipment in the image of the heating equipment to be diagnosed. It is the width of the thermal image. It is the height of the thermal image. It is the image width of the heating equipment to be diagnosed. It is the image height of the heating equipment to be diagnosed.

8. The method for fault diagnosis of a heating system based on artificial intelligence and Internet of Things technology according to claim 7, characterized in that: The ratio of the width of the target detection box in the thermal image to the overall width of the thermal image is obtained in the following way: in, It is the ratio of the width of the target detection box in the thermal image to the overall width of the thermal image. It is the ratio of the width of the target detection box of the heating equipment in the image of the heating equipment to be diagnosed to the width of the image of the heating equipment to be diagnosed.

9. A method for fault diagnosis of a heating system based on artificial intelligence and Internet of Things technologies according to claim 8, characterized in that: The ratio of the height of the target detection box in the thermal image to the overall height of the thermal image is specifically: in, It is the ratio of the height of the target detection bounding box in the thermal image to the overall height of the thermal image. It is the ratio of the height of the target detection box of the heating equipment in the image of the heating equipment to be diagnosed to the height of the image of the heating equipment to be diagnosed.

10. A method for fault diagnosis of a heating system based on artificial intelligence and Internet of Things technologies according to claim 9, characterized in that: In step S4, the average temperature within the target detection box range of the thermal image is used to determine whether the heating equipment to be diagnosed is malfunctioning. If an malfunction is detected, an alarm is triggered; otherwise, the detection process ends directly. Specifically: S401. Obtain the average temperature within the range of the target detection box in the thermal image, specifically: in, It is the average temperature within the bounding box of the target in the thermal image. It is the temperature value of the j-th pixel within the bounding box of the target in the thermal image. is the total number of pixels in the bounding box of the target in the thermal image, and j is the pixel number in the bounding box of the target in the thermal image; S402. Based on the type of heating equipment and the average temperature within the target detection box range in the thermal image, determine whether the heating equipment to be diagnosed is abnormal. If an abnormality is found, issue an alarm; otherwise, end the detection. Specifically: If the heating equipment type is a boiler, then when This indicates that the boiler has no heating malfunction, and the test is complete; if This indicates that the boiler temperature is too low, and a boiler temperature too low alarm will be issued; if If the temperature is too high, an alarm for excessively high boiler temperature will be triggered. If the heating equipment type is a radiator, then when This indicates that the radiator has no heating failure, and the test ends; if This indicates that the radiator temperature is too low, and a low radiator temperature alarm will be issued; if If the radiator temperature is too high, an overheating alarm will be triggered. If the heating equipment type is a heat exchanger, then when This indicates that there is no heating failure in the heat exchanger, and the test ends; if This indicates that the heat exchanger temperature is too low, and a low heat exchanger temperature alarm will be issued; if If the temperature is too high, it indicates that the heat exchanger temperature is too high and an alarm for excessive heat exchanger temperature will be issued. If the heating equipment type is a circulating pump, then when This indicates that the circulating pump has not experienced a heating failure, and the test is now complete; if This indicates that the circulating pump temperature is too low, and a low circulating pump temperature alarm will be issued; if If the temperature is too high, it indicates that the circulating pump temperature is too high, and an alarm for excessive circulating pump temperature will be issued.

Images