Fault monitoring and diagnosing method and system for flour processing equipment

Abstract

The invention relates to the technical field of flour equipment, in particular to a fault monitoring and diagnosing method and system for flour processing equipment. Comprising the following steps: S1, collecting electrostatic data in equipment operation through a field grinding type electrostatic sensor, and synchronously obtaining humidity data through a capacitive humidity sensor; calculating the charge cumulant of each link based on a charge accumulation formula; and S2, constructing a fault monitoring sequence based on the static data and the humidity data, extracting time sequence characteristics of a charge marking sequence, inputting the time sequence characteristics into the pre-trained LSTM model for classification, and generating an equipment state characteristic signal. According to the invention, electrostatic monitoring, humidity regulation and control and particle analysis are fused, and equipment fault diagnosis and quality optimization are realized; abnormality is accurately recognized based on charge classification and a dynamic threshold value, humidity balance particle fineness and electrostatic suppression are synchronously and adaptively adjusted, the false alarm risk is reduced in combination with a dynamic feedback mechanism, and a reliable intelligent solution is provided for complex working conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of flour equipment, and in particular to a fault monitoring and diagnosis method and system for flour processing equipment. Background Art

[0002] In industrial equipment monitoring, the efficient and stable operation of equipment is directly related to production efficiency and product quality. However, static electricity accumulation and uneven distribution of media caused by factors such as material friction and environmental humidity fluctuations during processing often lead to frequent equipment failures such as blockages, overheating, and even safety hazards. Traditional systems often rely on a single sensor and have difficulty capturing the nonlinear characteristics of multi-source coupled signals. In existing technologies, static electricity monitoring is usually limited to static charge detection and lacks the ability to dynamically analyze time-series charge change patterns. In addition, fixed threshold settings cannot adapt to the diverse operating conditions of equipment at different speeds, material moisture, and particle fineness.

[0003] Furthermore, the uniformity of medium distribution is a key indicator of processing quality, but current systems rarely integrate medium distribution data with equipment operating parameters for analysis, resulting in delayed fault warnings or increased misjudgment rates. Despite the gradual application of machine learning technology in industrial monitoring, it still faces significant challenges in flour processing scenarios: insufficient effective fusion and feature extraction of multi-source heterogeneous data make it difficult to build highly robust models. Furthermore, traditional algorithms have limited ability to capture the temporal dependencies and nonlinear relationships of charge marker sequences, making it impossible to accurately distinguish equipment operating states. Achieving multi-parameter collaborative analysis, dynamic threshold adaptive adjustment, and real-time accurate fault diagnosis has become a key challenge in improving the reliability and intelligence of flour processing equipment. Summary of the Invention

[0004] In order to overcome the shortcomings of insufficient dynamic multi-parameter collaborative monitoring, the present invention provides a fault monitoring and diagnosis method and system for flour processing equipment.

[0005] The technical solution of the present invention is: a fault monitoring and diagnosis method for flour processing equipment, comprising the following steps: S1: The electrostatic data during equipment operation is collected through a field-mill electrostatic sensor, and the humidity data is simultaneously obtained through a capacitive humidity sensor; and the charge accumulation amount of each link is calculated based on the charge accumulation formula; S2: Build a fault monitoring sequence based on electrostatic data and humidity data, extract the time series features of the charge mark sequence, and input it into the pre-trained LSTM model for classification to generate the device status feature signal; S3: Combine the charge labeling sequence pattern classification results with the media distribution data to identify abnormal patterns using a dynamic threshold adjustment formula. A reinforcement learning agent is then used to optimize the threshold parameters, with device stability as the reward objective. S4: Fault diagnosis is performed based on abnormal pattern recognition results. At the same time, the charge mark sequence, medium distribution and historical data are integrated to train the time series prediction model and trigger the hierarchical response strategy.

