A method and system for monitoring the running state of a jam-resistant rotor structure of a rotary unloader

Abstract

The application relates to the technical field of equipment state monitoring, and discloses a method and system for monitoring the operation state of a rotor structure of a rotary unloader. The method comprises the following steps: collecting motor current signals, cavity vibration signals, rotor radial runout signals and unloading port discharge flow signals of the rotor structure; constructing a phase grid according to a rotor rotation period to obtain multiple-source monitoring data slices; performing independent component analysis on each slice to separate independent components corresponding to rotor blade adhesion, dynamic balance imbalance and gap wear; constructing rotor blade cleanliness, dynamic balance health and gap health indexes based on the independent components; correcting the indexes by using real-time material characteristic parameters to obtain net health indexes; and inputting the net health indexes into a rotor state evolution model to output a jamming risk probability curve of the anti-jamming rotor structure in a future preset period. The application realizes accurate prediction of the jamming risk.

Description

Technical Field

[0001] This invention relates to the field of equipment condition monitoring technology, and in particular to a method and system for monitoring the operating status of an anti-jamming rotor structure of a rotary feeder. Background Technology

[0002] Rotary feeders are core equipment in powder and granular material conveying systems, and the operational status of their anti-jamming rotor structure directly affects the continuity and stability of the production process. In industries such as mining, chemical, and grain processing, rotary feeders are widely used for quantitative material conveying and sealed unloading. With the continuous improvement of industrial automation, the need for online monitoring and fault early warning of rotary feeder operation is becoming increasingly urgent. Accurately predicting the risk of jamming in the rotor structure can effectively avoid production interruptions caused by equipment jamming, reduce maintenance costs, and improve production efficiency.

[0003] Existing rotary feeder operation status monitoring technologies suffer from the following fundamental technical defects: Time asynchrony and phase deviation exist between multi-source monitoring data, making it impossible to establish true physical correlations between different signals, resulting in severe distortion of fault features extracted from aliased signals; rotor blade adhesion, dynamic imbalance, and clearance wear are three fault modes coupled in the monitoring signals, and existing technologies cannot separate their respective independent feature components, making it difficult to achieve accurate quantitative assessment of different fault types; changes in material properties significantly interfere with monitoring signals, and existing technologies confuse material interference with changes in the rotor's own state, causing health indicators to fluctuate drastically with operating conditions, making it impossible to obtain a stable and pure characterization of the rotor's health status. These defects collectively prevent existing technologies from accurately predicting material jamming risks, hindering the improvement of equipment reliability and intelligence.

[0004] Therefore, this invention proposes a method and system for monitoring the operating status of an anti-jamming rotor structure of a rotary feeder. Summary of the Invention

[0005] This invention provides a method and system for monitoring the operating status of the anti-jamming rotor structure of a rotary feeder. By aligning the phase grid, decoupling independent components, and correcting material coupling, it achieves accurate prediction of the risk of jamming in the rotary feeder, significantly improving the reliability and intelligence level of the equipment operation.

[0006] This invention provides a method for monitoring the operating status of an anti-jamming rotor structure of a rotary feeder, including: The intelligent sensor array collects the motor current signal, cavity vibration signal, rotor radial runout signal, and discharge flow rate signal of the anti-jamming rotor structure of the rotary feeder during the rotation process. A phase grid is constructed based on the rotor rotation cycle. The motor current signal, cavity vibration signal, rotor radial runout signal and discharge flow rate signal are mapped to each phase interval of the phase grid to obtain multi-source monitoring data slices in each phase interval. Independent component analysis is performed on the multi-source monitoring data slices in each phase interval. The independent current component related to rotor blade adhesion is separated from the motor current signal slice, the independent vibration component related to rotor dynamic imbalance is separated from the cavity vibration signal slice, and the independent runout component related to rotor clearance wear is separated from the rotor radial runout signal slice. Based on independent current components, independent vibration components, and independent runout components, rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are constructed respectively. Real-time material characteristic parameters of the material at the feed inlet are collected, and the rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are corrected using the real-time material characteristic parameters to obtain the net blade cleanliness index, net dynamic balance health index, and net clearance health index after removing material interference. The net blade cleanliness index, net dynamic balance health index, and net clearance health index are input into the pre-trained rotor state evolution model, and the output is the probability curve of the jamming risk of the anti-jamming rotor structure in the future preset period.

[0007] Furthermore, independent component analysis is performed on the multi-source monitoring data slices within each phase interval, including: For each phase interval of the motor current signal segmentation, the fast independent component analysis algorithm is used to extract the first independent component characterizing the rotor load fluctuation, which is used as the current independent component related to rotor blade adhesion. For each phase interval, the cavity vibration signal is segmented, and the empirical mode decomposition algorithm is used to extract the intrinsic mode function that characterizes the rotor unbalance vibration. The component in the intrinsic mode function that has the same frequency as the rotor rotation frequency is taken as the vibration independent component related to the rotor dynamic imbalance. For each phase interval, the rotor radial runout signal is segmented, and the wavelet packet decomposition algorithm is used to extract the detail coefficients that characterize the gap fluctuation. The reconstructed signal of the detail coefficients is used as the runout independent component related to rotor gap wear.

[0008] Furthermore, based on independent current components, independent vibration components, and independent runout components, rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are constructed, including: Calculate the cumulative energy of the independent current components over a complete rotor rotation cycle; Retrieve the reference current energy value that matches the current material type and current feed rate from the adaptive benchmark library; The ratio of the accumulated energy value to the reference current energy value is transformed by negative logarithm and used as the rotor blade cleanliness index. Perform Hilbert-Huang transform on the independent vibration components to obtain their instantaneous amplitude and instantaneous frequency. Calculate the instantaneous energy of the independent vibration components based on the instantaneous amplitude and instantaneous frequency; The ratio of instantaneous energy to preset benchmark energy is used as an indicator of rotor dynamic balance health. Extract the peak-to-peak value of the independent jitter component within each rotor rotation cycle; The ratio of peak-to-peak value to the original design clearance of the rotor is used as the rotor clearance wear coefficient. The difference between 1 and the rotor clearance wear coefficient is used as the rotor clearance health index.

[0009] Furthermore, the steps for building the adaptive benchmark library for operating conditions include: Multiple sets of benchmark operating data of the anti-jamming rotor structure under no-load operation were collected. The benchmark operating data included motor current signals, cavity vibration signals and rotor radial runout signals under different material types and different feeding amounts. Each set of baseline operating data is processed according to any of the above methods for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder to obtain the baseline current energy value, baseline vibration energy value and baseline gap value under the corresponding working conditions. The material type, feeding quantity, and operating condition parameters are associated and stored with the corresponding reference current energy value, reference vibration energy value, and reference gap value to build an adaptive benchmark library for operating conditions.

[0010] Furthermore, it also includes the step of updating the adaptive benchmark library for operating conditions online, including: When the anti-jamming rotor structure is detected to be in a fault-free and stable operating state, the motor current signal, cavity vibration signal and rotor radial runout signal under the current working conditions are collected in real time. Calculate the current energy value, vibration energy value, and gap value under the current working condition based on the above method for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder. The current energy value, vibration energy value, and gap value under the current operating condition are weighted and averaged with the corresponding benchmark value in the operating condition adaptive benchmark library. The benchmark value in the operating condition adaptive benchmark library is then updated with the weighted average result.

[0011] Furthermore, the rotor state evolution model is a bidirectional long short-term memory network incorporating an attention mechanism, wherein the bidirectional long short-term memory network includes: The input layer is used to receive the sequences of net leaf cleanliness index, net dynamic balance health index, and net gap health index from historical time periods. A bidirectional long short-term memory layer is used to extract the temporal dependency features of health index sequences from both forward and backward directions; The attention layer is used to perform weighted fusion of temporal dependency features at different time steps, highlighting key time step features that are highly correlated with material risk, and obtaining the weighted fused features. The output layer is used to generate the material jamming risk probability value for each time step within a preset future period based on the weighted fusion features, and connects the material jamming risk probability values ​​of all time steps into a material jamming risk probability curve.

