A dust removal fan fault detection method and system

Abstract

This invention relates to the field of dust collector fan technology, and more particularly to a method and system for detecting faults in dust collector fans. The method includes collecting normal operating data of the target fan; inputting the normal operating data into a neural network model; calculating and outputting the motor operating current and motor speed values ​​through the neural network model; and obtaining a neural network mathematical model through training. The advantages of this invention are: by using an edge warning terminal to collect multi-dimensional operating data (vibration, noise, temperature, speed, and current) in real time at the dust collector fan equipment site, and by continuously training the operating data to construct a neural network mathematical model, it can capture subtle changes in operating parameters that deviate from a healthy state, issuing warnings in the early stages or potential phases of fault occurrence. This allows maintenance personnel to shift from traditional "reactive maintenance" to "predictive maintenance," intervening early in the nascent stage of a fault and effectively avoiding production losses caused by unplanned downtime.

Description

Technical Field

[0001] This invention relates to the field of dust collector fan technology, and in particular to a method and system for detecting faults in dust collector fans. Background Technology

[0002] Dust collector fans are crucial equipment in coking plant production, responsible for collecting and treating dust within a specific area. Their operational status directly affects the continuity and stability of coking production. A malfunction in a dust collector fan can not only disrupt normal production but also cause environmental pollution and equipment damage, resulting in significant economic losses and operational challenges.

[0003] Currently, fault detection in coking plant dust collectors mainly relies on monitoring several key parameters, such as fan bearing temperature, motor bearing temperature, main turbine stator temperature, and fan vibration amplitude. The conventional approach is to set fixed alarm and interlock shutdown values ​​for these parameters. Once the detected data exceeds the threshold, the system will trigger an alarm or shut down the system, and notify maintenance personnel to come to the site for repair.

[0004] However, such traditional detection methods have the following limitations:

[0005] 1. It is a reactive response mechanism and cannot effectively predict or warn of failures before they occur;

[0006] 2. Relying on manual experience for troubleshooting and diagnosis results in low efficiency and may delay repair opportunities;

[0007] 3. The accuracy of fault identification is limited, and the ability to detect latent and gradual faults is insufficient;

[0008] 4. Frequent false alarms or omissions may affect production plans and increase the risk of unplanned downtime. Summary of the Invention

[0009] The purpose of this invention is to provide a method and system for detecting faults in dust collector fans. Based on multi-source sensor data, the system performs real-time analysis and diagnosis at the dust collector fan site, enabling early identification and warning of abnormalities in the dust collector fan's status, thereby maximizing the continuity and safety of production.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] A method for detecting faults in a dust collector fan includes:

[0012] Collect normal operating data of the target wind turbine;

[0013] Input the normal operating data of the target wind turbine into the neural network model;

[0014] The output motor current and motor speed are calculated using a neural network model.

[0015] A neural network mathematical model is obtained through training.

[0016] A neural network model consists of an input layer and an output layer, which are connected by neural network weight parameters.

[0017] The calculation formula for the neural network model is as follows:

[0018] ①

[0019] In formula ①, The weighting parameter 1 represents the bearing temperature value. The weighted parameter representing the fan oil temperature value The weighting parameter 1 represents the sound wave value of the wind turbine. The weighting parameter represents the bearing vibration value. The threshold value representing the output motor current value of the neural network model. This represents the motor current value calculated and output by the neural network model. This indicates the bearing temperature value. Indicates the fan oil temperature value. Indicates the sound wave value of the fan. Indicates the bearing vibration value;

[0020] ②

[0021] In formula ②, Weighting parameter 2 represents the bearing temperature value. Weighting parameters representing the motor winding temperature value. The weighting parameter 2 represents the sound wave value of the wind turbine. The weighting parameter represents the motor vibration value. The threshold value representing the motor speed output by the neural network model. This represents the motor speed value calculated and output by the neural network model. This indicates the temperature value of the motor windings. This indicates the motor vibration value.

[0022] A neural network mathematical model is obtained through training, including:

[0023] Establish a motor current deviation function between the motor current value calculated and output based on a neural network model and the motor current value in the target wind turbine operating data;

[0024] Establish a motor speed deviation function between the motor speed value calculated and output by a neural network model and the motor speed value in the target fan operating data;

[0025] Use the gradient descent algorithm to adjust the corresponding weight parameters and thresholds in the neural network model until the corresponding bias function value is minimized.

