Intelligent control system and method for an electric hoist

Abstract

This invention belongs to the field of engineering machinery and relates to intelligent control technology for lifting equipment. It addresses the problem that existing technologies cannot perform risk assessment and control decision analysis based on the physical mechanisms, damage rates, and severity of different types of rope tangling. Specifically, it is an intelligent control system and method for an electric hoist. The method first acquires vibration data and drum rotation position data, and performs equal-angle resampling to obtain angular domain vibration data. It then extracts the energy entropy of at least two frequency bands as the first feature vector characterizing the disorder of energy distribution and generates a weighted feature vector. This vector is matched with preset standard templates for multiple rope tangling types to determine the pre-identified rope tangling type and identification confidence level for the current cycle. Finally, based on the pre-identified type and confidence level, the corresponding control command is executed from at least two preset control strategies. This invention achieves refined identification and graded prevention and control of rope tangling faults, significantly improving the safety and reliability of electric hoist operation.

Description

Technical Field

[0001] This invention belongs to the field of engineering machinery and relates to intelligent control technology for lifting equipment, specifically an intelligent control system and method for an electric hoist. Background Technology

[0002] As a key lifting device in modern industry, the arrangement of the wire rope in electric hoists directly affects operational safety and equipment lifespan. Misalignment of the wire rope on the drum, leading to rope tangling, not only accelerates wire rope wear but can also cause heavy objects to fall, resulting in major safety accidents. Therefore, real-time monitoring and intelligent control of the wire rope status during electric hoist operation has been a long-standing focus of this field.

[0003] Existing wire rope condition monitoring methods mostly rely on installing mechanical limit switches near the drum or monitoring abnormal fluctuations in the motor's output torque to determine if rope tangling has occurred. When the limit switch is triggered or the torque exceeds a preset threshold, the control system executes a shutdown protection. In recent years, with the development of sensing technology, some solutions have attempted to introduce vibration sensors to indirectly assess the wire rope condition by analyzing the vibration signals of the hoist body. These methods typically collect vibration signals, calculate their total energy or single entropy value as a characteristic quantity, and compare this characteristic quantity with a fixed alarm threshold. When the characteristic quantity exceeds the threshold, it is determined that a rope tangling fault has occurred, and an alarm or shutdown is executed.

[0004] However, the limitations of the aforementioned existing technologies become increasingly apparent when faced with the complex actual working conditions of electric hoists. Wire rope tangling is not a single-form problem; during operation, it can cause the wire rope to climb onto adjacent strands due to axial slippage (slipping), become embedded in lower rope loops due to compression (rope pressing), or even loosen due to localized torque release (knotting). The physical mechanisms, damage rates, and severity of these three types of tangling are drastically different. Judgment methods based on a single characteristic quantity and a fixed threshold cannot distinguish these differentiated fault types. This "one-size-fits-all" approach leads to problems in practical applications. The dilemma is that if the threshold is set too low to ensure sensitivity, normal vibrations and load fluctuations during equipment operation can easily trigger frequent false alarms, disrupting normal production. If the threshold is set too high to ensure anti-interference capability, the system often reacts slowly to early-stage faults such as rope tangling, which have inconspicuous characteristics and evolve slowly, until serious damage occurs before triggering an alarm, missing the best intervention opportunity. Therefore, accurately identifying and distinguishing different types of early rope tangling faults from aliased vibration signals has become a core issue that urgently needs to be addressed to improve the operational safety and intelligent control level of electric hoists. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent control system and method for electric hoists, which solves the problem that existing technologies cannot perform risk assessment analysis and control decision analysis based on the physical mechanisms, damage speed and degree of harm of different types of tangled ropes. The technical problem to be solved by this invention is: how to provide an intelligent control system and method for electric hoists that can perform risk assessment analysis and control decision analysis based on the physical mechanism, damage speed and degree of harm of different types of tangled ropes.

[0006] The objective of this invention can be achieved through the following technical solutions: In a first aspect, the present invention discloses an intelligent control method for an electric hoist, comprising: Acquire vibration data and drum rotation position data collected during the operation of the electric hoist; Based on the rotation position data, the vibration data is resampled at equal angles to obtain angular domain vibration data corresponding to the drum rotation angle. Based on the angular domain vibration data, the energy entropy of at least two frequency bands is extracted to generate the first feature vector characterizing the disorder of energy distribution in each frequency band. Based on angular domain vibration data, the degree of angular clustering of abnormal vibrations in different rotation angle ranges of each frequency band is determined, and a second feature vector is generated to characterize the correlation between vibration events in each frequency band and the drum rotation angle. Based on the historical second feature vector obtained in the preset period, the weight coefficients of each frequency band in the first feature vector of the current period are dynamically adjusted, and the first feature vector is weighted to obtain the weighted feature vector of the current period. The weighted feature vector is matched with the standard template vectors corresponding to at least two types of tangled ropes, and the tangled rope pre-identification type and corresponding identification confidence level for the current period are determined based on the matching results. Based on the pre-identification type of the tangled rope and the identification confidence level, the corresponding control command is determined and executed from at least two preset control strategies of different levels.

[0007] Secondly, the present invention discloses an intelligent control system for an electric hoist, comprising: The data acquisition module is used to acquire vibration data and drum rotation position data collected during the operation of the electric hoist. The data preprocessing module is used to perform equal-angle resampling of vibration data based on rotational position data and generate a first feature vector characterizing the disorder of energy distribution in each frequency band; and to generate a second feature vector characterizing the correlation between vibration events in each frequency band and the drum rotation angle based on angular domain vibration data. The weighted analysis module is used to dynamically adjust the weight coefficients of each frequency band in the first feature vector of the current period based on the historical second feature vector obtained in the preset period, and to perform weighted processing on the first feature vector to obtain the weighted feature vector of the current period. The confidence analysis module is used to match the weighted feature vector with the standard template vectors corresponding to at least two preset types of tangled ropes, and determine the pre-identification type of tangled ropes in the current period and the corresponding identification confidence based on the matching results. The control decision module is used to determine and execute the corresponding control command from at least two preset control strategies of different levels based on the pre-identification type of the tangled rope and the identification confidence level.