[0012] Furthermore, the rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are corrected using real-time material characteristic parameters to obtain the net blade cleanliness index, net dynamic balance health index, and net clearance health index after removing material interference, including: Collect real-time material characteristic parameters of the material at the feed inlet, including material moisture content, material particle size distribution, and material flow index; Construct a coupling influence matrix between real-time material characteristic parameters and rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index; The rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are corrected using the coupling influence matrix to obtain the net blade cleanliness index, net dynamic balance health index, and net clearance health index after removing material interference.

[0013] Furthermore, it also includes: Based on historical fault data, equipment maintenance records, and material characteristic data, a knowledge graph of the rotor structure's operating status is constructed. The knowledge graph includes a jam fault node, a fault cause node, a rotor structure status node, a material characteristic node, and related edges connecting each node. When the probability curve of material jamming exceeds the preset warning threshold, extract the net blade cleanliness index, net dynamic balance health index and net gap health index at the current moment as query vectors. Retrieve and query historical fault case nodes in the knowledge graph whose vector similarity exceeds a similarity threshold; Based on the retrieved historical failure case nodes and associated failure cause nodes, output the material jamming cause analysis results and corresponding pre-intervention measures suggestions; When a material jamming failure actually occurs, the current net blade cleanliness index, net dynamic balance health index, net clearance health index, real-time material characteristic parameters, and material jamming failure type are added as new failure case nodes to the knowledge graph.

[0014] Furthermore, it also includes: Based on the material jamming risk probability curve, predict the expected moment when the material jamming risk probability will exceed the execution threshold within a preset period in the future; Based on the expected time, the control strategy matching the expected time is retrieved from the pre-control strategy library. The control strategies include speed adjustment strategy, material cleaning device start-up strategy and feed rate adjustment strategy. Before the expected arrival time, a control link is established in advance with the feeder control system, and the retrieved control strategy is loaded. When the expected time is reached, the loaded control strategy is executed through the pre-established control link.

[0015] This invention provides a monitoring system for the operating status of an anti-jamming rotor structure of a rotary feeder, comprising: The intelligent sensor array, including current sensors, vibration sensors, displacement sensors and flow sensors, is used to collect motor current signals, cavity vibration signals, rotor radial runout signals and discharge flow signals of the anti-jamming rotor structure of the rotary feeder during the rotation process. The phase grid mapping module is used to construct a phase grid based on the rotor rotation cycle, and map the motor current signal, cavity vibration signal, rotor radial runout signal and discharge flow signal from the discharge port to each phase interval of the phase grid, thereby obtaining multi-source monitoring data fragments within each phase interval. The independent component separation module is used to perform independent component analysis on the multi-source monitoring data slices in each phase interval. It separates the independent current component related to rotor blade adhesion from the motor current signal slice, the independent vibration component related to rotor dynamic imbalance from the cavity vibration signal slice, and the independent runout component related to rotor clearance wear from the rotor radial runout signal slice. The health index construction module is used to construct rotor blade cleanliness index, rotor dynamic balance health index and rotor clearance health index based on independent current component, independent vibration component and independent runout component, respectively. The material coupling correction module is used to collect real-time material characteristic parameters of the material at the feed inlet, and use the real-time material characteristic parameters to correct the rotor blade cleanliness index, rotor dynamic balance health index and rotor clearance health index, so as to obtain the net blade cleanliness index, net dynamic balance health index and net clearance health index after removing material interference. The risk probability prediction module is used to input the net blade cleanliness index, net dynamic balance health index, and net clearance health index into the pre-trained rotor state evolution model, and output the jamming risk probability curve of the anti-jamming rotor structure in the future preset period.

[0016] The beneficial effects of this invention compared to existing technologies are as follows: It overcomes the fundamental technical defects of existing technologies, such as the inability to establish physical correlation of signals due to asynchronous time and phase deviation of multi-source monitoring data; the lack of specificity in the construction of health indicators due to the mutual coupling of three fault modes—rotor blade adhesion, dynamic imbalance, and clearance wear—in monitoring signals; and the inability to stably represent the true health status of the rotor due to fluctuations in health indicators caused by changes in material properties. By constructing a phase grid to achieve precise spatiotemporal alignment of multi-source signals, by decoupling and separating the three fault modes through independent component analysis, by eliminating material interference through material coupling correction to obtain pure health indicators, and finally by achieving accurate prediction of future material jamming risks through a rotor state evolution model, this invention significantly improves the reliability and intelligence level of the rotary feeder, effectively avoids production interruptions caused by equipment jamming, greatly reduces equipment maintenance costs, and provides key technical support for improving enterprise production continuity and efficiency.

[0017] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in this application.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating the core method of a method for monitoring the operating status of an anti-jamming rotor structure of a rotary feeder, as described in an embodiment of the present invention. Figure 2 This is a flowchart illustrating the material coupling correction process in an embodiment of the present invention. Figure 3 This is a structural diagram of the rotor state evolution model in an embodiment of the present invention. Detailed Implementation

[0020] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] like Figure 1 , Figure 2 , Figure 3 As shown, the present invention provides an embodiment of a method for monitoring the operating status of an anti-jamming rotor structure of a rotary feeder, characterized in that it includes: The intelligent sensor array collects the motor current signal, cavity vibration signal, rotor radial runout signal, and discharge flow rate signal of the anti-jamming rotor structure of the rotary feeder during the rotation process. A phase grid is constructed based on the rotor rotation cycle. The motor current signal, cavity vibration signal, rotor radial runout signal and discharge flow rate signal are mapped to each phase interval of the phase grid to obtain multi-source monitoring data slices in each phase interval. Independent component analysis is performed on the multi-source monitoring data slices in each phase interval. The independent current component related to rotor blade adhesion is separated from the motor current signal slice, the independent vibration component related to rotor dynamic imbalance is separated from the cavity vibration signal slice, and the independent runout component related to rotor clearance wear is separated from the rotor radial runout signal slice. Based on independent current components, independent vibration components, and independent runout components, rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are constructed respectively. Real-time material characteristic parameters of the material at the feed inlet are collected, and the rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are corrected using the real-time material characteristic parameters to obtain the net blade cleanliness index, net dynamic balance health index, and net clearance health index after removing material interference. The net blade cleanliness index, net dynamic balance health index, and net clearance health index are input into the pre-trained rotor state evolution model, and the output is the probability curve of the jamming risk of the anti-jamming rotor structure in the future preset period.

[0022] In this embodiment, the intelligent sensor array is an integrated sensing device composed of current sensors, vibration sensors, displacement sensors and flow sensors. These sensors are installed at key positions of the rotary feeder to collect various physical quantity signals that reflect the operating status of the rotor structure in real time.

[0023] In this embodiment, the rotary feeder is an industrial device for quantitative conveying of powder or granular materials. It contains a rotatable rotor structure, and the continuous rotation of the rotor enables the stable conveying of materials from the feed inlet to the discharge outlet.

[0024] In this embodiment, the anti-jamming rotor structure is the core component of the rotary feeder. Its surface is specially designed to reduce the probability of material adhesion, but during long-term operation, fault states such as blade adhesion, dynamic imbalance or gap wear may still occur.

[0025] In this embodiment, an intelligent sensor array is used to collect motor current signals, cavity vibration signals, rotor radial runout signals, and discharge flow signals from the discharge port of the anti-jamming rotor structure of the rotary feeder during rotation. Specifically, a current sensor is used to collect the operating current waveform of the drive motor, a vibration sensor is used to collect the vibration amplitude and frequency of the cavity wall, a displacement sensor is used to collect the radial distance change between the outer edge of the rotor and the inner wall of the cavity, and a flow sensor is used to collect the instantaneous outflow of material at the discharge port. These signals together constitute multi-source monitoring data reflecting the rotor's operating status.

[0026] In this embodiment, the rotor rotation cycle refers to the time required for the anti-jamming rotor structure to complete one full 360-degree rotation. This time length is determined by the current rotation speed of the rotor and serves as the time reference unit for subsequent signal synchronization analysis.

[0027] In this embodiment, constructing a phase grid based on the rotor rotation cycle means uniformly dividing a complete rotor rotation cycle into multiple continuous phase intervals. Each phase interval corresponds to a specific angle range during the rotor rotation process. The total number of phase grid intervals can be preset according to the monitoring accuracy requirements.