[0026] The motor current deviation function is calculated as follows:

[0027] ③

[0028] In formula ③, This represents the motor current deviation function. This represents the motor current value in the target wind turbine's operating data.

[0029] The motor speed deviation function is calculated as follows:

[0030] ④

[0031] In formula ④, This represents a function indicating the deviation of the motor speed. This represents the motor speed value in the target fan's operating data.

[0032] Gradient descent algorithm, with and For example, the gradient formula is as follows:

[0033] ⑤

[0034] ⑥

[0035] gradient The calculation formula is the same as that in formula ⑤. The calculation formula is the same as that in formula ⑥;

[0036] The weights are updated according to formulas ⑤ and ⑥, as follows:

[0037] ⑦

[0038] ⑧

[0039] In formulas ⑦ and ⑧, Indicates weight The updated value, Indicates weight initial value, Indicates threshold The updated value, Indicates threshold initial value, Describing the gradient d , Describing the gradient d , This represents the preset self-learning efficiency, with a value range of 0.0001 to 0.5.

[0040] The target fan's normal operating data includes bearing temperature, fan oil temperature, motor winding temperature, fan sound wave, bearing vibration, motor vibration, motor current, and motor speed.

[0041] A dust collector fan fault detection system includes a temperature sensor, a winding temperature sensor, a vibration sensor, a noise sensor, and a speed sensor for each dust collector fan. The temperature sensors are respectively installed on the fan bearings and the fan gearbox to detect the bearing temperature and the fan oil temperature, respectively. The winding temperature sensor is installed in the motor to detect the motor winding temperature. The vibration sensors are respectively installed on the fan bearings and the motor base to detect the bearing vibration and the motor vibration, respectively. The noise sensor is installed in the motor to detect the fan sound wave value. The speed sensor is installed on the fan bearings to detect the motor speed.

[0042] Each dust collector fan also includes an edge warning terminal. The edge warning terminal is connected to the temperature sensor, winding temperature sensor, vibration sensor, noise sensor and speed sensor of the corresponding dust collector fan through ports. Each edge warning terminal is connected to the server. The edge warning terminal is fixedly installed in the corresponding control box, which is fixedly installed next to the motor of the dust collector fan.

[0043] Each dust collector fan also includes an ammeter, which is fixedly installed in the control box next to the corresponding dust collector fan. The ammeter is used to detect the motor current value of the corresponding dust collector fan and is connected to the edge warning terminal through a port.

[0044] Compared with the prior art, the beneficial effects of the present invention are:

[0045] 1. By using edge early warning terminals to collect multi-dimensional operating data (vibration, noise, temperature, speed, current) in real time at the dust removal fan equipment site, and using the operating data for continuous training, a neural network mathematical model is built to capture subtle changes in operating parameters that deviate from the healthy state, and issue early warnings in the early stage or potential stage of failure. This allows maintenance personnel to shift from traditional "reactive maintenance" to "predictive maintenance", intervene in advance at the bud stage of failure, and effectively avoid production losses caused by unplanned downtime.

[0046] 2. By adopting a neural network model, data acquisition, analysis, and diagnostic calculations are carried out at the edge early warning terminal of each dust collector fan on-site. There is no need to upload massive amounts of raw data to the cloud or platform early warning system server for calculation, reducing the latency of data round trips and achieving millisecond-level local real-time diagnosis and response. This overcomes the problem of early warning delays that may be caused by network latency in the traditional cloud center computing mode, and significantly improves the timeliness and reliability of fault diagnosis.

[0047] 3. The edge early warning terminal performs high-speed calculations and judgments only locally. It only needs to upload key early warning information and other result data to the upper-level platform early warning system server, or upload process data for a period of time before and after the early warning is triggered for experts to analyze in depth. This processing method reduces the amount of data transmitted over the network and stored in the cloud, effectively saving bandwidth resources and platform early warning system server storage costs, and is particularly suitable for large-scale deployment applications.

[0048] 4. The neural network model used has a self-learning function, which can continuously iterate and optimize the weights and thresholds. The obtained neural network model of the dust removal fan infinitely approximates the normal operating state of the fan. Compared with the traditional alarm method with fixed thresholds, this invention improves the diagnostic accuracy and the intelligence level of the system.