[0006] Preferably, the collecting of electrostatic data during operation of the equipment by a field mill electrostatic sensor and the synchronous acquisition of humidity data by a capacitive humidity sensor include: The electrostatic data is the amount of charge generated in each link during the operation of the flour processing equipment; The humidity data is the humidity parameter required for controlling the charge amount during operation of the device.

[0007] Preferably, the charge accumulation amount of each link is calculated based on the charge accumulation formula, including: the charge accumulation formula is as follows, in, For the The charge of a device, For the The operating speed coefficient of each device is the ratio of the device speed to the reference speed, in m / s. is the humidity parameter, is the humidity compensation coefficient, is the medium fineness correction factor, is the coupling coefficient of the preceding link, For the The charge of a device, is the attenuation factor.

[0008] Preferably, the method of constructing a fault monitoring sequence based on electrostatic data and humidity data, extracting the time series features of the charge mark sequence and inputting them into a pre-trained LSTM model for classification to generate a device status feature signal includes: Charge polarity marking: The charge generated by each production link of the flour processing equipment is marked with polarity, where the positive charge is marked with a value of 1 and the negative charge is marked with a value of -1; Pattern classification processing: Static mode determination: When the charge mark values are arranged alternately as [1, -1, 1, -1, ...], it is determined to be in static working mode; Dynamic mode determination: When the charge mark sequence does not conform to the alternating arrangement rule, it is determined to be a dynamic working mode; The difference between adjacent marking values in the incremental marking rule increases in step size 1; Mixed-mode processing: When the charge marking sequence does not conform to either the static or dynamic mode definition, the incremental marking rule is executed: Step 1: Initialize the marking parameters and mark the first positive charge position as the reference value 1; Step 2: Process the subsequent charge quantities in the chronological order of the production links, and keep the current mark value when a positive charge is detected; Step 3: In a preset section where no positive charge appears continuously, perform a value increment operation, with each increment step being 1; Step 4: Repeat steps 2 to 3 until the entire sequence analysis is completed; The dynamic mode is defined as a non-alternating charge mark sequence mode, and the difference between adjacent mark values in the incremental mark rule maintains a constant incremental relationship.

[0009] Preferably, the combining of the charge mark sequence pattern classification result and the medium distribution data to identify abnormal patterns through a dynamic threshold adjustment formula includes: The processed flour is irradiated by a lighting device, and the irradiated flour is photographed by a camera to obtain medium distribution data; Based on the charge polarity labeling result and the medium distribution data, a dynamic threshold adjustment formula is used to dynamically monitor abnormal modes during the operation of flour processing equipment.

[0010] Preferably, the dynamic monitoring of abnormal modes during the operation of the flour processing equipment using a dynamic threshold adjustment formula includes: Based on the medium distribution data, a medium uniformity correction value is obtained using a medium uniformity correction formula; The dynamic threshold adjustment formula is as follows: in, is the dynamic threshold, is the current charge, is the reference charge, Set a reference value for humidity, is the charge mode weight coefficient, is the humidity sensitivity coefficient, It is the correction value of medium uniformity.

[0011] Preferably, the medium uniformity correction value is obtained based on the medium distribution data using a medium uniformity correction formula, including: the medium uniformity correction formula is as follows: in, is the medium uniformity correction value, is the medium uniformity index, and ∈(0,1].

[0012] Preferably, the fault diagnosis based on the abnormal pattern recognition result includes: If the medium uniformity correction value exceeds a third preset threshold, a shutdown recommendation signal is immediately generated; If the medium uniformity correction value exceeds a second preset threshold and the charge amount is abnormal, an alarm is triggered; In other cases, normal monitoring is maintained; Among them, the third preset threshold>the second preset threshold>the first preset threshold.

[0013] Preferably, the fault diagnosis based on the abnormal pattern recognition result includes: If the medium uniformity correction value continuously exceeds the preset first threshold value and the flour processing equipment is in a static working mode, humidity-related abnormality diagnosis is triggered.