[0008] The present invention has the following beneficial effects: This invention constructs a composite feature that characterizes the state of wire rope from two dimensions: "energy distribution disorder" and "fault location clustering," by extracting the multi-band energy entropy and angular correlation coefficient of the angular domain vibration signal. This overcomes the shortcomings of existing technologies that rely solely on vibration amplitude or single-band energy and cannot distinguish fault types. By matching the weighted feature vector with three preset standard templates—spiral rope tangling, rope biting, and knotting—the invention can accurately output the pre-identified type of rope tangling in the current cycle and the corresponding identification confidence level, providing a precise decision basis for subsequent differentiated control. Based on the severity and evolution of rope tangling types, this invention establishes a tiered control mechanism: For the most instantaneously destructive kinking faults, emergency shutdown is implemented to ensure absolute safety; for wear-accumulating rope biting faults, a dynamic time window confirmation mechanism is introduced, adaptively adjusting the window length based on load torque and recognition confidence, ensuring that faults are not ignored while avoiding frequent shutdowns that affect efficiency; for reversible spiral rope tangling faults, a periodic anomaly index is constructed and compared with a dynamic threshold to trigger automatic correction in the early stages of the fault. This tiered strategy solves the problems of false alarms or delayed intervention caused by the "one-size-fits-all" alarms in existing technologies. This invention dynamically adjusts the frequency band weights of the first feature vector by using the historical second feature vector, enabling the fault identification model to adaptively focus on the sensitive frequency bands that are currently strongly correlated with mechanical defects. Simultaneously, it incrementally updates the benchmark weight coefficients based on the feedback effect of control decisions, allowing the system to continuously optimize its judgment preferences during long-term operation. Furthermore, it dynamically adjusts the vibration data sampling rate and the number of wavelet packet decomposition layers according to fluctuations in identification confidence, rationally allocating computational resources while ensuring analytical accuracy, thus achieving a balance between monitoring performance and operational efficiency. By identifying and precisely intervening in rope tangling faults at an early stage, this invention effectively avoids severe wear on wire ropes caused by the cumulative deviation of the rope guide, reduces fatigue damage to the rope body caused by persistent rope biting, and eliminates rope breakage accidents that may be caused by kinking faults. The hierarchical control strategy reduces unnecessary downtime and extends the service life of the wire rope and rope guide mechanism. At the same time, the accurate output of fault type information provides maintenance personnel with clear repair directions, significantly reducing fault diagnosis time and unplanned downtime losses, and providing complete technical support for the intelligent health management of electric hoists. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart of the method according to Embodiment 1 of the present invention; Figure 2 This is a system block diagram of Embodiment 2 of the present invention. Detailed Implementation

[0011] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] In the complex operating conditions of electric hoists, the orderly winding of the wire rope on the drum is fundamental to achieving safe lifting and stable operation. From a physical perspective, the formation of tangled wire rope is essentially a process of dynamic instability in the rope arrangement system. This is characterized by the disruption of the relative motion between the wire rope and the drum grooves, leading to the generation and transmission of abnormal vibration energy. Ideally, the rope arranger and drum move in tandem, with the wire rope arranged sequentially at a fixed pitch, resulting in a stable distribution of vibration energy in both the frequency and angular domains.

[0013] However, existing technologies lack the ability to finely deconstruct the mechanical state information contained in vibration signals, especially lacking an effective mechanism to jointly characterize the evolution of rope tangling from two dimensions: "energy distribution disorder" and "abnormal event angular clustering." This results in the inability to accurately distinguish between different types of rope tangling: one type is rope biting faults, characterized by continuous friction between the wire rope and adjacent loops, with vibration energy dispersed across a wide frequency band and abnormal events randomly occurring at different angular positions; another type is spiral rope tangling faults, originating from rope arranger jamming, characterized by intermittent impacts in a specific frequency band and abnormal events repeatedly occurring within a fixed angular range. More serious kinking faults manifest as instantaneous spikes in energy entropy across the entire frequency band, with a suddenness and destructiveness far exceeding the first two types. This confusion of fault types prevents the control system from implementing differentiated intervention strategies, leading to premature shutdown of rope biting faults, impacting operational efficiency, or failure to correct spiral rope tangling in a timely manner, thus exacerbating the fault.

[0014] For example, in the actual operation of an electric hoist in a heavy machinery processing workshop, the system detected a persistently high energy entropy in the mid-to-high frequency band through vibration monitoring. Traditional methods would then diagnose this as a rope-biting fault and implement load reduction. However, the actual cause of the fault was a slight deformation of the rope guide rail leading to early-stage spiral rope tangling. Because the correction control was not triggered, the rope guide's offset gradually accumulated, eventually causing severe wear of the wire rope during subsequent heavy-load operation. Simultaneously, when another hoist experienced localized wire rope kinking due to improper operation, the system only triggered an alarm based on excessive vibration amplitude, failing to identify the fundamental difference in energy entropy distribution patterns between this and the rope-biting fault. This resulted in the emergency stop command being executed, but the cause of the fault being incorrectly recorded, affecting the formulation of subsequent maintenance strategies.

[0015] If the above problems are not resolved, the intelligent monitoring system of the electric hoist will continuously lose its ability to accurately identify the condition of the wire rope. Specifically, misjudging rope biting and spiral rope tangling will lead to a mismatch between control strategies and fault types. Reduced load operation will fail to prevent the continued deterioration of the rope guide's deviation, while automatic correction will be ineffective against friction-type faults, resulting in wasted control resources. Although kinking faults trigger emergency shutdowns, the lack of type identification prevents maintenance personnel from obtaining accurate fault diagnosis information. Consequently, the system's response to rope tangling faults will remain at a simple, one-size-fits-all alarm level, failing to achieve tiered prevention and precise intervention. Ultimately, this will lead to shortened wire rope life, increased equipment downtime, and even rope breakage accidents due to undetected kinking, severely restricting the safety and economy of electric hoist operation.

[0016] Example 1: As Figure 1 As shown, an intelligent control method for an electric hoist includes: Step S1: Acquire vibration data and drum rotation position data collected during the operation of the electric hoist; The vibration data specifically refers to triaxial vibration signals collected by an accelerometer installed at the bearing housing of the drum; the rotational position data specifically refers to angular pulse signals collected by an encoder installed at the end of the drum shaft. In the specific implementation of step S1, the system first needs to complete the raw data acquisition of the electric hoist's operating status. In this embodiment, vibration data is acquired through a triaxial accelerometer installed at the drum bearing seat. The sensor's installation position is selected based on the principle of the shortest vibration propagation path, ensuring that it can sensitively capture the weak vibration signals generated when the wire rope interacts with the drum. Considering the characteristic frequency distribution range of wire rope tangling faults, this embodiment configures the accelerometer's sampling rate to 5120Hz, a value sufficient to cover the entire frequency band from low-frequency rope skipping impact to high-frequency friction noise. Rotational position data is acquired through an encoder installed at the drum shaft end. This encoder outputs 1024 pulses per revolution, corresponding to an angular resolution of 360 / 1024≈0.35 degrees, which can accurately locate the drum's angular position at any given time.