[0028] In this embodiment, the motor current signal, cavity vibration signal, rotor radial runout signal, and discharge flow rate signal are mapped to each phase interval of the phase grid to obtain multi-source monitoring data slices in each phase interval. Specifically, according to the rotor rotation angle corresponding to the acquisition time of each signal, the continuous waveform data of each signal on the time axis is allocated to the corresponding phase interval, so that each phase interval contains data segments of the four signals within that specific angle range.

[0029] In this embodiment, the phase interval refers to each angle range obtained by uniformly dividing a complete rotor rotation cycle. For example, if the 360-degree rotation cycle is divided into 36 phase intervals, then each phase interval corresponds to a rotation angle range of 10 degrees.

[0030] In this embodiment, the independent current component related to rotor blade adhesion refers to the characteristic signal component separated from the motor current signal that can reflect the load fluctuation of the rotor blade due to material adhesion. This component is separated from the normal motor operating current waveform and is the basis for constructing the blade cleanliness index.

[0031] In this embodiment, the vibration independent component related to rotor dynamic imbalance refers to the vibration component separated from the cavity vibration signal whose frequency is exactly the same as the rotor rotation frequency. The magnitude of this component directly reflects the intensity of unbalanced vibration caused by uneven mass distribution of the rotor.

[0032] In this embodiment, the runout independent component related to rotor clearance wear refers to the characteristic signal component separated from the rotor radial runout signal, which characterizes the distance fluctuation between the outer edge of the rotor and the inner wall of the cavity. The fluctuation amplitude of this component directly reflects the degree of rotor clearance wear.

[0033] In this embodiment, rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are constructed based on independent current component, independent vibration component, and independent runout component, respectively. Specifically, blade cleanliness is obtained by calculating the cumulative energy value of the independent current component per unit time and comparing it with the benchmark value; dynamic balance health is obtained by calculating the instantaneous energy of the independent vibration component and comparing it with the benchmark value; and clearance health is obtained by calculating the peak-to-peak value of the independent runout component and comparing it with the original design clearance.

[0034] In this embodiment, the rotor blade cleanliness index is a value ranging from zero to one. The closer the value is to one, the cleaner the rotor blade surface and the less material adhesion. The closer the value is to zero, the more serious the adhesion on the blade surface and the closer it is to a jamming failure state.

[0035] In this embodiment, the rotor dynamic balance health index is a value ranging from zero to one. The closer the value is to one, the better the rotor dynamic balance and the less unbalanced vibration. The closer the value is to zero, the more serious the dynamic imbalance and the more intense the rotor vibration.

[0036] In this embodiment, the rotor clearance health index is a value ranging from zero to one. The closer the value is to one, the less the rotor clearance wear and the smaller the radial runout. The closer the value is to zero, the more severe the clearance wear and the closer the rotor is to contacting the inner wall of the cavity.

[0037] In this embodiment, collecting real-time material characteristic parameters of the material at the feed inlet refers to using a near-infrared sensor, a laser particle size analyzer, and a flowability tester installed at the feed inlet to detect in real time the current moisture content percentage, particle size distribution range, and quantitative index of the ease of material flow of the material entering the rotary feeder.

[0038] In this embodiment, the net blade cleanliness index, net dynamic balance health index, and net clearance health index refer to the pure index values ​​that reflect only the health status of the rotor structure itself after the three health indices are coupled and calculated with real-time material characteristic parameters, and the index fluctuations caused by changes in material characteristics are eliminated.

[0039] In this embodiment, the pre-trained rotor state evolution model refers to a deep learning model constructed using a bidirectional long short-term memory network. During the training phase, the model uses a large amount of historical operating data, taking the sequence of net blade cleanliness index, net dynamic balance health index, and net clearance health index from past time periods as inputs, and using the actual time of material jamming after these historical data as the output label for training, so that the model learns the mapping relationship between the trend of health index changes and the future risk of material jamming. In the application phase, the model receives three net health index sequences from multiple consecutive time points before the current time and outputs the probability values ​​of material jamming risk at multiple future time points.

[0040] In this embodiment, the future preset time period refers to a pre-set time length extending backward from the current moment, such as the next thirty minutes or the next two hours. This time length can be flexibly set according to the importance of the equipment and the maintenance response speed.

[0041] In this embodiment, the jamming risk probability curve is a continuous curve with future time as the horizontal axis and the probability of jamming occurrence as the vertical axis. Each time point on the curve corresponds to a probability value between zero and one hundred, which represents the likelihood of the rotor structure experiencing a jamming failure at that future moment. The curve as a whole reflects the evolution trend of jamming risk over time.

[0042] Furthermore, independent component analysis is performed on the multi-source monitoring data slices within each phase interval, including: For each phase interval of the motor current signal segmentation, the fast independent component analysis algorithm is used to extract the first independent component characterizing the rotor load fluctuation, which is used as the current independent component related to rotor blade adhesion. For each phase interval, the cavity vibration signal is segmented, and the empirical mode decomposition algorithm is used to extract the intrinsic mode function that characterizes the rotor unbalance vibration. The component in the intrinsic mode function that has the same frequency as the rotor rotation frequency is taken as the vibration independent component related to the rotor dynamic imbalance. For each phase interval, the rotor radial runout signal is segmented, and the wavelet packet decomposition algorithm is used to extract the detail coefficients that characterize the gap fluctuation. The reconstructed signal of the detail coefficients is used as the runout independent component related to rotor gap wear.

[0043] In this embodiment, the Fast Independent Component Analysis (FIC) algorithm is a signal processing method based on higher-order statistics. This method finds a linear transformation of the observed signal so that the transformed components are statistically as independent as possible, thereby separating the independent source signal components from the mixed signal.

[0044] In this embodiment, the first independent component characterizing rotor load fluctuation refers to the signal component separated from the motor current signal by the fast independent component analysis algorithm, which reflects the change in rotor rotational resistance caused by material adhering to the blades. The amplitude fluctuation of this component is directly related to the degree of adhesion on the blade surface.

[0045] In this embodiment, for each phase interval of the motor current signal segment, a fast independent component analysis algorithm is used to extract the first independent component characterizing the rotor load fluctuation, which is used as the current independent component related to rotor blade adhesion. Specifically, the motor current data segment in each phase interval is input into the fast independent component analysis algorithm. The algorithm finds the transformation matrix that maximizes the independence between the output components through iterative calculation, and separates the independent component related to the load fluctuation from the mixed signal. This component is used as the current independent component reflecting the degree of blade adhesion.

[0046] In this embodiment, the empirical mode decomposition algorithm is an adaptive signal time-frequency analysis method. This method decomposes a complex signal into multiple intrinsic mode functions arranged from high frequency to low frequency through a screening process. Each intrinsic mode function represents the oscillation mode at different time scales in the signal.

[0047] In this embodiment, for each phase interval, the cavity vibration signal is segmented, and the empirical mode decomposition algorithm is used to extract the intrinsic mode functions that characterize the rotor unbalanced vibration. Specifically, the vibration data segment in each phase interval is input into the empirical mode decomposition algorithm. The algorithm continuously identifies the local extreme points of the signal and fits the upper and lower envelopes to decompose the original vibration signal into a series of intrinsic mode functions. These functions contain the energy distribution information of the rotor vibration at different frequency scales.

[0048] In this embodiment, the component in the intrinsic mode function that has the same frequency as the rotor rotation frequency is taken as the independent vibration component related to the rotor dynamic imbalance. Specifically, from a series of intrinsic mode functions obtained by empirical mode decomposition, the intrinsic mode function whose fluctuation frequency is exactly the same as the current rotation frequency of the rotor is selected. The magnitude of this function directly reflects the intensity of the unbalanced vibration caused by the uneven mass distribution of the rotor.

[0049] In this embodiment, the component in the intrinsic mode function that has the same frequency as the rotor rotation frequency refers to the specific intrinsic mode function whose instantaneous frequency fluctuates around the rotor rotation frequency among the many intrinsic mode functions decomposed from the vibration signal. This component represents the periodic vibration component caused by rotor dynamic imbalance.

[0050] In this embodiment, the wavelet packet decomposition algorithm is a signal processing method based on wavelet transform. This method decomposes the signal layer by layer into a low-frequency approximation part and a high-frequency detail part, and continues to decompose the two parts at each layer, thereby obtaining the fine decomposition results of the signal in different frequency bands.