[0014] Preferably, the fault monitoring and diagnosis system for flour processing equipment includes: Multi-source data acquisition module: The field-grinding electrostatic sensor collects the charge of each link of the equipment in real time, the capacitive humidity sensor obtains humidity parameters, and the cumulative amount is calculated based on the charge accumulation formula; Fault monitoring sequence construction module: constructs a fault monitoring sequence based on charge and humidity data, marks the charge polarity, and extracts charge marking sequence features; Charge pattern classification module: static mode is determined by alternating arrangement, non-alternating mode is dynamic mode, and mixed mode executes incremental marking rule to generate device status characteristic signal; Dynamic threshold anomaly identification module: Combined with media distribution data, it uses dynamic threshold formulas to monitor anomalies in real time; Multi-level fault diagnosis module: adjusts the equipment working mode according to the medium distribution uniformity and charge state; Output optimization module: Generates control suggestions based on anomaly detection results, including environmental parameter adjustment strategies.

[0015] Beneficial effects: The present invention achieves dual optimization of flour processing equipment fault diagnosis and quality control by integrating electrostatic monitoring, humidity control and medium distribution analysis. Based on charge labeling sequence pattern classification and dynamic threshold adjustment, the system can accurately identify abnormal equipment conditions, and synchronously adjust humidity parameters in combination with real-time electrostatic data to effectively balance particle fineness and electrostatic suppression requirements. Compared with the traditional static threshold method, the present invention significantly reduces the risk of false alarms caused by sudden environmental changes or equipment aging through dynamic weight distribution and pattern feedback mechanism. The present invention provides the flour processing industry with an intelligent solution that takes into account flexibility, reliability and energy efficiency, and is suitable for quality control and equipment health management under complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of the fault monitoring and diagnosis method for flour processing equipment of the present invention; Figure 2 It is a structural diagram of the fault monitoring and diagnosis system for flour processing equipment of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] Example 1: A fault monitoring and diagnosis method for flour processing equipment, such as Figure 1 As shown, the following steps are included: S1: The electrostatic data during equipment operation is collected through a field-mill electrostatic sensor, and the humidity data is simultaneously obtained through a capacitive humidity sensor; and the charge accumulation amount of each link is calculated based on the charge accumulation formula; S2: Build a fault monitoring sequence based on electrostatic data and humidity data, extract the time series features of the charge mark sequence, and input it into the pre-trained LSTM model for classification to generate the device status feature signal; Further explanation is that the timing characteristics of the charge marking sequence specifically include: Polarity encoding: mark positive polarity as 1 and negative polarity as -1 to generate a charge label sequence; Static mode detection: If the charge mark sequence satisfies the alternating arrangement rule (such as [1, -1, 1, -1, ...]), it is determined to be a static mode; Dynamic pattern detection: If the charge mark sequence does not satisfy the alternation, a sliding window statistics is performed to calculate the variance of the polarity change within the window. If it is higher than the threshold, it is determined to be a dynamic pattern; The extracted charge mark sequence time series features are input into the pre-trained LSTM model for classification to generate the device status feature signal, including: Timing feature channel: input charge mark sequence , extracting temporal features through a bidirectional LSTM layer; Environmental parameter channel: input static parameters , encoded into a feature vector through a fully connected layer; Feature fusion layer: Use the attention mechanism to fusion the time series feature vector and the environmental feature vector Perform dynamic fusion. First, through the time series projection matrix and the environment projection matrix Respectively and Mapped to the same dimensional space, we get and Then, through the inner product or learnable weight matrix Calculate the similarity score between the two Then, the Sigmoid function is used to map the scores to environmental feature weights Finally, the fusion features are and The weighted sum of For example, in the case of humidity anomaly ( When the similarity score increases, the weight of the environmental feature Increase, the fusion features pay more attention to the abnormalities of environmental parameters, thereby improving the sensitivity of the classification model to changes in working conditions.

[0019] Classification output layer: Uses the Softmax function to output the probability distribution of device status (normal / abnormal / warning).

[0020] The training process is as follows: a time window sliding sampling is used to generate a training set, the loss function is the cross entropy loss, the optimizer is Adam, and the learning rate decay strategy is: .