[0017] Regarding the synchronization of data acquisition, this embodiment employs a hardware triggering mechanism to ensure time alignment between vibration data and rotational position data. Each time the encoder outputs a pulse, a sampling trigger signal is simultaneously generated. This signal also acts on the data acquisition buffer of the accelerometer, ensuring that each frame of vibration data strictly corresponds to the absolute angular position at the current moment. This synchronization mechanism eliminates phase errors introduced by time drift in subsequent data processing, establishing a precise time reference for angular domain resampling.

[0018] The acquired raw vibration signal needs to be preprocessed to remove irrelevant interference. In this embodiment, a bandpass filter is used to process the triaxial vibration signal, with the filter passband set to 10Hz to 2400Hz. The lower cutoff frequency of 10Hz is used to filter out low-frequency swaying of the electric hoist structure itself and environmental vibration interference. Although these low-frequency components have large amplitudes, they are weakly correlated with wire rope tangling faults. The upper cutoff frequency of 2400Hz is the Nyquist frequency corresponding to a sampling rate of 5120Hz, used to prevent high-frequency aliasing. The filtered vibration signal retains the mid-to-high frequency components where the tangling fault characteristics are most concentrated, while significantly improving the signal-to-noise ratio.

[0019] The processing of rotational position data also requires conversion; the pulse signal output by the encoder is accumulated by a counter to generate an absolute angle value from the start time. Every 1024 pulses accumulated, the system records one complete rotation, resets the angle value to 0, and continues accumulating. At the same time, by calculating the angle change per unit time, the system calculates the instantaneous rotational speed of the drum in real time. This speed information will be used as one of the references for judging the working condition in subsequent steps.

[0020] After the above preprocessing, the system obtains time-aligned vibration signal sequences and angle signal sequences. These two sets of data constitute the basic input for all subsequent analyses. The vibration signals carry the wire rope's state information, while the angle signals provide a reference for locating this information to specific mechanical positions. It is worth noting that this embodiment considers the coordinated cooperation of multiple technical details, such as sensor selection, sampling rate configuration, synchronization mechanism, and filtering processing, during the data acquisition stage. These seemingly basic parameter settings actually provide reliable data quality assurance for subsequent complex feature extraction and pattern recognition.

[0021] Step S2: Based on the rotation position data, the vibration data is resampled at equal angles to obtain the angular domain vibration data corresponding to the drum rotation angle. Based on the angular domain vibration data, the energy entropy of at least two frequency bands is extracted to generate the first feature vector characterizing the disorder of energy distribution in each frequency band. The specific extraction process of energy entropy for at least two frequency bands includes: performing wavelet packet decomposition on the angular domain vibration data to obtain wavelet packet coefficients for multiple frequency bands; for each frequency band, calculating the energy entropy of that frequency band based on the energy proportion of each wavelet packet coefficient within that frequency band, and the energy entropy is negatively correlated with the uniformity of energy distribution within that frequency band.

[0022] After completing the data acquisition and preprocessing in step S1, the system enters the core processing stage in step S2. This stage aims to convert the time-domain vibration signal into the angular-domain vibration signal and extract the first feature vector that can characterize the state of the wire rope.

[0023] Because the rotational speed of an electric hoist fluctuates during actual operation, directly analyzing vibration signals collected at equal time intervals can lead to the same mechanical angle corresponding to a vibration event being stretched or compressed in different revolutions, thus disrupting the fixed correspondence between the vibration signal and the mechanical position. To solve this problem, this embodiment first performs equal-angle resampling on the vibration data based on the rotational position data, mapping the vibration signal from the time domain to the angle domain. Specifically, the system uses 1024 pulses output per encoder revolution as a reference, and resamples the vibration data within the angle range of each revolution from 0 to... The target angle is evenly divided into 1024 equally spaced angular positions, with each angular position corresponding to a sampling point. For each target angular position... The system uses rotational position data to find the precise moment when that angular position occurs. Then, using the vibration signal V(t) in Cubic spline interpolation is performed on nearby data points to calculate the vibration amplitude corresponding to that angular position. Repeat this process for all target angular positions to obtain the complete angular domain vibration signal for that cycle. The subscript k represents the kth rotation cycle. After equal-angle resampling, each rotation's angular domain vibration signal contains 1024 data points, and these points strictly correspond to the same mechanical angular position, laying a unified data benchmark for subsequent frequency band analysis and angular correlation analysis.

[0024] To explore the energy distribution characteristics of angular domain vibration signals at different frequency scales, this embodiment employs wavelet packet decomposition for each angular domain vibration signal. Multi-resolution analysis is performed. Wavelet packet decomposition can simultaneously decompose a signal into a low-frequency approximation and a high-frequency detail, and further decompose both in subsequent layers, thereby achieving fine segmentation of the signal frequency band. This embodiment uses the Daubechies4 wavelet basis (db4), which has good orthogonality and tight support, effectively extracting transient features from vibration signals. The decomposition layer is set to 4 layers. After 4 layers of wavelet packet decomposition, the original signal is decomposed into... Each frequency band node has a uniform bandwidth. Each frequency band node corresponds to a set of wavelet packet coefficients, denoted as... , where i is the frequency band number (i=1,2,...,16). The number of coefficients in this frequency band depends on the length of the original signal and the number of decomposition layers.

[0025] For each frequency band i, first calculate the sum of squares of its wavelet packet coefficients to obtain the energy of that frequency band: The frequency band energy reflects the intensity of the vibration signal within that band, but a simple energy value cannot distinguish whether the energy is concentrated on a few coefficients or dispersed across multiple coefficients, the latter being crucial for identifying different types of tangled ropes. Therefore, this embodiment further calculates the energy entropy of each frequency band to quantify the degree of disorder in the energy distribution. Specifically, for frequency band i, the energy proportion of each wavelet packet coefficient is calculated: Then the energy entropy of that frequency band Defined as ;in It is a very small positive number, used to avoid taking the logarithm of zero. The physical meaning of energy entropy is: if the energy in this frequency band is concentrated on a few coefficients (for example, a strong impact at a single frequency), then... The distribution of energy is highly uneven, resulting in a small entropy value; if the energy is dispersed across multiple coefficients (e.g., broadband frictional noise), then... The distribution is uniform, and the entropy value is relatively large. Therefore, energy entropy can sensitively reflect the energy distribution characteristics of different types of rope tangles in a specific frequency band. For example, rope biting faults usually manifest as broadband friction in the mid-to-high frequency band, resulting in a significant increase in energy entropy in the corresponding frequency band; while spiral rope tangles may manifest as intermittent impacts in a specific frequency band, causing the entropy value to fluctuate between normal and abnormal.