[0051] In this embodiment, for the rotor radial runout signal slices in each phase interval, the wavelet packet decomposition algorithm is used to extract the detail coefficients representing the gap fluctuations. Specifically, the radial runout data segments in each phase interval are input into the wavelet packet decomposition algorithm. The algorithm decomposes the signal into different frequency bands through multi-level wavelet transform and extracts the detail coefficients containing high-frequency fluctuation information. These detail coefficients reflect the rapidly changing components in the rotor radial runout.

[0052] In this embodiment, the reconstructed signal of the detail coefficients refers to the time-domain signal that is resynthesized after performing inverse wavelet transform on the detail coefficients reflecting the gap fluctuations obtained by wavelet packet decomposition. This signal retains the fluctuation characteristics related to gap wear in the original radial runout signal, while filtering out other irrelevant components.

[0053] In this embodiment, the reconstructed signal of the detail coefficients is used as the independent component of the jump related to rotor clearance wear. Specifically, the time-domain signal obtained by reconstructing the detail coefficients extracted by wavelet packet decomposition is used as the independent component of the jump reflecting the degree of rotor clearance wear. The amplitude and fluctuation pattern of this signal directly characterize the change of the gap between the outer edge of the rotor and the inner wall of the cavity.

[0054] Furthermore, based on independent current components, independent vibration components, and independent runout components, rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are constructed, including: Calculate the cumulative energy of the independent current components over a complete rotor rotation cycle; Retrieve the reference current energy value that matches the current material type and current feed rate from the adaptive benchmark library; The ratio of the accumulated energy value to the reference current energy value is transformed by negative logarithm and used as the rotor blade cleanliness index. Perform Hilbert-Huang transform on the independent vibration components to obtain their instantaneous amplitude and instantaneous frequency. Calculate the instantaneous energy of the independent vibration components based on the instantaneous amplitude and instantaneous frequency; The ratio of instantaneous energy to preset benchmark energy is used as an indicator of rotor dynamic balance health. Extract the peak-to-peak value of the independent jitter component within each rotor rotation cycle; The ratio of peak-to-peak value to the original design clearance of the rotor is used as the rotor clearance wear coefficient. The difference between 1 and the rotor clearance wear coefficient is used as the rotor clearance health index.

[0055] In this embodiment, the complete rotor rotation cycle refers to the total time length during which the anti-jamming rotor structure starts rotating from a certain reference angle, goes through 360 degrees, and returns to the same reference angle. The length of this cycle is determined by the real-time rotation speed of the rotor at the current moment.

[0056] In this embodiment, calculating the cumulative energy value of the independent current component within a complete rotor rotation cycle specifically refers to summing the sum of the squares of all instantaneous values ​​of the independent current component within a complete rotor rotation cycle to obtain the total energy of current fluctuations within that cycle. This cumulative value reflects the amount of extra energy consumed by the rotor due to blade adhesion within a rotation cycle.

[0057] In this embodiment, the adaptive benchmark library is a pre-built database that stores the benchmark current energy value, benchmark vibration energy value, and benchmark gap value collected and calculated by the rotary feeder under no-load stable operation for different material types and different feeding amounts. These benchmark values ​​serve as reference standards for subsequent health index calculations.

[0058] In this embodiment, the current material type refers to the type of material being conveyed by the rotary feeder at the current moment, such as corn, wheat, cement, or plastic granules. This information can be obtained through manual input or automatically by the material identification device at the feed inlet.

[0059] In this embodiment, the current feed rate refers to the actual weight or volume of material conveyed by the rotary feeder per unit time at the current moment. This data is obtained in real time by the discharge flow sensor at the feed port.

[0060] In this embodiment, a reference current energy value matching the current material type and current feeding quantity is retrieved from the adaptive benchmark library. Specifically, based on the material type identified at the current moment and the measured feeding quantity, the reference current energy value pre-stored under the corresponding working condition is searched in the adaptive benchmark library, and this value is used as the standard current energy consumption when the rotor is in a healthy state under the current working condition.

[0061] In this embodiment, the ratio of the accumulated energy value to the reference current energy value, after undergoing a negative logarithmic transformation, is used as the rotor blade cleanliness index. Specifically, this involves first calculating the accumulated current energy value within the current rotor rotation cycle, dividing it by the reference current energy value under the corresponding operating condition to obtain a ratio, then taking the natural logarithm of this ratio and taking its negative value, which is then divided by a reference value. This ensures that the calculated index value is between zero and one. The smaller the ratio, the closer the cleanliness index is to one, indicating cleaner blades. The negative logarithmic transformation formula is: The blade cleanliness index is calculated as -ln(S / S0) / ln(C), where S is the current energy accumulation value, S0 is the reference current energy value, and C is the calibration constant (taken as 10, ensuring that the index value is 0-1 when S / S0∈(0,10)).

[0062] In this embodiment, the vibration independent components are subjected to Hilbert-Huang transform to obtain the instantaneous amplitude and instantaneous frequency of the vibration independent components. Specifically, the vibration independent components are first subjected to empirical mode decomposition to obtain intrinsic mode functions, and then Hilbert transform is performed on each intrinsic mode function. The instantaneous amplitude and instantaneous frequency at each moment are calculated from the analytical signal obtained after the transformation.

[0063] In this embodiment, the instantaneous energy of the vibration independent component is calculated based on the instantaneous amplitude and instantaneous frequency. Specifically, the instantaneous amplitude at each moment is squared to obtain the instantaneous energy value at that moment. This energy value reflects the vibration intensity of the rotor caused by dynamic imbalance at the current moment.

[0064] In this embodiment, the preset reference energy is a fixed value that represents the standard energy level of the vibration independent component when the rotor is running in a fully dynamic balanced state. It is usually determined by measuring the vibration energy of a newly manufactured or recently dynamically balanced rotor when it is running under no-load conditions.

[0065] In this embodiment, the ratio of instantaneous energy to preset reference energy is used as the rotor dynamic balance health index. Specifically, it means calculating the instantaneous energy of the independent vibration component at the current moment and dividing it by the preset reference energy to obtain a ratio. The closer the ratio is to one, the closer the current vibration energy is to the standard level, and the better the rotor dynamic balance state.

[0066] In this embodiment, the peak-to-peak value of the independent runout component is extracted in each rotor rotation cycle. Specifically, for each complete rotor rotation cycle, the maximum and minimum values ​​of the independent runout component in that cycle are found, and the difference between the two is calculated. This difference reflects the maximum change in radial runout of the rotor in one rotation cycle.

[0067] In this embodiment, the ratio of peak-to-peak value to the original design clearance of the rotor is used as the rotor clearance wear coefficient. Specifically, the extracted peak-to-peak value is divided by the original design clearance value of the rotor in a brand-new state to obtain a ratio between zero and one. The larger the ratio, the more severe the clearance wear.

[0068] In this embodiment, the original design clearance of the rotor refers to the standard distance between the outer edge of the rotor and the inner wall of the cavity, which is designed and determined at the factory. This value is a fixed constant and serves as a reference for subsequent calculation of the degree of clearance wear.

[0069] In this embodiment, the difference between 1 and the rotor clearance wear coefficient is used as the rotor clearance health index. Specifically, the calculated rotor clearance wear coefficient is subtracted from the value 1, so that the health index takes a value between zero and one. The closer the wear coefficient is to zero, the closer the health index is to one, indicating that the clearance wear is more slight.

[0070] Furthermore, the steps for building the adaptive benchmark library for operating conditions include: Multiple sets of benchmark operating data of the anti-jamming rotor structure under no-load operation were collected. The benchmark operating data included motor current signals, cavity vibration signals and rotor radial runout signals under different material types and different feeding amounts. Each set of baseline operating data is processed according to any of the above methods for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder to obtain the baseline current energy value, baseline vibration energy value and baseline gap value under the corresponding working conditions. The material type, feeding quantity, and operating condition parameters are associated and stored with the corresponding reference current energy value, reference vibration energy value, and reference gap value to build an adaptive benchmark library for operating conditions.

[0071] In this embodiment, the no-load operation state refers to the operating mode in which no material passes through the rotary feeder and only the rotor structure rotates idly under the drive of the motor. At this time, the collected monitoring signals only reflect the operating characteristics of the rotor structure itself and are not affected by material flow and material characteristics.