[0049] 5. By using this method to continuously and accurately monitor the status of core components of the fan (such as bearings) and provide early warnings, the risk of machine shutdown due to sudden failures can be minimized, ensuring the continuous and stable operation of the entire coking plant's production system. At the same time, it extends the service life of the equipment and reduces maintenance costs and safety risks. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of a dust collector fan fault detection structure. Figure 1 .

[0051] Figure 2 This is a schematic diagram of a dust collector fan fault detection structure. Figure 2 .

[0052] Figure 3 This is a diagram of the neural network model structure.

[0053] In the diagram: 1. Temperature sensor; 2. Winding temperature sensor; 3. Speed ​​sensor; 4. Vibration sensor; 5. Noise sensor; 6. Motor; 7. Control box; 8. Local control box; 9. Fan gearbox. Detailed Implementation

[0054] The present invention will now be described in detail with reference to the accompanying drawings, but it should be noted that the implementation of the present invention is not limited to the following embodiments.

[0055] The following embodiments are implemented based on the technical solution of the present invention, providing detailed implementation methods and specific operation processes. However, the scope of protection of the present invention is not limited to the following embodiments. Unless otherwise specified, the methods used in the following embodiments are conventional methods.

[0056] Example 1

[0057] See Figure 1, Figure 2 A dust collector fan fault detection system includes an early warning system platform server. Each dust collector fan includes an edge early warning terminal, an ammeter, a temperature sensor 1, a winding temperature sensor 2, a vibration sensor 4, a noise sensor 5, and a speed sensor 3. The edge early warning terminal uses the BL410 model, but other models can also be used. The temperature sensor 1 is installed on the fan bearing and the fan gearbox 9, respectively, to detect the bearing temperature and the fan oil temperature. The winding temperature sensor 2 is installed in the motor to detect the motor winding temperature. The vibration sensor 4 is installed on the fan bearing and the motor base, respectively, to detect the bearing vibration and the motor vibration. The noise sensor 5 is installed in the motor to detect the fan sound wave value. The speed sensor 3 is installed on the fan bearing to detect the motor speed. Temperature sensor 1, winding temperature sensor 2, vibration sensor 4, noise sensor 5, and speed sensor 3 are each connected to their respective edge early warning terminals via ports. Each edge early warning terminal is connected to the early warning system platform server via Ethernet, fiber optic, or 4G module. Each edge early warning terminal is fixedly installed in its corresponding control box 7, which is fixedly installed next to the motor. Each dust collector fan is equipped with a local control box 8, which contains an ammeter connected to a current transformer. The ammeter is used to detect the motor current value of the corresponding dust collector fan and is connected to the corresponding edge early warning terminal via a port.

[0058] A method for detecting faults in a dust collector fan includes:

[0059] S1. Collect target wind turbine operating data;

[0060] The target fan operating data includes bearing temperature, fan oil temperature, motor winding temperature, fan sound wave, bearing vibration, motor vibration, motor current, and motor speed.

[0061] S2. Input the target wind turbine operating data into the neural network model.

[0062] S3. Calculate and output the motor current and motor speed using a neural network model;

[0063] A neural network model consists of an input layer and an output layer, which are connected by neural network weight parameters.

[0064] The calculation formula for the neural network model is as follows:

[0065] ①

[0066] In formula ①, The weighting parameter 1 represents the bearing temperature value. The weighted parameter representing the fan oil temperature value The weighting parameter 1 represents the sound wave value of the wind turbine. The weighting parameter represents the bearing vibration value. The threshold value representing the output motor current value of the neural network model. This represents the motor current value calculated and output by the neural network model. This indicates the bearing temperature value. Indicates the fan oil temperature value. Indicates the sound wave value of the fan. Indicates the bearing vibration value;

[0067] ②

[0068] In formula ②, Weighting parameter 2 represents the bearing temperature value. Weighting parameters representing the motor winding temperature value. The weighting parameter 2 represents the sound wave value of the wind turbine. The weighting parameter represents the motor vibration value. The threshold value representing the motor speed output by the neural network model. This represents the motor speed value calculated and output by the neural network model. This indicates the temperature value of the motor windings. This indicates the motor vibration value.