[0026] After performing the above calculations on the angular domain vibration signal of the k-th cycle, we obtain the energy entropy values ​​of 16 frequency bands. Arranging them in frequency band order constitutes the first eigenvector of the k-th cycle: This vector characterizes the degree of disorder in the distribution of vibration energy across 16 frequency bands within the current rotation cycle, serving as a fundamental feature for subsequent identification of rope tangling types. It is noteworthy that the energy entropy calculation is entirely based on the inherent statistical characteristics of the angular domain vibration signal, requiring no prior knowledge; therefore, it can adaptively reflect changes in vibration state under different working conditions. In this embodiment, after each complete rotation, the system calculates and outputs a first feature vector as a digital description of the wire rope's state within that cycle.

[0027] Step S3: Based on the angular domain vibration data, determine the degree of angular clustering of abnormal vibrations in different rotation angle ranges of each frequency band, and generate a second feature vector characterizing the correlation between vibration events in each frequency band and the drum rotation angle; The specific process for determining the degree of angular clustering of abnormal vibrations in different rotation angle intervals of each frequency band includes: dividing the angular domain vibration data of each revolution into multiple continuous angular intervals; for each frequency band, detecting whether the local vibration energy in each angular interval meets the preset anomaly judgment criteria, marking the angular intervals that meet the criteria as abnormal events, and determining the center angle of the interval as the angular position corresponding to the abnormal event; based on the angular positions of the marked abnormal events within a preset number of historical revolutions, calculating the angular correlation coefficient that characterizes the degree of clustering of these angular positions, and using it as the second feature vector element of the frequency band.

[0028] After obtaining the angular domain vibration signal and the first feature vector in step S2, step S3 further explores the distribution pattern of vibration events in the angular domain, aiming to generate a second feature vector characterizing the correlation between vibration events in each frequency band and the drum rotation angle. The core of this step is to reveal whether abnormal vibrations repeatedly occur at fixed mechanical positions, thereby distinguishing between random interference and deterministic faults caused by mechanical structural defects.

[0029] Angle range division and local energy entropy calculation: To achieve the above objectives, it is first necessary to subdivide the angular domain vibration signal of each revolution in spatial dimension. In this embodiment, the angle range of each revolution is 0~ The angle is uniformly divided into Q continuous and non-overlapping angle intervals. The value of Q is chosen to balance angle resolution and computational efficiency; here, Q=32 is selected, meaning that each interval covers an angle span of... / 32≈11.25 degrees. For each interval, the degree of anomaly in its internal vibration will be assessed independently later.

[0030] For each frequency band i, local energy entropy calculation is required for the vibration data within each angular interval. Since wavelet packet decomposition has been completed in step S2, the wavelet packet coefficients of each frequency band are arranged in angular order, so the corresponding wavelet packet coefficients within each interval can be directly extracted. Let the subset of wavelet packet coefficients contained in the k-th circle, the i-th frequency band, and the q-th angular interval (q=1,2,...,32) be (the subset of wavelet packet coefficients is not specified in the original text). ,in The number of coefficients within this interval is determined by dividing the total number of wavelet packet coefficients in this frequency band by the number of intervals Q=32 and then rounding down. The last interval accommodates the remaining coefficients. First, the total energy of the coefficients within this interval is calculated: Where r is the index number of the wavelet packet coefficients within the interval; then the energy proportion of each coefficient is calculated: Then the local energy entropy of this interval is: ;in It is a very small positive number. Local energy entropy reflects the degree of disorder in the distribution of vibration energy within this angular interval. If there is strong impact or friction in this interval, the energy distribution tends to be uniform and the entropy value increases; if it is a stable background vibration, the entropy value is low.

[0031] Anomaly Marking and Angle Position Recording: To identify which angle intervals have experienced significant anomalies, an objective anomaly judgment benchmark needs to be established. This embodiment uses statistical characteristics within a sliding window to dynamically determine the benchmark value. Specifically, the system maintains a benchmark with a length of... (Pick =10) A sliding window of cycles is used. For each frequency band i and each interval q, the local energy entropy sequence of all cycles within the window is collected. Calculate its mean and standard deviation When the local energy entropy of the q-th interval of the current cycle satisfies When the threshold is reached, an abnormal event is considered to have occurred within that interval. This threshold (mean plus 1.5 times the standard deviation) corresponds to a statistically significant confidence level of approximately 93%, ensuring sensitivity to weak anomalies while avoiding frequent mislabeling caused by noise.

[0032] Once a certain interval is marked as an abnormal event, the system records the center angle of that interval as the angular position corresponding to the event. Since each interval covers a fixed range of angles, its center angle can be expressed as: This angle value will be used for subsequent angular clustering analysis. It is worth noting that for each frequency band i, multiple intervals within the same circle may be marked as anomalous events, or no events may occur.

[0033] Calculation of angular correlation coefficient: In order to quantify the degree of clustering of anomalous events in each frequency band in the angular domain, this embodiment is based on the most recent Calculate the angular correlation coefficient for the angular positions of all anomaly events marked within the circle. The design idea of ​​this coefficient is: if the anomaly events are randomly distributed across various angles, their vector sums cancel each other out, and the magnitude approaches 0; if the anomaly events are concentrated at one or several fixed angles, their vector sums are aligned, and the magnitude approaches 1.

[0034] Specifically, for frequency band i, in the k-th analysis period, the data within the sliding window (current cycle and...) is collected. (Circle) All angle positions marked as abnormal events. Let there be M events in total, and the corresponding angle position sequence is... Treat these angular positions as unit vectors The argument is calculated, its vector sum is taken, and the magnitude is then divided by the total number of events M for normalization, resulting in the angle correlation coefficient: Where j is the imaginary unit; if M=0, then there are no abnormal events within the window, and the window is defined. =0. The value range is [0, 1]; the larger the value, the more concentrated the angular distribution of the abnormal event, that is, the vibration abnormality in this frequency band is strongly correlated with a specific mechanical position; conversely, it indicates that the abnormal event is dispersed and is more likely to be random interference.

[0035] Generate the second feature vector: Calculate the angular correlation coefficient for all 16 frequency bands. Then, arranged in frequency band order, they form the second eigenvector of the k-th ring. This vector characterizes the correlation between vibration events in each frequency band of the current cycle and the drum rotation angle, serving as a crucial basis for subsequent dynamic weight adjustment and rope tangling type identification. The second feature vector, together with the first feature vector, constitutes a two-dimensional characterization of the wire rope's state: the first feature vector focuses on the disorder of energy distribution across frequency bands, while the second feature vector focuses on the clustering of abnormal events at different angles; the two complement each other. For example, early-stage spiral rope tangling typically manifests as intermittent impacts in a specific frequency band, with the impact location fixed by the rope guide jamming point, thus significantly increasing its angle correlation coefficient. Conversely, while broadband friction noise generated by rope biting faults has a high energy entropy, its angle correlation coefficient may be relatively low because the friction occurs over a larger angular range. This differentiated response characteristic lays the physical foundation for accurately distinguishing rope tangling types.