[0072] In this embodiment, multiple sets of benchmark operating data of the anti-jamming rotor structure under no-load operation are collected. Specifically, when the rotating feeder is not conveying any material, the equipment is started and the rotor structure is rotated stably. At the same time, multiple sets of motor current signals, cavity vibration signals and rotor radial runout signals are collected through the intelligent sensor array. Each set of data corresponds to different simulated working conditions.

[0073] In this embodiment, the motor current signal, cavity vibration signal, and rotor radial runout signal under different material types and different feeding amounts refer to the three signals collected under each simulated operating condition by setting different material type identifiers and different feeding amount values ​​in an unloaded operating state. Although these signals are collected under unloaded conditions, they are labeled as reference data corresponding to specific material types and feeding amounts.

[0074] In this embodiment, each set of reference operating data is processed according to any of the above-mentioned methods for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder to obtain the reference current energy value, reference vibration energy value, and reference gap value under the corresponding working condition. Specifically, the reference operating data collected under each no-load condition are sequentially processed through phase grid mapping, independent component analysis, and index construction steps to finally calculate the cumulative current energy value, instantaneous vibration energy value, and radial runout peak value under the simulated working condition, and these values ​​are used as the reference values ​​for the working condition.

[0075] In this embodiment, the reference current energy value, reference vibration energy value, and reference gap value under the corresponding working condition refer to three values ​​obtained after a complete processing flow of no-load operation data under simulated working conditions with a specific material type and a specific combination of feeding amount. They respectively represent the standard current energy level, standard vibration energy level, and standard gap fluctuation level when the rotor is in a healthy state under this working condition.

[0076] In this embodiment, the material type, feeding quantity, and operating condition parameters are associated and stored with the corresponding reference current energy value, reference vibration energy value, and reference gap value to construct an adaptive benchmark library. Specifically, a database table is established, and each row of the table contains three fields as indexes, namely the material type and feeding quantity value, and three fields as data values, namely the reference current energy value, reference vibration energy value, and reference gap value under that operating condition. In this way, a benchmark database that can be called for subsequent real-time monitoring is formed.

[0077] Furthermore, it also includes the step of updating the adaptive benchmark library for operating conditions online, including: When the anti-jamming rotor structure is detected to be in a fault-free and stable operating state, the motor current signal, cavity vibration signal and rotor radial runout signal under the current working conditions are collected in real time. Calculate the current energy value, vibration energy value, and gap value under the current working condition based on the above method for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder. The current energy value, vibration energy value, and gap value under the current operating condition are weighted and averaged with the corresponding benchmark value in the operating condition adaptive benchmark library. The benchmark value in the operating condition adaptive benchmark library is then updated with the weighted average result.

[0078] In this embodiment, detecting that the anti-jamming rotor structure is in a fault-free and stable operating state means that the real-time monitoring system determines that the rotor is not jammed, the blade adhesion is slight, the dynamic balance is good, and the gap wear does not exceed the threshold. At the same time, all monitoring indicators remain stable without significant fluctuations within a certain period of time. At this time, the rotor is in a healthy and reliable operating stage.

[0079] In this embodiment, real-time acquisition of motor current signal, cavity vibration signal and rotor radial runout signal under the current operating conditions means that after confirming that the rotor is in a fault-free and stable operating state, the intelligent sensor array is immediately started to synchronously acquire the motor current waveform, cavity vibration waveform and rotor radial runout waveform at the current moment, so as to obtain the latest real-time data reflecting the rotor operating status under the current operating conditions.

[0080] In this embodiment, the current energy value, vibration energy value and gap value under the current working condition are calculated according to the above-mentioned method for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder. Specifically, the real-time collected motor current signal, cavity vibration signal and rotor radial runout signal are sequentially processed through phase grid mapping, independent component analysis and index construction steps to finally calculate the current energy accumulation value, vibration instantaneous energy value and radial runout peak value under the current working condition.

[0081] In this embodiment, the current energy value, vibration energy value, and gap value under the current working condition are weighted and averaged with the corresponding benchmark values ​​in the working condition adaptive benchmark library. The benchmark values ​​in the working condition adaptive benchmark library are then updated using the weighted average result. Specifically, the original benchmark current energy value, benchmark vibration energy value, and benchmark gap value that match the current material type and current feed amount are first retrieved from the working condition adaptive benchmark library. Then, the newly calculated current energy value, vibration energy value, and gap value are weighted and averaged with the corresponding original benchmark values ​​according to preset weights. Finally, the calculated weighted average value is used to replace the original benchmark value of the corresponding working condition in the working condition adaptive benchmark library, thereby realizing the online dynamic update of the benchmark library.

[0082] Furthermore, the rotor state evolution model is a bidirectional long short-term memory network incorporating an attention mechanism, wherein the bidirectional long short-term memory network includes: The input layer is used to receive the sequences of net leaf cleanliness index, net dynamic balance health index, and net gap health index from historical time periods. A bidirectional long short-term memory layer is used to extract the temporal dependency features of health index sequences from both forward and backward directions; The attention layer is used to perform weighted fusion of temporal dependency features at different time steps, highlighting key time step features that are highly correlated with material risk, and obtaining the weighted fused features. The output layer is used to generate the material jamming risk probability value for each time step within a preset future period based on the weighted fusion features, and connects the material jamming risk probability values ​​of all time steps into a material jamming risk probability curve.

[0083] In this embodiment, the sequence of net blade cleanliness index, the sequence of net dynamic balance health index, and the sequence of net clearance health index within a historical period refers to a set of data formed by arranging the net blade cleanliness index values ​​calculated at multiple consecutive time points before the current moment in chronological order. Similarly, the net dynamic balance health index and net clearance health index at the corresponding time points are also arranged into groups. These three sets of sequences together constitute time-series data reflecting the changes in the historical health status of the rotor structure.

[0084] In this embodiment, extracting the temporal dependency features of the health index sequence from both forward and backward directions means simultaneously inputting three sets of net health index sequences from historical periods into the two processing directions of a bidirectional long short-term memory network. The forward network learns the index change patterns sequentially from early to late time, while the backward network learns the index change patterns in reverse order from late to early time. Finally, the features learned from the two directions are concatenated to obtain complete temporal features that simultaneously contain dependencies in both the past-to-future and future-to-past directions.

[0085] In this embodiment, the temporal dependency features of different time steps are weighted and fused to highlight the key time step features that are highly correlated with the risk of material loss. The weighted fused features are obtained by applying an attention mechanism to the time step features output by the bidirectional long short-term memory network. The attention weight is obtained by calculating the degree of correlation between each time step feature and the final risk of material loss. Time steps with high correlation are given larger weights, and time steps with low correlation are given smaller weights. Then, the features of all time steps are weighted and summed according to their respective weights to obtain a comprehensive feature vector that integrates the information of key time steps.

[0086] In this embodiment, the key time step features that are highly correlated with the risk of material jamming refer to the features corresponding to those time points in the historical health index sequence that have an important indicative role in predicting whether material jamming will occur in the future. For example, the starting point of a rapid decline in the health index, the time point when the index value falls below a certain threshold, or the time point when the index fluctuation pattern changes significantly. The features of these time points can more effectively reflect the changing trend of material jamming risk.

[0087] In this embodiment, the probability value of material jamming risk at each time step within a future preset period is generated based on the weighted fused features. Specifically, the weighted fused feature vector output from the attention layer is input into the fully connected network of the output layer. The network maps the features to a probability value between zero and one through nonlinear transformation, and outputs an independent probability value for each time point within the future preset period, representing the likelihood of the rotor structure experiencing a material jamming failure at that future time point.

[0088] In this embodiment, the probability values ​​of material jamming risk at all time steps are connected to form a material jamming risk probability curve. Specifically, the probability values ​​of material jamming risk at each time point within a future preset time period generated by the output layer are arranged in chronological order to form a continuous curve with future time as the horizontal axis and the probability of material jamming as the vertical axis. This curve can intuitively show the evolution trend of material jamming risk over time.