[0069] S4. A neural network model of wind turbine operating parameters is obtained through training, which is used to judge the wind turbine operating status in real time. Neural network training is to obtain a set of neural network weight parameters k11~k24 and neural network values ​​b1, b2, so as to minimize the bias function of the neural network, that is, to minimize the difference between the prediction bias and the actual bias of the neural network, and to minimize the deviation between the calculated operating parameters and the actual operating parameters, thereby obtaining an optimized neural network model. The process of solving for the neural network weight parameters k11~k24 and neural network thresholds b1, b2 is the neural network training process.

[0070] Specifically, it includes:

[0071] S41. Establish a motor current deviation function between the motor current value calculated and output based on the neural network model and the motor current value in the target wind turbine operating data;

[0072] The motor current deviation function is calculated as follows:

[0073] ③

[0074] In formula ③, This represents the motor current deviation function. This represents the motor current value in the target fan's operating data; This represents the motor current value calculated and output by the neural network model.

[0075] S42. Establish a motor speed deviation function between the motor speed value calculated and output based on the neural network model and the motor speed value in the target fan operating data;

[0076] The motor speed deviation function is calculated as follows:

[0077] ④

[0078] In formula ④, This represents a function indicating the deviation of the motor speed. This represents the motor speed value in the target fan's operating data. This represents the motor speed value calculated and output by the neural network model.

[0079] S43. Use the gradient descent algorithm to adjust the corresponding weight parameters and thresholds in the neural network model until the corresponding deviation function value is minimized.

[0080] Gradient descent algorithm, with and For example, the gradient formula is as follows:

[0081] ⑤

[0082] ⑥

[0083] gradient The calculation formula is the same as that in formula ⑤. The calculation formula is the same as that in formula ⑥;

[0084] The weights are updated according to formulas ⑤ and ⑥, as follows:

[0085] ⑦

[0086] ⑧

[0087] In formulas ⑦ and ⑧, Indicates weight The updated value, Indicates weight initial value, Indicates threshold The updated value, Indicates threshold initial value, Describing the gradient d , Describing the gradient d , This represents the preset self-learning efficiency, with a value range of 0.0001 to 0.5.

[0088] Example 2

[0089] In this embodiment, a dust collector fan fault detection method and system are the same as in Embodiment 1, except that a dust collector fan fault detection process is added based on Embodiment 1 and / or Embodiment 2.

[0090] First, initialize a set of weight data. and threshold data ;

[0091] Initial weights =0.1, =0.2, =0.3, =0.8, =6, =4, =6, =4; Initial threshold =0.1, =0.2.

[0092] Take a set of normal operating data of a dust collector fan as an example:

[0093] =70, =30, =130; =50, =0.06, =4, =30, =1300.

[0094] in, This indicates the bearing temperature value. Indicates the fan oil temperature value. This indicates the temperature value of the motor windings. Indicates the sound wave value of the fan. Indicates the bearing vibration value. This indicates the motor vibration value. Indicates the motor current value. This indicates the motor's operating speed.

[0095] A set of weight data will be initialized. Threshold data Normal operating data of the fan , , , , , Substituting into formulas ① and ② of the neural network model, we obtain the output of the neural network model. and : =28.148, =1256.2.

[0096] Based on formulas ③ and ④, the deviation of the motor current value is obtained. =1.66 and the deviation of the motor speed value =959.22 From the bias results, the output of the neural network model is... , , and the operating current of the dust collector fan motor and rotational speed There is a discrepancy between the two:

[0097] The gradient descent method is used to adjust the corresponding weight parameters and thresholds in the neural network model. The corresponding weights in the neural network model are then obtained according to formulas ⑤ and ⑥. and threshold data gradient:

[0098] =(28.148-30) 70 = -129.64;

[0099] =(28.148-30) 30 = -55.56;

[0100] =(28.148-30) 50 = -92.6;

[0101] =(28.148-30) 0.06 = -0.11112;

[0102] =(1256.2-1300) 70 = -3066;

[0103] =(1256.2-1300) 130 = -5694;

[0104] =(1256.2-1300) 50 = -2190;

[0105] =(1256.2-1300) 4 = -175.2;

[0106] =28.148-30=-1.852;

[0107] =1256.2-1300=-43.8;

[0108] The weights are updated according to formulas ⑤ and ⑥, and the weights are taken from the learning efficiency = 0.0001.