[0036] Step S4: Dynamically adjust the weight coefficients of each frequency band in the first feature vector of the current period according to the historical second feature vector obtained in the preset period, and perform weighted processing on the first feature vector to obtain the weighted feature vector of the current period; The specific adjustment process of the weight coefficients of each frequency band in the first feature vector includes: obtaining the angular correlation coefficients of each frequency band in the second feature vector of the previous analysis period; weighting and combining the angular correlation coefficients of each frequency band with the reference weight coefficients of each frequency band to obtain the initial weights of each frequency band; normalizing the initial weights of each frequency band to generate the weight coefficients of each frequency band in the current period; wherein, the reference weight coefficients are preset to the same value when the system starts and are dynamically updated according to the execution effect of subsequent control decisions.

[0037] After obtaining the second feature vector characterizing the correlation between vibration events in each frequency band and the drum rotation angle in step S3, the core task of step S4 is to dynamically adjust the weight coefficients of each frequency band in the first feature vector of the current period based on the second feature vector of the historical periods, thereby generating a weighted feature vector for subsequent type identification. The design concept of this step is that if a certain frequency band has shown a strong correlation with the drum rotation angle in the past, it indicates that the vibration anomaly of that frequency band is more likely to originate from mechanical structural defects rather than random interference. Therefore, a higher weight should be assigned to that frequency band in the type identification of the current period.

[0038] Dynamic generation mechanism of weighting coefficients: In this embodiment, the dynamic adjustment of weighting coefficients is based on the core idea of ​​"historical correlation guiding current judgment". Specifically, for the k-th analysis period, the weighting coefficients of each frequency band are not fixed values, but are derived from the angular correlation coefficients of each frequency band in the second feature vector R(k-1) of the previous period (i.e., the (k-1)-th period). With the preset reference weighting coefficients for each frequency band The decision is made jointly. The physical basis of this design is that the angular correlation coefficient reflects the degree of clustering of abnormal events at mechanical locations. The higher the coefficient, the more likely the vibration anomaly in that frequency band is related to a fixed mechanical location, and therefore the more reliable it is in subsequent identification.

[0039] Calculation of initial weights: First, for each frequency band i, calculate its angular correlation coefficient from the previous cycle. The reference weighting coefficient of this frequency band By performing weighted combination, the initial weights for this frequency band are obtained. : Where α is a preset gain coefficient, which is set to 0.8 in this embodiment to control the contribution of historical correlation information to the current weight. Here is the basic weighting coefficient for frequency band i, which is preset to the same value at system startup (e.g., for all frequency bands). All are set to 0.2) to ensure that each frequency band has an equal basic contribution in the early stage of system operation and when historical information is insufficient. At the same time, this coefficient will be dynamically updated according to the execution effect of subsequent control decisions (see step S6 for details), so that the system has long-term self-learning ability.

[0040] Weight normalization processing: due to the initial weights of each frequency band The initial weights are calculated independently, and their sum is not necessarily 1. Subsequent template matching requires that the weights for each frequency band have a uniform scale. Therefore, the initial weights need to be normalized to obtain the final weight coefficients. : Normalized weight vector satisfy This ensures that the contributions of each frequency band are comparable.

[0041] Generating a weighted feature vector: After obtaining the weighting coefficients for each frequency band, these coefficients are applied to the first feature vector H(k) of the current period. This involves weighting the energy entropy values ​​of each frequency band to obtain the weighted feature vector for the current period. : This vector integrates the energy distribution information of the current cycle (the first feature vector) and the angular correlation information of the historical cycles (introduced through weighting coefficients), and serves as the direct input for subsequent identification of the tangled rope type.

[0042] Scenario 1: If the angular correlation coefficient of the previous cycle of a certain frequency band is close to (e.g.) =0.9, indicating that the abnormal vibrations in this frequency band are concentrated at fixed mechanical locations, possibly caused by mechanical defects such as rope assembly jamming. According to the weighting calculation formula, the initial weight of this frequency band is... This will significantly increase (for example), and its contribution weight in the current cycle after normalization will be much higher than that of other frequency bands. Scenario 2: If the angular correlation coefficient of a frequency band in the previous cycle is close to 0 ( =0.1), indicating that the abnormal vibrations in this frequency band are randomly distributed and more likely to be environmental interference, thus its initial weight is low (e.g., after normalization, its contribution in the current cycle is suppressed. This dynamic adjustment mechanism enables the system to adaptively focus on those frequency bands that are strongly correlated with the mechanical rotation period, thereby improving the ability to identify deterministic faults under complex operating conditions.

[0043] Initialization and updating of benchmark weighting coefficients: It is worth noting that the benchmark weighting coefficients... The initial values ​​are set to be the same, based on the simple principle that "all frequency bands are equal when there is no prior knowledge." However, as system operating experience accumulates, the importance of different frequency bands in fault identification may show statistical differences. For example, some frequency bands may frequently become the bands that contribute the most in multiple correct interventions, indicating that these frequency bands are more stably associated with specific types of tangled ropes. To address this, this embodiment designs a parameter self-learning mechanism in step S6, which dynamically adjusts the benchmark weight coefficients of relevant frequency bands according to the feedback type of each control decision, enabling the system to optimize its judgment preferences over the long term and achieve a more accurate result with use. This closed-loop design ensures that the weight generation mechanism in step S4 not only has instantaneous adaptability but also long-term evolutionary capability.

[0044] Step S5: Match the weighted feature vector with the standard template vectors corresponding to at least two types of tangled ropes, and determine the tangled rope pre-identification type and the corresponding identification confidence level for the current period based on the matching results; The specific process for determining the pre-identification type of tangled rope in the current period and the corresponding identification confidence includes: calculating the similarity between the weighted feature vector and the preset spiral tangled rope template, biting rope template, and knot template respectively; if the highest similarity is lower than the preset benchmark threshold, the current period is determined to be in a normal state; otherwise, the tangled rope type corresponding to the highest similarity is determined as the pre-identification type of tangled rope in the current period; the identification confidence is determined based on the difference between the highest similarity and the second highest similarity and the mean of the frequency band angle correlation coefficients in the second feature vector of the current period.

[0045] After obtaining the weighted feature vector for the current period in step S4, step S5 proceeds to the core identification stage of the tangled rope type. This step matches the weighted feature vector with a pre-constructed standard template vector, determines the pre-identified type of tangled rope for the current period by quantifying the similarity, and calculates the corresponding identification confidence level, providing a basis for subsequent hierarchical control decisions.