[0021] in, For a training round ( ) Dynamically adjusted learning rate value, is the initial learning rate, is the parameter, For rounding operation, The number of current training rounds. To prevent overfitting, a Dropout layer is added after the LSTM layer.

[0022] S3: Combine the charge labeling sequence pattern classification results with the media distribution data to identify abnormal patterns using a dynamic threshold adjustment formula. A reinforcement learning agent is then used to optimize the threshold parameters, with device stability as the reward objective. Further explanation is that the above-mentioned reinforcement learning agent is used to optimize the threshold parameters, with device stability as the reward objective, including: Input parameter normalization: charge ,humidity , medium uniformity Perform standardization (such as Z-score normalization); Weight coefficient calculation: Use historical data to train a linear regression model to determine the charge mode weight and humidity sensitivity coefficient ; Pattern classification and threshold parameter association rules: For example, Static mode (alternating sequence): =0.8, reduce the charge weight to tolerate fluctuations; Dynamic mode (non-alternating sequence): =1.2, improve charge sensitivity; Blending modes (incremental): =1.0, maintaining neutral weight. Humidity sensitivity coefficient Fixed to 0.5, calibrated by historical data.

[0023] Dynamic threshold generation: According to the formula , calculate the threshold in real time; Reinforcement Learning Optimization: Designing Reward Functions , is the comprehensive reward value, is the precision weight, is the recall weight, is the correct anomaly recognition rate, For the abnormal capture rate, the threshold parameter is updated through Q-learning 、 , improve equipment stability.

[0024] Normalization formula: , ; in, 、 are the mean and standard deviation of the charge, 、 is the mean and standard deviation of humidity.

[0025] S4: Fault diagnosis is performed based on abnormal pattern recognition results. At the same time, the charge mark sequence, medium distribution and historical data are integrated to train the time series prediction model and trigger the hierarchical response strategy.

[0026] The electrostatic data during equipment operation is collected through a field-milled electrostatic sensor, and the humidity data is obtained synchronously through a capacitive humidity sensor, including: The electrostatic data is the amount of charge generated in each link during the operation of the flour processing equipment; The humidity data is the humidity parameter required for controlling the charge amount during operation of the device.

[0027] The charge accumulation of each link is calculated based on the charge accumulation formula, including: the charge accumulation formula is as follows, in, For the The charge of a device, For the The operating speed coefficient of each device is the ratio of the device speed to the reference speed, in m / s. is the humidity parameter, is the humidity compensation coefficient, is the medium fineness correction factor, is the coupling coefficient of the preceding link, For the The charge of a device, is the attenuation factor.

[0028] To further explain, the formula is divided into two parts, the current device's own charge generation: , the decay effect of the charge of the previous device: the current device (the The amount of charge accumulated per device) Affected by upstream equipment ( devices, < ) The transfer effect of charge, the intensity of which decays exponentially with time. Specifically, Attenuation formula: ; Physical meaning: The previous device transfers charge to the current device through mechanical connection or electromagnetic induction, but the amount of charge transferred will vary with time. Increase and decrease (by the attenuation factor control); Example: If the preceding device exist When the charge is generated = 10, after Δt = 5 minutes (assuming =0.1), its impact on the current device is 10*e -0.1*5 ≈6.07; Characterizes the operating intensity of the equipment. The faster the speed, the greater the amount of frictional static electricity generated. is the humidity parameter, The humidity offset is the amount of humidity. The higher the humidity, the less static electricity is generated. Used to quantify the efficiency of humidity in suppressing static electricity, which is determined by the surface characteristics of the material. is the correction factor for the fineness of the medium. For example, the finer the flour particles, the larger the friction surface area and the stronger the electrostatic intensity. Used to correct the amplification effect of medium fineness on charge, Reflect the The charge of the device is proportional to the current device The transmission efficiency (such as the tightness of the mechanical connection), For the The charge of a device, Used to control the decay rate of the preceding charge effect, The larger it is, the faster the influence of historical charge disappears. For the current device timestamp, For the preceding device timestamp, The time difference between the current device and the previous device.