[0109] =0.1-0.0001 (-129.64) = 0.202964;

[0110] =0.2-0.0001 (-55.56) = 0.205556;

[0111] =0.3-0.0001 (-92.6) = 0.30926;

[0112] =0.8-0.0001 (-0.11112) = 0.800011;

[0113] =6-0.0001 (-3066) = 6.3066;

[0114] =4-0.0001 (-5694) = 4.5694;

[0115] =6-0.0001 (-2190) = 6.219;

[0116] =4-0.0001 (-175.2) = 4.01752;

[0117] = 0.1-0.0001 (-1.852) = 0.1001852;

[0118] =2-0.0001 (-43.8) = 2.00438;

[0119] Substituting the updated weights and thresholds into formulas ① and ②, let's take a set of normal operating data of a dust collector fan as an example: =70, =30, =130, =50, =0.06, =4, thus obtaining the output of the neural network model. and : =35.98, =1364.51. (Compared to using initial weight data) and threshold The obtained neural network output =28.148、 =1256.2 and target =30, A comparison with 1300 reveals that the direction of weight and threshold adjustment increases the output of the neural network, indicating that the direction of parameter adjustment is correct.

[0120] Substitute the updated weights and thresholds into formulas ① and ②, and then recalculate another set of outputs based on another set of target wind turbine operating data. , The loss bias is obtained using the bias function formulas ③ and ④. Then, the neural network weight parameters are further optimized according to the negative gradient direction and formulas ⑤ to ⑧. and neural network threshold Using a large amount of historical data on the normal operation of wind turbines, the neural network is trained through continuous iterative calculations until the gradient of the calculated weights and values ​​is obtained. The goal is to minimize the sum of db, or to reach a set number of iterations, thereby obtaining a set of optimized neural network weight parameters. The neural network mathematical model is obtained by using neural network thresholds b1 and b2 to obtain the target wind turbine operating parameters.

[0121] This neural network mathematical model can predict the wind turbine current under normal operating conditions. and rotational speed The corresponding bearing temperature value of the fan Fan oil temperature value Motor winding temperature value Fan sound wave value Bearing vibration value and motor vibration value The range.

[0122] When the real-time collected wind turbine current or rotational speed , and the current of the wind turbine neural network mathematical model or rotational speed When they are the same, compare the real-time collected values ​​of the fan bearing temperature. Fan oil temperature value Motor winding temperature value Fan sound wave value Bearing vibration value Motor vibration value When the real-time collected value deviates from the predicted value by more than a set value (e.g., the deviation function calculation result is greater than 1%), it is determined that the wind turbine has failed.

[0123] This invention utilizes edge early warning terminals to collect multi-dimensional operational data (vibration, noise, temperature, speed, and current) in real time at the dust collector fan equipment site. Through continuous training of this data, a neural network mathematical model is constructed to capture subtle changes in operating parameters that deviate from a healthy state, issuing early warnings in the early stages or potential phases of faults. This allows maintenance personnel to shift from traditional "reactive maintenance" to "predictive maintenance," intervening early in the nascent stage of faults and effectively avoiding production losses caused by unplanned downtime. By employing a neural network model, data collection, analysis, and diagnostic calculations are performed at the edge early warning terminal on each dust collector fan site, eliminating the need to upload massive amounts of raw data to the cloud or platform early warning system server for calculation. This reduces data round-trip latency, achieving millisecond-level local real-time diagnosis and response. It overcomes the early warning lag problems that may result from network latency in traditional cloud-centric computing models, significantly improving the timeliness and reliability of fault diagnosis. The edge early warning terminal performs high-precision calculations only locally. The system allows for rapid calculation and judgment, requiring only the uploading of key early warning information and other results to the upper-level platform's early warning system server, or uploading process data for a period before and after triggering an early warning for in-depth expert analysis. This approach reduces the amount of data transmitted over the network and stored in the cloud, effectively saving bandwidth resources and platform early warning system server storage costs, making it particularly suitable for large-scale deployment applications. The adopted neural network model has self-learning capabilities, enabling it to continuously iterate and optimize weights and thresholds. The resulting dust collector fan neural network model closely approximates the normal operating state of the fan. Compared to traditional fixed-threshold alarm methods, this invention improves diagnostic accuracy and system intelligence. By using this method for continuous and accurate status monitoring and early warning of core fan components (such as bearings), the system can minimize the risk of complete machine shutdown due to sudden failures, ensuring the continuous and stable operation of the entire coking plant's production system, while extending equipment lifespan and reducing maintenance costs and safety risks.