[0046] Presetting and Construction of Standard Vector Templates: In this embodiment, standard template vectors corresponding to three types of tangled ropes are preset, namely the spiral tangled rope template. , rope biting template and knot template These templates were not arbitrarily set, but obtained through offline experimental calibration: In a laboratory environment, three typical rope tangling faults were simulated, and using the same processing flow as steps S1 to S4, a large number of weighted feature vectors of fault samples were collected. The average of the sample vectors for each type of fault was taken to obtain the standard template for that type of fault. For example, the spiral rope tangling template reflects the typical distribution pattern of this type of fault in energy entropy across various frequency bands: the entropy value in the mid-to-low frequency band shows an intermittent increase, while the high frequency band is relatively stable. The rope biting template shows a continuously high entropy value in the mid-to-high frequency band, while the knotted template shows an instantaneous spike in the entropy value across the entire frequency band. The template vector is 16-dimensional, consistent with the dimension of the weighted feature vector, ensuring the feasibility of matching.

[0047] Weighted similarity calculation: weighted feature vector for the current period The similarity between the vector and three standard templates is calculated. This embodiment uses cosine similarity as the metric, which effectively measures the directional closeness between two vectors, is insensitive to changes in vector magnitude, and is suitable for pattern matching. The formula for calculating cosine similarity is: ;in Weighted eigenvectors The i-th component, This is the i-th component of the spiral tangled rope template; similarly, it is calculated. and The values ​​of the three similarities are all in the range of [−1, 1]. The closer the value is to 1, the higher the degree of matching between the current feature and the corresponding template.

[0048] Pre-identification type determination: After obtaining three similarities, the system first determines the maximum value among them. And the corresponding type. To distinguish between normal and fault states, this embodiment sets a baseline threshold. The value is 0.5. If < This indicates that the current weighted feature vector has a low degree of matching with all fault templates, and the system determines that the current cycle is in a normal state, i.e., no rope tangling has occurred. If ≥ Then, the type of tangled rope corresponding to the maximum value is determined as the pre-identified type of tangled rope for the current period. For example, if If the maximum exceeds the threshold, then =A tangled spiral.

[0049] Calculation of Recognition Confidence: Determining the type solely based on the highest similarity score is insufficient to comprehensively assess the reliability of the recognition; the clarity of the recognition must also be considered. For example, if the three similarities are very close, even if the highest value exceeds a threshold, the risk of misjudgment is high. Therefore, this embodiment introduces an identification confidence score C(k), which comprehensively reflects two factors: first, the difference between the highest and second-highest similarity scores; a larger difference indicates a clearer recognition result; second, the mean of the angular correlation coefficients of each frequency band in the second feature vector of the current period. This mean characterizes the correlation between the current overall vibration and the mechanical rotation period; a higher mean indicates a stronger angular clustering of abnormal events and a more solid physical basis for the recognition. The formula for calculating the identification confidence score is as follows: ,in It is the second highest value among the three similarity scores. Let R(k) be the components of the second feature vector R(k) in the current period. The confidence level C(k) ranges from [0,1]. The larger the value, the more reliable the recognition result.

[0050] Output of recognition results: Through the above calculations, step S5 finally outputs two key results: the pre-identification type of the tangled rope in the current period. The system also calculates the corresponding recognition confidence level C(k). These two results serve as direct inputs to the hierarchical control decision in step S6. For example, when the pre-identification type is a kink, the system will trigger an emergency shutdown regardless of the confidence level. When the pre-identification type is a rope bite, the confidence level will participate in the dynamic adjustment of the time window. When the pre-identification type is a spiral tangled rope, although the confidence level is not directly used for threshold judgment, it will affect the update weight of the baseline weight coefficient in subsequent parameter self-learning. This design ensures that the recognition results are fully utilized in different control strategies, reflecting the system's refinement and intelligence.

[0051] Step S6: Based on the pre-identification type of the tangled rope and the identification confidence level, determine and execute the corresponding control command from at least two preset control strategies of different levels: If the pre-identified type of tangled rope is Type 1, then execute the emergency stop control command. Type 1 is the kink type. If the rope tangling pre-identification type is the second type, then the continuous timeout judgment is initiated. The time window length is dynamically determined based on the average torque of the current cycle and the identification confidence level. When the second type is continuously identified within the time window, the load reduction operation control command is executed. The second type is the rope biting type. If the pre-identified type of tangled rope is the third type, then a periodic anomaly index is constructed by fusing the first feature vector and the second feature vector. The periodic anomaly index is the weighted sum of the energy entropy and angle correlation coefficient of each frequency band. When the periodic anomaly index exceeds the dynamic threshold updated in real time based on the quantile of historical data, an automatic correction control command is executed. The third type is the spiral tangled rope type. The method also includes a parameter self-learning step: determining the feedback type of the current control decision based on preset judgment rules, and updating the baseline weight coefficients. If the feedback type is positive feedback, then increase the baseline weight coefficient of at least one frequency band that contributed the most when triggering this decision. If the feedback type is negative, then reduce the baseline weight coefficient of at least one frequency band that contributed the most when triggering this decision. The positive feedback type and negative feedback type are determined based on at least one factor among the control command type, vibration data characteristics within a preset time period after execution, and operator reset operation.

[0052] In step S5, the pre-identification type of tangled rope for the current cycle is obtained. After obtaining the corresponding recognition confidence level C(k), step S6 proceeds to the hierarchical control decision-making and parameter self-learning stage. This step executes differentiated control strategies based on different pre-identification types, and simultaneously utilizes feedback information from the control effects to dynamically optimize the baseline weight coefficients of the feature extraction stage, enabling the system to possess adaptive evolutionary capabilities. Furthermore, the system dynamically adjusts data acquisition parameters based on fluctuations in recognition confidence level to balance computational resources and analytical accuracy.

[0053] Hierarchical control decision-making: The system presets control strategies for three types of rope tangling, denoted as Type 1 (knotting), Type 2 (rope biting), and Type 3 (spiral tangling). For each analysis cycle, based on the output of step S5... C(k) executes the corresponding control logic.

[0054] Type 1 (Knot) Handling: If For kink-type faults, regardless of confidence level, the system immediately executes an emergency stop control command. Kink faults often lead to local structural damage to the wire rope, and continued operation may cause a rope breakage accident; therefore, the highest level of safety response is adopted. The emergency stop command includes cutting off the motor power supply, activating the mechanical brake, and displaying the alarm message "Kink Fault - Emergency Stop" on the human-machine interface. The irreversibility of this control strategy also makes it a strong signal for determining the correct intervention during subsequent parameter self-learning.