[0089] Furthermore, the rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are corrected using real-time material characteristic parameters to obtain the net blade cleanliness index, net dynamic balance health index, and net clearance health index after removing material interference, including: Collect real-time material characteristic parameters of the material at the feed inlet, including material moisture content, material particle size distribution, and material flow index; Construct a coupling influence matrix between real-time material characteristic parameters and rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index; The rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are corrected using the coupling influence matrix to obtain the net blade cleanliness index, net dynamic balance health index, and net clearance health index after removing material interference.

[0090] In this embodiment, real-time material characteristic parameters of the material at the feed inlet are collected. These real-time material characteristic parameters include material moisture content, material particle size distribution, and material flow index. Specifically, this means that the percentage of water currently contained in the material is detected in real time by a near-infrared moisture meter installed at the feed inlet, the size distribution range of the material particles is detected in real time by a laser particle size analyzer, and the quantitative value of the ease of flow of the material under standard conditions is detected in real time by a flow performance tester.

[0091] In this embodiment, the material moisture content refers to the percentage of water mass in the material relative to the total mass of the material. This value directly affects the adhesion tendency between the material and the rotor blades; a higher moisture content makes it easier for the material to adhere to the blade surface. The material particle size distribution refers to the size range of the material particles and their proportion, such as the proportion of fine powder with a diameter less than 0.1 mm, medium-sized particles from 0.1 to 1 mm, and coarse particles larger than 1 mm. Particle size distribution affects the material's flowability and wear characteristics within the rotor gap. The material flow index is a dimensionless quantitative value calculated by measuring the time or angle it takes for the material to pass through a screen with a specified aperture under standard vibration conditions. A higher value indicates better material flowability and less likelihood of accumulation within the rotor cavity.

[0092] In this embodiment, a coupling influence matrix is ​​constructed between real-time material characteristic parameters and rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index. Specifically, based on historical operating data, statistical analysis is performed to determine the degree of influence of each unit change in material moisture content on rotor blade cleanliness index as the first coefficient, the degree of influence of each unit change in material particle size distribution on rotor clearance health index as the second coefficient, and the degree of influence of each unit change in material flow index on rotor dynamic balance health index as the third coefficient. Then, these three coefficients are used as the diagonal elements of the matrix to construct a three-row, three-column matrix, and the off-diagonal elements of the matrix are set to zero. Specifically, the coupling influence matrix is ​​a 3×3 diagonal matrix, with the diagonal coefficients being the influence coefficients of material moisture content on blade cleanliness k1 (0.02-0.05 / 1% moisture content), material flow index on dynamic balance health k2 (0.01-0.03 / unit flow index), and material particle size distribution on gap health k3 (0.03-0.06 / 10% fine powder percentage), all calibrated through batch material experiments; correction amount = coupling influence matrix × (real-time material characteristic parameters - baseline material characteristic parameters), and net health index = original health index - correction amount.

[0093] In this embodiment, the rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are corrected using a coupling influence matrix to obtain net blade cleanliness index, net dynamic balance health index, and net clearance health index after removing material interference. Specifically, the material interference vector is first calculated based on the degree to which the current real-time material characteristic parameters deviate from the baseline state. This vector is then multiplied by the coupling influence matrix to obtain the material interference correction amount. Finally, the corresponding material interference correction amount is subtracted from the three health indices constructed in the original embodiment to obtain three net health indices that reflect only the health status of the rotor structure itself and are not affected by material characteristics.

[0094] Furthermore, it also includes: Based on historical fault data, equipment maintenance records, and material characteristic data, a knowledge graph of the rotor structure's operating status is constructed. The knowledge graph includes a jam fault node, a fault cause node, a rotor structure status node, a material characteristic node, and related edges connecting each node. When the probability curve of material jamming exceeds the preset warning threshold, extract the net blade cleanliness index, net dynamic balance health index and net gap health index at the current moment as query vectors. Retrieve and query historical fault case nodes in the knowledge graph whose vector similarity exceeds a similarity threshold; Based on the retrieved historical failure case nodes and associated failure cause nodes, output the material jamming cause analysis results and corresponding pre-intervention measures suggestions; When a material jamming failure actually occurs, the current net blade cleanliness index, net dynamic balance health index, net clearance health index, real-time material characteristic parameters, and material jamming failure type are added as new failure case nodes to the knowledge graph.

[0095] In this embodiment, historical fault data refers to all information related to material jamming events recorded by the rotary feeder during its past operation, including the time of each jamming event, the equipment operating parameters at the time of the event, the trend of health indicators before the event, and the cause of the fault confirmed after troubleshooting.

[0096] In this embodiment, the equipment maintenance record refers to a detailed log of all maintenance activities performed on the rotary feeder throughout its entire life cycle, including the time of each maintenance, maintenance type, replaced parts, adjusted parameters, and the equipment operating status evaluation results after maintenance.

[0097] In this embodiment, material characteristic data refers to the historical characteristic information of various materials processed by the rotary feeder, including the moisture content range, particle size distribution characteristics, flow index value, and statistical information of these materials in relation to jamming events in past operations.

[0098] In this embodiment, a knowledge graph of the rotor structure's operating status is constructed based on historical fault data, equipment maintenance records, and material characteristic data. The knowledge graph includes jamming fault nodes, fault cause nodes, rotor structure status nodes, material characteristic nodes, and connecting edges between the nodes. Specifically, each jamming event in the historical fault data is abstracted as a jamming fault node, the various causes of jamming are abstracted as fault cause nodes, the rotor's health indicators at different times are abstracted as rotor structure status nodes, and the various characteristic parameters of the material are abstracted as material characteristic nodes. Then, connecting edges are established based on the causal relationships and correlation between these nodes to form a structured knowledge network.

[0099] In this embodiment, the jamming fault node is a basic unit in the knowledge graph representing a specific jamming event. Each jamming fault node contains information such as the time of the jamming, its severity, and the handling method. The fault cause node is a basic unit in the knowledge graph representing various causes of jamming, such as severe blade adhesion, excessive dynamic balance imbalance, excessive clearance wear, and excessively high material moisture content. The rotor structure state node is a basic unit in the knowledge graph representing the rotor's health status at a specific moment, containing the specific values ​​of the net blade cleanliness index, net dynamic balance health index, and net clearance health index at that moment. The material characteristic node is a basic unit in the knowledge graph representing material characteristic parameters, containing the specific values ​​of the material moisture content, particle size distribution, and material flow index during that operation.

[0100] In this embodiment, the associated edge is a line segment in the knowledge graph that connects different nodes. Each associated edge represents a certain relationship between two nodes. For example, the associated edge between the jamming fault node and the fault cause node indicates that the jamming was caused by the cause. The associated edge between the rotor structure state node and the jamming fault node indicates that jamming occurred after the state. The associated edge between the material characteristic node and the jamming fault node indicates that jamming is likely to occur under the material characteristic.

[0101] In this embodiment, the preset warning threshold is a pre-set probability value, for example, set to 80%. When the probability value at a certain future time point on the material risk probability curve exceeds this value, the system determines that a warning needs to be issued and initiates the subsequent knowledge graph reasoning process.

[0102] In this embodiment, the jamming risk probability curve exceeding the preset warning threshold means that on the future jamming risk probability curve output by the rotor state evolution model, there are one or more time points with probability values ​​higher than the preset warning values, indicating that the rotor has a high probability of jamming failure in the future.

[0103] In this embodiment, the net blade cleanliness index, net dynamic balance health index, and net gap health index at the current moment are extracted as query vectors. Specifically, the three net health index values ​​obtained after material coupling correction at the current moment are combined into a three-dimensional vector, which serves as the query basis for retrieving similar historical cases in the knowledge graph.

[0104] In this embodiment, the similarity threshold is a pre-set similarity value, for example, set to 85%, used to determine whether the similarity between the query vector and historical fault case nodes in the knowledge graph meets the retrieval criteria. This invention preferably uses cosine similarity as the node similarity calculation method, with a preset similarity threshold of 0.85 (i.e., cases are considered similar when cosine similarity ≥ 0.85); Euclidean distance is used as a backup method, with a corresponding preset threshold of 0.2 (cases are considered similar when distance ≤ 0.2). In this embodiment, retrieving historical fault case nodes in the knowledge graph whose similarity to the query vector exceeds a similarity threshold specifically means calculating the similarity between the current query vector and the rotor structure state nodes associated with all historical fault case nodes in the knowledge graph, and selecting all historical fault case nodes with similarity higher than a preset threshold as reference cases.