Claims

1. A method for detecting faults in a dust collector fan, characterized in that, include: Collect normal operating data of the target wind turbine; Input the normal operating data of the target wind turbine into the neural network model; The output motor current and motor speed are calculated using a neural network model. A neural network mathematical model is obtained through training; The neural network model includes an input layer and an output layer, which are connected by neural network weight parameters. The calculation formula for the neural network model is as follows: ① In formula ①, The first weighted parameter representing the bearing temperature value. The weighted parameter representing the fan oil temperature value The first weighted parameter representing the sound wave value of the wind turbine. The weighting parameter represents the bearing vibration value. The threshold value representing the output motor current value of the neural network model. This represents the motor current value output by the neural network model. This indicates the bearing temperature value. Indicates the fan oil temperature value. Indicates the sound wave value of the fan. Indicates the bearing vibration value; ② In formula ②, The second weighted parameter representing the bearing temperature value, Weighting parameters representing the motor winding temperature value. The second weighted parameter represents the sound wave value of the wind turbine. The weighting parameter represents the motor vibration value. The threshold value representing the motor speed output by the neural network model. This represents the motor speed value output by the neural network model. This indicates the temperature value of the motor windings. Indicates the motor vibration value; The method of obtaining a neural network mathematical model through training includes: Establish a motor current deviation function between the motor current value calculated and output based on a neural network model and the motor current value in the target wind turbine operating data; Establish a motor speed deviation function between the motor speed value calculated and output by a neural network model and the motor speed value in the target fan operating data; Use the gradient descent algorithm to adjust the corresponding weight parameters and thresholds in the neural network model until the corresponding bias function value is minimized.

2. The method for detecting faults in a dust collector fan according to claim 1, characterized in that, The motor current deviation function is defined by the following formula: ③ In formula ③, This represents the motor current deviation function. This represents the motor current value in the target wind turbine's operating data.

3. The method for detecting faults in a dust collector fan according to claim 1, characterized in that, The motor speed deviation function is defined by the following formula: ④ In formula ④, This represents a function indicating the deviation of the motor speed. This represents the motor speed value in the target fan's operating data.

4. The method for detecting faults in a dust collector fan according to claim 1, characterized in that, The gradient descent algorithm described above, with and For example, the gradient formula is as follows: ⑤ ⑥ gradient The calculation formula is the same as that in formula ⑤. The calculation formula is the same as that in formula ⑥; The weights are updated according to formulas ⑤ and ⑥, as follows: ⑦ ⑧ In formulas ⑦ and ⑧, Indicates weight The updated value, Indicates weight initial value, Indicates threshold The updated value, Indicates threshold initial value, Gradient , Gradient , This represents the preset self-learning efficiency, with a value range of 0.0001 to 0.

5.

5. The method for detecting faults in a dust collector fan according to claim 1, characterized in that, The target wind turbine's normal operating data includes bearing temperature, wind turbine oil temperature, motor winding temperature, wind turbine sound wave, bearing vibration, motor vibration, motor current, and motor speed.

6. A dust collector fan fault detection system that implements the method of any one of claims 1-5, characterized in that, Each dust collector fan includes a temperature sensor, a winding temperature sensor, a vibration sensor, a noise sensor, and a speed sensor. The temperature sensors are located on the fan bearings and the fan gearbox, respectively, to detect the bearing temperature and the fan oil temperature. The winding temperature sensor is located in the motor to detect the motor winding temperature. The vibration sensors are located on the fan bearings and the motor base, respectively, to detect the bearing vibration and the motor vibration. The noise sensor is located in the motor to detect the fan sound wave. The speed sensor is located on the fan bearings to detect the motor speed.

7. A dust collector fan fault detection system according to claim 6, characterized in that, Each dust collector fan also includes an edge warning terminal. The edge warning terminal is connected to the temperature sensor, winding temperature sensor, vibration sensor, noise sensor and speed sensor of the corresponding dust collector fan through ports. Each edge warning terminal is connected to the server. The edge warning terminal is fixedly installed in the corresponding control box, which is fixedly installed next to the motor of the dust collector fan.

8. A dust collector fan fault detection system according to claim 6, characterized in that, Each dust collector fan also includes an ammeter, which is fixedly installed in the control box next to the corresponding dust collector fan. The ammeter is used to detect the motor current value of the corresponding dust collector fan and is connected to the edge warning terminal through a port.

Images