[0055] Type 2 (rope biting) treatment: If For rope biting issues, the system activates a continuous timeout detection mechanism. Rope biting faults typically manifest as friction between the wire rope and adjacent loops. While less harmful than kinking, prolonged occurrences can accelerate wire rope wear. To avoid frequent shutdowns impacting operational efficiency, the system employs a dynamic time window for verification, only executing a load reduction command when rope biting is continuously detected within the window. Determining the dynamic time window length: Time window length (Unit: revolutions) Based on the average torque of the current cycle The identification confidence level C(k) is dynamically adjusted, and the calculation formula is as follows: ,in Based on the base window length, this embodiment takes... =5 laps, t1 and t2 are adjustment coefficients, take t1=2.0, t2=3.0, This refers to the rated torque of the electric hoist. This is the average torque estimated by the motor current or measured by the torque sensor during the current cycle. The physical meaning of this formula is: when the load is relatively light (… When the load is small or the confidence level is low, the window length is extended to reduce the risk of false alarms; conversely, when the load is heavy and the confidence level is high, the window is shortened to speed up the response. Continuous recognition and judgment: The system maintains a length of The sliding buffer records the pre-identified type for each recent cycle. When the pre-identified type for all cycles in the buffer is "rope biting," it is determined to be a continuous rope biting state, and a load reduction operation control command is executed. The load reduction command includes: reducing the motor operating speed to 70% of the rated speed via the frequency converter, and simultaneously outputting a "rope biting fault - load reduction operation" prompt. If a non-rope biting type occurs within the window period, the buffer is cleared and counting restarts.

[0056] Type 3 (Spiral Tangle) Handling: If For the spiral rope tangling type, the system constructs a periodic anomaly index I(k) and compares it with a dynamic threshold to determine whether to execute an automatic correction control command. Early spiral rope tangling typically manifests as a slight misalignment of the rope arranger; active correction can prevent the fault from worsening.

[0057] By fusing the first eigenvector (energy entropy) and the second eigenvector (angular correlation coefficient), the anomaly index for k periods is defined as I(k): ;in Let be the energy entropy of the i-th frequency band (the first eigenvector element). is the angular correlation coefficient (second eigenvector element) for the i-th frequency band. This index reflects both the degree of disorder in the distribution of vibration energy and its correlation with the mechanical rotation angle. When the spiral rope becomes tangled, the energy entropy of a specific frequency band increases and the abnormal events are concentrated in a fixed angle region, resulting in a significant increase in I(k).

[0058] Dynamic threshold calculation: threshold The system updates the data in real time based on the statistical quantiles of historical anomaly indices to avoid the inability of fixed thresholds to adapt to changes in operating conditions. The system maintains a sliding window of length Nth = 50 cycles, storing the anomaly index sequence for each cycle within the window. Calculate the quantiles of this sequence. , as the dynamic threshold of the current period : Quantities can be calculated using interpolation, such as the PERCENTILE.INC function in Excel. When I(k) > If the system detects a tangled rope and requires intervention, it will execute an automatic correction control command. Automatic correction is achieved by fine-tuning the stepper motor of the rope arranger's lateral movement mechanism, causing the rope arranger to move in the opposite direction by a preset step length (e.g., 0.5mm), while simultaneously recording the correction action for subsequent effect evaluation.

[0059] The system dynamically updates the baseline weight coefficients used in step S4 based on the effect of each control decision. (i=1,2,...,16) to achieve long-term performance optimization. The update mechanism is based on feedback type determination and contribution frequency band identification.

[0060] After each control command is executed, the system determines whether the intervention is a correct intervention (positive feedback) or a false alarm (negative feedback) based on the following rules: Emergency Stop (Knot): If the operator checks and resets the machine after the stop, and there was no manual adjustment of the wire rope before the reset (judged by sensors or operation log), it is considered a correct intervention; if the system alarms again immediately after the operator resets the machine, or the operator marks it as a false alarm, it is considered a false alarm. Load reduction operation (rope biting): After executing the load reduction command, monitor the identification results for the next 5 consecutive cycles. If the pre-identified type within 5 cycles is not rope biting, and the average vibration energy entropy decreases by more than 20%, it is considered a correct intervention; otherwise, it is considered a false alarm. Automatic correction (spiral tangled rope): After correction, monitor the abnormality index I(k) for the next 10 consecutive cycles. If I(k) drops below the dynamic threshold and remains stable, it is considered a correct intervention; otherwise, it is considered a false alarm.

[0061] After determining the feedback type, identify the frequency band that contributed the most to triggering this decision. The magnitude of the contribution is measured by the product of each frequency band in the dot product of the weighted feature vector and the corresponding standard template vector. Taking a tangled spiral as an example, let the weighted feature vector... With spiral template The dot product is ; where each product term This reflects the contribution of frequency band i to the matching result. The top M frequency bands with the largest contributions (M=3 in this embodiment) are selected and denoted as the set. .

[0062] right For each frequency band i, the reference weighting coefficient Update as follows: ;in To adjust the step size, this embodiment takes... =0.05; γ is the feedback direction sign, γ=+1 for positive feedback and -1 for negative feedback. After updating, it is necessary to... The frequency bands are limited to the range [0,1], and if they exceed the boundary, the boundary value is used. The reference weight coefficients for other frequency bands remain unchanged. This mechanism gradually increases the weight of frequency bands that are stable in association with faults, while decreasing the weight of frequency bands that are unrelated to faults or prone to false alarms, thereby improving long-term identification accuracy.

[0063] After determining the type of tangled ropes to be identified in the current period and the corresponding identification confidence level, the following steps are also included: Based on the identification confidence level of each period within a preset historical window, the fluctuation statistic of the identification confidence level is calculated. The fluctuation statistic is used to characterize the dispersion of the identification confidence level in each period. The data collection status for the current period is determined based on the identification confidence level and the confidence level fluctuation statistics. Adjust the vibration data sampling rate and / or wavelet packet decomposition level for subsequent cycles based on the data acquisition status.

[0064] After adjusting the sampling rate of vibration data in subsequent cycles and / or the number of wavelet packet decomposition levels, the following is also included: The adjusted sampling rate and / or wavelet packet decomposition level will be applied to the vibration data acquisition and wavelet packet decomposition processing in the next analysis cycle.

[0065] Confidence fluctuation analysis and adaptive sampling rate: To further improve the adaptability of data acquisition, the system also introduces confidence fluctuation analysis in step S6 to dynamically adjust the sampling rate and wavelet packet decomposition level in subsequent cycles.

[0066] Confidence fluctuation statistic: Maintain a statistic of length A sliding window of 20 cycles is used to store the recognition confidence C(k) for each cycle within the window, and its standard deviation is calculated. As a statistic of fluctuation: ;in The mean of the built-in confidence level of the window; In the initial stage of system startup, if the current cycle number k < If the existing confidence data from k periods is used, the mean and standard deviation are calculated, but parameter adjustments are not performed until the window is full, at which point the adaptive adjustment mechanism is activated. Alternatively, the initial default parameters (sampling rate 5120Hz, decomposition layers 4) can be used as a baseline, and dynamic adjustments can be made once the window meets the conditions.