[0105] In this embodiment, the similarity calculation method between historical fault case nodes and query vectors includes calculating the Euclidean distance or cosine similarity between the query vector and the rotor structure state node associated with each historical fault case node. The smaller the distance or the closer the cosine value is, the higher the similarity.

[0106] In this embodiment, the associated fault cause nodes refer to those fault cause nodes that are directly connected to the retrieved historical fault case nodes through association edges. These nodes represent the specific reasons that caused the historical material jam event.

[0107] In this embodiment, based on the retrieved historical failure case nodes and associated failure cause nodes, the analysis results of material jamming causes and corresponding pre-intervention measures are output. Specifically, this means summarizing the failure causes and their frequencies in all similar historical cases, taking the most frequent cause as the main possible cause of the current material jamming risk, and extracting successful handling measures for the cause from historical cases as pre-intervention measures for the current situation.

[0108] In this embodiment, the jamming cause analysis result refers to the judgment conclusion output by the system regarding the possible causes of the current jamming risk, such as severe rotor blade adhesion or excessive rotor clearance wear.

[0109] In this embodiment, the pre-intervention measures suggestion refers to the specific operational suggestions proposed by the system based on historical experience to prevent material jamming, such as suggesting to start the material clearing device, suggesting to reduce the rotor speed, or suggesting to reduce the feed rate.

[0110] In this embodiment, an actual jamming failure means that despite the system's warning and intervention, a jamming event still occurs in the rotor structure. In this case, the system will record all relevant information about the jamming event.

[0111] In this embodiment, the current net blade cleanliness index, net dynamic balance health index, net clearance health index, real-time material characteristic parameters, and jamming fault type are added as new fault case nodes to the knowledge graph. Specifically, after a jamming event occurs, the system automatically combines the three net health index values ​​before the event, the material characteristic parameter values ​​at the time of the event, and the jamming fault type confirmed afterward into a new fault case node, and connects it with the corresponding rotor structure status node and material characteristic node through association edges to realize the self-updating and expansion of the knowledge graph.

[0112] In this embodiment, the jamming fault type refers to a label that classifies the jamming event, such as mild jamming, moderate jamming, severe jamming, or a label that classifies the event based on the specific cause, such as adhesive jamming, foreign object jamming, gap jamming, etc.

[0113] Furthermore, it also includes: Based on the material jamming risk probability curve, predict the expected moment when the material jamming risk probability will exceed the execution threshold within a preset period in the future; Based on the expected time, the control strategy matching the expected time is retrieved from the pre-control strategy library. The control strategies include speed adjustment strategy, material cleaning device start-up strategy and feed rate adjustment strategy. Before the expected arrival time, a control link is established in advance with the feeder control system, and the retrieved control strategy is loaded. When the expected time is reached, the loaded control strategy is executed through the pre-established control link.

[0114] In this embodiment, the execution threshold is a preset probability value that is higher than the warning threshold, for example, set to 95%. When the probability value at a certain future time point on the jamming risk probability curve exceeds this value, the system determines that active control measures must be taken to prevent jamming from occurring.

[0115] In this embodiment, based on the material jamming risk probability curve, the expected moment when the probability of material jamming risk exceeds the execution threshold within a preset time period is predicted. Specifically, this means analyzing the future material jamming risk probability curve output by the rotor state evolution model, finding the future time point corresponding to the first probability value on the curve exceeding the execution threshold, and determining this time point as the moment when active control measures need to be implemented.

[0116] In this embodiment, the expected time refers to the specific point in time predicted based on the material jamming risk probability curve, at which the future material jamming risk probability will exceed the execution threshold, such as three or five minutes after the current time.

[0117] In this embodiment, the pre-control strategy library is a pre-built database that stores various control strategy schemes for different risk levels, different expected times, and different operating conditions. Each strategy scheme includes specific control parameters and execution methods.

[0118] In this embodiment, retrieving a control strategy that matches the expected time from the pre-control strategy library specifically means searching the pre-control strategy library for the most suitable control strategy for the current situation based on the predicted time length between the expected time and the current time, the current risk probability value, and the current material type and feed quantity.

[0119] In this embodiment, the speed adjustment strategy refers to a control scheme that adjusts the rotor rotation speed by changing the speed of the drive motor, for example, appropriately reducing the speed to reduce the impact and adhesion of material to the blades. The cleaning device activation strategy refers to a control scheme that activates a cleaning device such as a vibrator or scraper installed on the rotor cavity wall to remove material that has begun to adhere. The feed rate adjustment strategy refers to a control scheme that reduces or suspends the entry of material into the feeder by linking with the upstream feeding equipment, thereby reducing the rotor load.

[0120] In this embodiment, the preset time point before the expected moment refers to a preparation time reserved before the expected execution moment, such as three or five seconds before the expected moment. This time is used to establish the control link and load the control strategy in advance to ensure that it can be executed immediately at the expected moment.

[0121] In this embodiment, a control link is pre-established with the feeder control system, and the retrieved control strategy is loaded. Specifically, at a preset time point before the expected time arrives, the system actively establishes a communication connection with the feeder control unit and writes the control strategy parameters retrieved from the pre-controlled strategy library into the control unit's buffer, waiting for the execution instruction.

[0122] In this embodiment, when the expected time is reached, the loaded control strategy is executed through the pre-established control link. Specifically, at the exact moment the expected time is reached, the system sends an execution command to the control unit through the established communication connection. The control unit immediately starts operations such as speed adjustment, material clearing device start-up, or feed rate adjustment according to the pre-loaded strategy parameters, thereby realizing proactive intervention against the risk of material jamming.

[0123] This invention provides an implementation method for a monitoring system for the operating status of an anti-jamming rotor structure of a rotary feeder, comprising: The intelligent sensor array, including current sensors, vibration sensors, displacement sensors and flow sensors, is used to collect motor current signals, cavity vibration signals, rotor radial runout signals and discharge flow signals of the anti-jamming rotor structure of the rotary feeder during the rotation process. The phase grid mapping module is used to construct a phase grid based on the rotor rotation cycle, and map the motor current signal, cavity vibration signal, rotor radial runout signal and discharge flow signal from the discharge port to each phase interval of the phase grid, thereby obtaining multi-source monitoring data fragments within each phase interval. The independent component separation module is used to perform independent component analysis on the multi-source monitoring data slices in each phase interval. It separates the independent current component related to rotor blade adhesion from the motor current signal slice, the independent vibration component related to rotor dynamic imbalance from the cavity vibration signal slice, and the independent runout component related to rotor clearance wear from the rotor radial runout signal slice. The health index construction module is used to construct rotor blade cleanliness index, rotor dynamic balance health index and rotor clearance health index based on independent current component, independent vibration component and independent runout component, respectively. The material coupling correction module is used to collect real-time material characteristic parameters of the material at the feed inlet, and use the real-time material characteristic parameters to correct the rotor blade cleanliness index, rotor dynamic balance health index and rotor clearance health index, so as to obtain the net blade cleanliness index, net dynamic balance health index and net clearance health index after removing material interference. The risk probability prediction module is used to input the net blade cleanliness index, net dynamic balance health index, and net clearance health index into the pre-trained rotor state evolution model, and output the jamming risk probability curve of the anti-jamming rotor structure in the future preset period.

[0124] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for monitoring the operating status of an anti-jamming rotor structure of a rotary feeder, characterized in that, include: The intelligent sensor array collects the motor current signal, cavity vibration signal, rotor radial runout signal, and discharge flow rate signal of the anti-jamming rotor structure of the rotary feeder during the rotation process. A phase grid is constructed based on the rotor rotation cycle. The motor current signal, cavity vibration signal, rotor radial runout signal and discharge flow rate signal are mapped to each phase interval of the phase grid to obtain multi-source monitoring data slices in each phase interval. Independent component analysis is performed on the multi-source monitoring data slices in each phase interval. The independent current component related to rotor blade adhesion is separated from the motor current signal slice, the independent vibration component related to rotor dynamic imbalance is separated from the cavity vibration signal slice, and the independent runout component related to rotor clearance wear is separated from the rotor radial runout signal slice. Based on independent current components, independent vibration components, and independent runout components, rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are constructed respectively. Real-time material characteristic parameters of the material at the feed inlet are collected, and the rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are corrected using the real-time material characteristic parameters to obtain the net blade cleanliness index, net dynamic balance health index, and net clearance health index after removing material interference. The net blade cleanliness index, net dynamic balance health index, and net clearance health index are input into the pre-trained rotor state evolution model, and the output is the probability curve of the jamming risk of the anti-jamming rotor structure in the future preset period.