[0029] Real-time charge generation: the device's own operating status (speed ) and ambient humidity ( ) jointly determine the current amount of static electricity generated, the fineness of the medium ( ) can amplify or weaken this effect.

[0030] Historical charge accumulation: The charge of the previous device is accumulated through mechanical coupling ( ) is passed to the current device, but its effect decays exponentially over time ( ), reflecting the "memory effect" of charge.

[0031] Charge mark sequence generation rule: according to the charge accumulation The positive and negative polarity generates a charge label sequence , the specific rules are: For example, if the charge sequence of a link is [0.5,-0.3,1.2], the corresponding charge label sequence is [1,-1,1].

[0032] Based on the above-mentioned construction of a fault monitoring sequence based on electrostatic data and humidity data, the time series features of the charge mark sequence are extracted and input into the pre-trained LSTM model for classification to generate a device status feature signal, including: Charge polarity marking: The charge generated by each production link of the flour processing equipment is marked with polarity, where the positive charge is marked with a value of 1 and the negative charge is marked with a value of -1; Pattern classification processing: Static mode determination: When the charge mark values are arranged alternately as [1, -1, 1, -1, ...], it is determined to be in static working mode; Dynamic mode determination: When the charge mark sequence does not conform to the alternating arrangement rule, it is determined to be a dynamic working mode; The difference between adjacent marking values in the incremental marking rule increases in step size 1; Mixed-mode processing: When the charge marking sequence does not conform to either the static or dynamic mode definition, the incremental marking rule is executed: Step 1: Initialize the marking parameters and mark the first positive charge position as the reference value 1; Step 2: Process the subsequent charge quantities in the chronological order of the production links, and keep the current mark value when a positive charge is detected; Step 3: In a preset section where no positive charge appears continuously, perform a value increment operation, with each increment step being 1; Step 4: Repeat steps 2 to 3 until the entire sequence analysis is completed; The dynamic mode is defined as a non-alternating charge mark sequence mode, and the difference between adjacent mark values in the incremental mark rule maintains a constant incremental relationship.

[0033] A further explanation is that the basic classification is to use polarity alternation to distinguish between static mode (ideal state) and dynamic mode (non-ideal state).

[0034] For example, the static sequence [1,-1,1,-1] indicates stable equipment operation; the dynamic sequence [1,1,-1,1] indicates the risk of charge accumulation.

[0035] Dynamic pattern refinement: Further determine whether the non-alternating sequence complies with the difference increasing rule (the difference between adjacent markers increases with a step size of 1).

[0036] For example, the sequence [1, 2, 3, 4] satisfies the adjacent difference of +1 and is judged to be a dynamic mode, which may reflect the linear wear of the equipment.

[0037] Hybrid mode fallback: For sequences that do not meet the above conditions, the incremental marking rule is executed and the mark increment is controlled by the preset segment.

[0038] For example, the sequence [1,-1,-1,-1] triggers mixed processing: the initial mark is 1, and then it increases to [1,2,3,4] when there is no positive charge, to help locate the abnormal segment.

[0039] Rule execution example: The original sequence is [1,−1,−1,1], and the processing flow is as follows: Initialize the markers to [1,0,0,0]; The second negative charge is detected (no positive charge), incrementing to [1,2,0,0]; The third negative charge continues to increase to [1,2,3,0]; The fourth positive charge holds the current label [1,2,3,3].

[0040] The final label sequence [1,2,3,3] reflects the anomalous segment (charges 2-3).

[0041] Combining the charge mark sequence pattern classification results with the medium distribution data, abnormal patterns are identified through a dynamic threshold adjustment formula, including: The processed flour is irradiated by a lighting device, and the irradiated flour is photographed by a camera to obtain medium distribution data; Based on the charge polarity labeling result and the medium distribution data, a dynamic threshold adjustment formula is used to dynamically monitor abnormal modes during the operation of flour processing equipment.

[0042] Dynamic threshold adjustment formula is used to dynamically monitor abnormal modes during the operation of flour processing equipment, including: Based on the medium distribution data, a medium uniformity correction value is obtained using a medium uniformity correction formula; The dynamic threshold adjustment formula is as follows: in, is the dynamic threshold, is the current charge, is the reference charge, Set a reference value for humidity, is the charge mode weight coefficient, is the humidity sensitivity coefficient, It is the correction value of medium uniformity.

[0043] To further explain, the formula consists of four parts: Charge base ratio: (the ratio of the current charge to the reference value), charge mode weight: (Adjust the charge influence weight according to the device mode), humidity deviation correction: (compensation coefficient for humidity deviation from the set value), medium uniformity correction value: (Adjustment factor of medium uniformity correction value to threshold value), It is used to reflect the static electricity accumulation level of the equipment in real time. The higher the charge, the higher the threshold value needs to be to capture the anomaly. Used to calibrate the relative change of charge (such as >1 means the charge exceeds the limit). Used to distinguish the equipment operation mode (static / dynamic / mixed), and increase the weight in dynamic mode (such as high-speed operation) ( =1.2), lowered in static mode ( =0.8), Set the reference value for humidity. If the humidity is lower than the set value ( < ), the threshold needs to be increased to compensate for the electrostatic risk, Used to control the adjustment range of humidity deviation to threshold value (such as = 0.5 indicates medium moisture sensitivity), is the medium uniformity correction value (such as , x∈(0,1]), the lower the medium uniformity (x→0), the more significant the threshold correction (M→+∞).

[0044] Charge dominance: charge ratio Reflects the risk of static abnormality, weight Dynamic adjustment by mode.

[0045] Humidity compensation item: humidity deviation When the humidity is positive (insufficient), the threshold is amplified to cope with the risk of increased static electricity; Controls sensitivity to humidity effects.

[0046] Medium uniformity correction value: uneven distribution of medium ( When the value is large, the threshold is further increased to avoid misjudgment due to fluctuations in material status.

[0047] Based on the medium distribution data, a medium uniformity correction formula is used to obtain a medium uniformity correction value, including: the medium uniformity correction formula is as follows: in, is the medium uniformity correction value, is the medium uniformity index, and ∈(0,1].

[0048] Fault diagnosis based on abnormal pattern recognition results, including: If the medium uniformity correction value exceeds a third preset threshold, a shutdown recommendation signal is immediately generated; If the medium uniformity correction value exceeds a second preset threshold and the charge amount is abnormal, an alarm is triggered; In other cases, normal monitoring is maintained; Among them, the third preset threshold>the second preset threshold>the first preset threshold.

[0049] Further explanation is that the priority rule is: the thresholds are strictly increasing (third preset threshold > second preset threshold > first preset threshold), ensuring that serious faults are handled first.

[0050] Hierarchical response logic: Level 3 exception (highest priority): If the medium uniformity correction value exceeds the third threshold, a shutdown recommendation signal is directly generated regardless of the charge state.

[0051] Example: Evenness 150% (third threshold 130%) → Shutdown for maintenance.

[0052] Level 2 abnormality (lowest priority): If the medium uniformity correction value exceeds the second threshold and the charge amount is abnormal, an alarm is triggered and parameters are adjusted.

[0053] Example: Uniformity is 120% (second threshold 110%), and the charge suddenly increases → an alarm is issued to indicate a risk.

[0054] Other situations (default processing): When the above conditions are not met (such as only exceeding the first threshold or the charge is normal), maintain normal monitoring.

[0055] Example: Uniformity 105% (first threshold 100%), charge normal → Continue observing.

[0056] Fault diagnosis based on abnormal pattern recognition results, including: If the medium uniformity correction value continuously exceeds the preset first threshold value and the flour processing equipment is in a static working mode, humidity-related abnormality diagnosis is triggered.