2. The method for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder according to claim 1, characterized in that, Independent component analysis is performed on the multi-source monitoring data slices within each phase interval, including: For each phase interval of the motor current signal segmentation, the fast independent component analysis algorithm is used to extract the first independent component characterizing the rotor load fluctuation, which is used as the current independent component related to rotor blade adhesion. For each phase interval, the cavity vibration signal is segmented, and the empirical mode decomposition algorithm is used to extract the intrinsic mode function that characterizes the rotor unbalance vibration. The component in the intrinsic mode function that has the same frequency as the rotor rotation frequency is taken as the vibration independent component related to the rotor dynamic imbalance. For each phase interval, the rotor radial runout signal is segmented, and the wavelet packet decomposition algorithm is used to extract the detail coefficients that characterize the gap fluctuation. The reconstructed signal of the detail coefficients is used as the runout independent component related to rotor gap wear.

3. The method for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder according to claim 1, characterized in that, Based on independent current components, independent vibration components, and independent runout components, rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are constructed, including: Calculate the cumulative energy of the independent current components over a complete rotor rotation cycle; Retrieve the reference current energy value that matches the current material type and current feed rate from the adaptive benchmark library; The ratio of the accumulated energy value to the reference current energy value is transformed by negative logarithm and used as the rotor blade cleanliness index. Perform Hilbert-Huang transform on the independent vibration components to obtain their instantaneous amplitude and instantaneous frequency. Calculate the instantaneous energy of the independent vibration components based on the instantaneous amplitude and instantaneous frequency; The ratio of instantaneous energy to preset benchmark energy is used as an indicator of rotor dynamic balance health. Extract the peak-to-peak value of the independent jitter component within each rotor rotation cycle; The ratio of peak-to-peak value to the original design clearance of the rotor is used as the rotor clearance wear coefficient. The difference between 1 and the rotor clearance wear coefficient is used as the rotor clearance health index.

4. The method for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder according to claim 3, characterized in that, The steps for building an adaptive benchmark library for operating conditions include: Multiple sets of benchmark operating data of the anti-jamming rotor structure under no-load operation were collected. The benchmark operating data included motor current signals, cavity vibration signals and rotor radial runout signals under different material types and different feeding amounts. Each set of benchmark operating data is processed according to the method for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder according to any one of claims 1 to 3, to obtain the benchmark current energy value, benchmark vibration energy value and benchmark gap value under the corresponding working conditions; The material type, feeding quantity, and operating condition parameters are associated and stored with the corresponding reference current energy value, reference vibration energy value, and reference gap value to build an adaptive benchmark library for operating conditions.

5. The method for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder according to claim 4, characterized in that, It also includes the step of updating the adaptive benchmark library online, including: When the anti-jamming rotor structure is detected to be in a fault-free and stable operating state, the motor current signal, cavity vibration signal and rotor radial runout signal under the current working conditions are collected in real time. The current energy value, vibration energy value and gap value under the current working condition are calculated according to the anti-jamming rotor structure operation status monitoring method of the rotary feeder in claim 4. The current energy value, vibration energy value, and gap value under the current operating condition are weighted and averaged with the corresponding benchmark value in the operating condition adaptive benchmark library. The benchmark value in the operating condition adaptive benchmark library is then updated with the weighted average result.

6. The method for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder according to claim 1, characterized in that, The rotor state evolution model is a bidirectional long short-term memory network that incorporates an attention mechanism, wherein the bidirectional long short-term memory network includes: The input layer is used to receive the sequences of net leaf cleanliness index, net dynamic balance health index, and net gap health index from historical time periods. A bidirectional long short-term memory layer is used to extract the temporal dependency features of health index sequences from both forward and backward directions; The attention layer is used to perform weighted fusion of temporal dependency features at different time steps, highlighting key time step features that are highly correlated with material risk, and obtaining the weighted fused features. The output layer is used to generate the material jamming risk probability value for each time step within a preset future period based on the weighted fusion features, and connects the material jamming risk probability values ​​of all time steps into a material jamming risk probability curve.

7. The method for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder according to claim 1, characterized in that, The rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are corrected using real-time material characteristic parameters to obtain the net blade cleanliness index, net dynamic balance health index, and net clearance health index after removing material interference, including: Collect real-time material characteristic parameters of the material at the feed inlet, including material moisture content, material particle size distribution, and material flow index; Construct a coupling influence matrix between real-time material characteristic parameters and rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index; The rotor blade cleanliness index, rotor dynamic balance health index, and rotor clearance health index are corrected using the coupling influence matrix to obtain the net blade cleanliness index, net dynamic balance health index, and net clearance health index after removing material interference.

8. The method for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder according to claim 1, characterized in that, Also includes: Based on historical fault data, equipment maintenance records, and material characteristic data, a knowledge graph of the rotor structure's operating status is constructed. The knowledge graph includes a jam fault node, a fault cause node, a rotor structure status node, a material characteristic node, and related edges connecting each node. When the probability curve of material jamming exceeds the preset warning threshold, extract the net blade cleanliness index, net dynamic balance health index and net gap health index at the current moment as query vectors. Retrieve and query historical fault case nodes in the knowledge graph whose vector similarity exceeds a similarity threshold; Based on the retrieved historical failure case nodes and associated failure cause nodes, output the material jamming cause analysis results and corresponding pre-intervention measures suggestions; When a material jamming failure actually occurs, the current net blade cleanliness index, net dynamic balance health index, net clearance health index, real-time material characteristic parameters, and material jamming failure type are added as new failure case nodes to the knowledge graph.

9. The method for monitoring the operating status of the anti-jamming rotor structure of the rotary feeder according to claim 1, characterized in that, Also includes: Based on the material jamming risk probability curve, predict the expected moment when the material jamming risk probability will exceed the execution threshold within a preset period in the future; Based on the expected time, the control strategy matching the expected time is retrieved from the pre-control strategy library. The control strategies include speed adjustment strategy, material cleaning device start-up strategy and feed rate adjustment strategy. Before the expected arrival time, a control link is established in advance with the feeder control system, and the retrieved control strategy is loaded. When the expected time is reached, the loaded control strategy is executed through the pre-established control link.

10. A monitoring system for the operating status of an anti-jamming rotor structure of a rotary feeder, characterized in that, include: The intelligent sensor array, including current sensors, vibration sensors, displacement sensors and flow sensors, is used to collect motor current signals, cavity vibration signals, rotor radial runout signals and discharge flow signals of the anti-jamming rotor structure of the rotary feeder during the rotation process. The phase grid mapping module is used to construct a phase grid based on the rotor rotation cycle, and map the motor current signal, cavity vibration signal, rotor radial runout signal and discharge flow signal from the discharge port to each phase interval of the phase grid, thereby obtaining multi-source monitoring data fragments within each phase interval. The independent component separation module is used to perform independent component analysis on the multi-source monitoring data slices in each phase interval. It separates the independent current component related to rotor blade adhesion from the motor current signal slice, the independent vibration component related to rotor dynamic imbalance from the cavity vibration signal slice, and the independent runout component related to rotor clearance wear from the rotor radial runout signal slice. The health index construction module is used to construct rotor blade cleanliness index, rotor dynamic balance health index and rotor clearance health index based on independent current component, independent vibration component and independent runout component, respectively. The material coupling correction module is used to collect real-time material characteristic parameters of the material at the feed inlet, and use the real-time material characteristic parameters to correct the rotor blade cleanliness index, rotor dynamic balance health index and rotor clearance health index, so as to obtain the net blade cleanliness index, net dynamic balance health index and net clearance health index after removing material interference. The risk probability prediction module is used to input the net blade cleanliness index, net dynamic balance health index, and net clearance health index into the pre-trained rotor state evolution model, and output the jamming risk probability curve of the anti-jamming rotor structure in the future preset period.

Images