[0057] A further explanation is that when the particle count exceeds the standard continuously in static mode (e.g., exceeding the threshold of 105% three times), humidity correlation diagnosis is triggered (e.g., analyzing the impact of humidity in stages) to locate the cause of the anomaly.

[0058] Example 2: Based on Example 1, a fault monitoring and diagnosis system for flour processing equipment, such as Figure 2 Shown, including: Multi-source data acquisition module: The field-grinding electrostatic sensor collects the charge of each link of the equipment in real time, the capacitive humidity sensor obtains humidity parameters, and the cumulative amount is calculated based on the charge accumulation formula; Fault monitoring sequence construction module: constructs a fault monitoring sequence based on charge and humidity data, marks the charge polarity, and extracts charge marking sequence features; Charge pattern classification module: static mode is determined by alternating arrangement, non-alternating mode is dynamic mode, and mixed mode executes incremental marking rule to generate device status characteristic signal; Dynamic threshold anomaly identification module: Combined with media distribution data, it uses dynamic threshold formulas to monitor anomalies in real time; Multi-level fault diagnosis module: adjusts the equipment working mode according to the medium distribution uniformity and charge state; Output optimization module: Generates control suggestions based on anomaly detection results, including environmental parameter adjustment strategies.

[0059] The above is a detailed introduction to the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of the present application. At the same time, for those skilled in the art, based on the idea of the present application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A fault monitoring and diagnosis method for flour processing equipment, characterized in that: The following steps are involved: S1: The electrostatic data of the equipment during operation is collected through a field-milled electrostatic sensor, and the humidity data is obtained synchronously through a capacitive humidity sensor; And calculate the charge accumulation of each link based on the charge accumulation formula; S2: Build a fault monitoring sequence based on electrostatic data and humidity data, extract the time series features of the charge mark sequence, and input it into the pre-trained LSTM model for classification to generate the device status feature signal; S3: Combine the charge labeling sequence pattern classification results with the media distribution data to identify abnormal patterns using a dynamic threshold adjustment formula. A reinforcement learning agent is then used to optimize the threshold parameters, with device stability as the reward objective. S4: Fault diagnosis is performed based on abnormal pattern recognition results. At the same time, the charge mark sequence, medium distribution and historical data are integrated to train the time series prediction model and trigger the hierarchical response strategy.

2. A method for fault monitoring and diagnosis of flour processing equipment according to claim 1, characterized in that: The method of collecting electrostatic data during operation of the equipment by using a field-milling electrostatic sensor and synchronously obtaining humidity data by using a capacitive humidity sensor includes: The electrostatic data is the amount of charge generated in each link during the operation of the flour processing equipment; The humidity data is the humidity parameter required for controlling the charge amount during operation of the device.

3. A method for fault monitoring and diagnosis of flour processing equipment according to claim 1, characterized in that: The charge accumulation amount of each link is calculated based on the charge accumulation formula, including: the charge accumulation formula is as follows, in, For the The charge of a device, For the The operating speed coefficient of each device is the ratio of the device speed to the reference speed, in m / s. is the humidity parameter, is the humidity compensation coefficient, is the medium fineness correction factor, is the coupling coefficient of the preceding link, For the The charge of a device, is the attenuation factor.

4. A method for fault monitoring and diagnosis of flour processing equipment according to claim 1, characterized in that: The fault monitoring sequence is constructed based on the electrostatic data and humidity data, and the time series features of the charge mark sequence are extracted and input into the pre-trained LSTM model for classification to generate the device status feature signal, including: Charge polarity marking: The charge generated by each production link of the flour processing equipment is marked with polarity, where the positive charge is marked with a value of 1 and the negative charge is marked with a value of -1; Pattern classification processing: Static mode determination: When the charge mark values are arranged alternately as [1, -1, 1, -1, ...], it is determined to be in static working mode; Dynamic mode determination: When the charge mark sequence does not conform to the alternating arrangement rule, it is determined to be a dynamic working mode; The difference between adjacent marking values in the incremental marking rule increases in step size 1; Mixed-mode processing: When the charge marking sequence does not conform to either the static or dynamic mode definition, the incremental marking rule is executed: Step 1: Initialize the marking parameters and mark the first positive charge position as the reference value 1; Step 2: Process the subsequent charge quantities in the chronological order of the production links, and keep the current mark value when a positive charge is detected; Step 3: In a preset section where no positive charge appears continuously, perform a value increment operation, with each increment step being 1; Step 4: Repeat steps 2 to 3 until the entire sequence analysis is completed; The dynamic mode is defined as a non-alternating charge mark sequence mode, and the difference between adjacent mark values in the incremental mark rule maintains a constant incremental relationship.

5. The method for fault monitoring and diagnosis of flour processing equipment according to claim 1, wherein: The method combines the charge mark sequence pattern classification results with the medium distribution data and identifies abnormal patterns through a dynamic threshold adjustment formula, including: The processed flour is irradiated by a lighting device, and the irradiated flour is photographed by a camera to obtain medium distribution data; Based on the charge polarity labeling result and the medium distribution data, a dynamic threshold adjustment formula is used to dynamically monitor abnormal modes during the operation of flour processing equipment.

6. A method for fault monitoring and diagnosis of flour processing equipment according to claim 5, characterized in that: The method of dynamically monitoring abnormal modes during the operation of flour processing equipment using a dynamic threshold adjustment formula includes: Based on the medium distribution data, a medium uniformity correction value is obtained using a medium uniformity correction formula; The dynamic threshold adjustment formula is as follows: in, is the dynamic threshold, is the current charge, is the reference charge, Set a reference value for humidity, is the charge mode weight coefficient, is the humidity sensitivity coefficient, It is the correction value of medium uniformity.

7. A method for fault monitoring and diagnosis of flour processing equipment according to claim 6, characterized in that: The medium uniformity correction formula is used to obtain the medium uniformity correction value based on the medium distribution data, including: the medium uniformity correction formula is as follows: in, is the medium uniformity correction value, is the medium uniformity index, and ∈(0,1].

8. The method for fault monitoring and diagnosis of flour processing equipment according to claim 1, wherein: The fault diagnosis based on the abnormal pattern recognition result includes: If the medium uniformity correction value exceeds a third preset threshold, a shutdown recommendation signal is immediately generated; If the medium uniformity correction value exceeds a second preset threshold and the charge amount is abnormal, an alarm is triggered; In other cases, normal monitoring is maintained; Among them, the third preset threshold>the second preset threshold>the first preset threshold.

9. The method for fault monitoring and diagnosis of flour processing equipment according to claim 1, wherein: The fault diagnosis based on the abnormal pattern recognition result includes: If the medium uniformity correction value continuously exceeds the preset first threshold value and the flour processing equipment is in a static working mode, humidity-related abnormality diagnosis is triggered.

10. A fault monitoring and diagnosis system for flour processing equipment, used to implement the fault monitoring and diagnosis method for flour processing equipment according to any one of claims 1 to 9, characterized in that: include: Multi-source data acquisition module: The field-grinding electrostatic sensor collects the charge of each link of the equipment in real time, the capacitive humidity sensor obtains humidity parameters, and the cumulative amount is calculated based on the charge accumulation formula; Fault monitoring sequence construction module: constructs a fault monitoring sequence based on charge and humidity data, marks the charge polarity, and extracts charge marking sequence features; Charge pattern classification module: static mode is determined by alternating arrangement, non-alternating mode is dynamic mode, and mixed mode executes incremental marking rule to generate device status characteristic signal; Dynamic threshold anomaly identification module: Combined with media distribution data, it uses dynamic threshold formulas to monitor anomalies in real time; Multi-level fault diagnosis module: adjusts the equipment working mode according to the medium distribution uniformity and charge state; Output optimization module: Generates control suggestions based on anomaly detection results, including environmental parameter adjustment strategies.

Images