[0067] Data acquisition status determination and parameter adjustment: based on the confidence level C(k) and its fluctuation statistics for the current period. The data collection status is divided into three categories: Steady state: C(k) ≥ 0.8 and ≤0.1. At this point, the recognition result is reliable. To save computing resources, the sampling rate is reduced to 80% of the original value (i.e., 4096Hz), while the number of wavelet packet decomposition layers remains unchanged at 4. Fluctuation state: 0.5 ≤ C(k) < 0.8 or 0.1 < ≤0.2. At this point, maintain the current sampling rate (5120Hz) and the number of decomposition layers (4 layers); Unstable state: C(k) < 0.5 or >0.2. At this point, the identification uncertainty is high. To improve the analysis accuracy, the sampling rate is increased to 120% of the original value (i.e., 6144Hz), and the number of wavelet packet decomposition layers is increased to 5 (corresponding to 32 frequency bands). The above thresholds (0.8, 0.5, 0.1, 0.2) are preferred values ​​in this embodiment, and can be adjusted according to specific working conditions in actual applications.

[0068] After parameter adjustment, starting from the next analysis cycle, the system collects vibration data according to the new sampling rate and performs wavelet packet decomposition according to the new number of layers, thereby forming a closed-loop adaptive system.

[0069] The aforementioned hierarchical control, parameter self-learning, and adaptive adjustment mechanisms fully demonstrate the closed-loop characteristics of intelligent control methods.

[0070] Based on the determined data acquisition status, the vibration data sampling rate for the next analysis cycle is adjusted. The wavelet packet decomposition level L is dynamically adjusted. The adjustment rules are as follows: Steady state: the state where the sampling rate is reduced to the baseline value, i.e. ,That The baseline sampling rate is 5120Hz (adjusted to 4096Hz); the number of wavelet packet decomposition layers remains unchanged at L=4. This reduces the amount of data and computational burden while ensuring basic monitoring capabilities. Fluctuation state: Maintain sampling rate = =5120Hz, wavelet packet decomposition level L=4 remains unchanged; Unstable state: The sampling rate is increased to the baseline value, i.e. =6144Hz; at the same time, the number of wavelet packet decomposition layers is increased to L=5 layers. The higher sampling rate can capture vibration information in a wider frequency band, while the deeper wavelet packet decomposition (5 layers corresponding to 32 frequency bands) can provide finer frequency band division, which helps to extract weaker fault features.

[0071] The above-mentioned adjusted sampling rate The wavelet packet decomposition level L will take effect from the next analysis cycle (i.e., the (k+1)th cycle). Specifically, in step S1 of the next cycle, the system acquires vibration data according to the new sampling rate; in step S2, the angular domain vibration signal is decomposed into wavelet packets according to the new decomposition level, and the corresponding number of frequency band energy entropies is extracted (5 levels of decomposition correspond to 32 frequency bands). This adaptive mechanism ensures that the system can dynamically optimize the data acquisition strategy based on the recent fluctuations in identification confidence, saving resources when identification reliability is high and improving analysis accuracy when identification uncertainty is high, thereby maintaining optimal overall performance in long-term operation.

[0072] It should be noted that when the wavelet packet decomposition layer is adjusted from 4 to 5 layers, the dimensions of the first and second feature vectors will change from 16 to 32. To adapt to this change, the standard template vector also needs to be expanded to 32 dimensions accordingly. In this embodiment, the system pre-stores template vectors in both dimensions and automatically switches to use the corresponding dimension template for matching after parameter adjustment. Furthermore, the baseline weight coefficients... The number is also adjusted accordingly, and the baseline weighting coefficient of the new frequency band is initialized to the same value (e.g., 0.2 / 32), and is gradually updated with the feedback of subsequent control decisions.

[0073] Through steps S1 to S6 above, this invention constructs a complete closed loop from vibration signal acquisition to hierarchical control execution. First, the vibration signal is resampled at equal angles based on rotational position data, eliminating the influence of rotational speed fluctuations and ensuring subsequent analysis strictly corresponds to the mechanical angular position. Then, wavelet packet decomposition is used to extract the energy entropy of each frequency band, forming a first feature vector characterizing the disorder of energy distribution. Simultaneously, the angular correlation coefficient is calculated using the angular clustering of abnormal events, forming a second feature vector characterizing the degree of correlation between vibration and mechanical position. The first feature vector is dynamically weighted based on the historical second feature vector, causing the identification model to adaptively focus on frequency bands strongly correlated with mechanical defects. Then, by matching with preset fault templates, three typical fault types—kinking, rope biting, and spiral rope tangling—are accurately distinguished, and the identification confidence level is output. Finally, differentiated control is executed based on the fault type and confidence level: kinking faults result in immediate emergency shutdown; rope biting faults are confirmed through a dynamic time window and then run at reduced load; and spiral rope tangling faults are automatically corrected when the anomaly index exceeds the dynamic threshold. Simultaneously, frequency band weights are continuously optimized through parameter self-learning, and the sampling rate and wavelet packet decomposition level are adaptively adjusted through confidence fluctuation analysis. This invention enables precise identification and graded prevention and control of electric hoist rope tangling faults, effectively avoiding false alarms and missed alarms, significantly improving the safety and reliability of equipment operation, extending the service life of wire ropes, reducing unplanned downtime, and providing a complete technical solution for intelligent health management of industrial lifting equipment.

[0074] Example 2: Figure 2As shown, an intelligent control system for an electric hoist includes: The data acquisition module is used to acquire vibration data and drum rotation position data collected during the operation of the electric hoist. The data preprocessing module is used to perform equal-angle resampling of vibration data based on rotational position data and generate a first feature vector characterizing the disorder of energy distribution in each frequency band; and to generate a second feature vector characterizing the correlation between vibration events in each frequency band and the drum rotation angle based on angular domain vibration data. The weighted analysis module is used to dynamically adjust the weight coefficients of each frequency band in the first feature vector of the current period based on the historical second feature vector obtained in the preset period, and to perform weighted processing on the first feature vector to obtain the weighted feature vector of the current period. The confidence analysis module is used to match the weighted feature vector with the standard template vectors corresponding to at least two preset types of tangled ropes, and determine the pre-identification type of tangled ropes in the current period and the corresponding identification confidence based on the matching results. The control decision module is used to determine and execute the corresponding control command from at least two preset control strategies of different levels based on the pre-identification type of the tangled rope and the identification confidence level.

[0075] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0076] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0077] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An intelligent control method for an electric hoist, characterized in that, include: Acquire vibration data and drum rotation position data collected during the operation of the electric hoist; Based on the rotation position data, the vibration data is resampled at equal angles to obtain angular domain vibration data corresponding to the drum rotation angle. Based on the angular domain vibration data, the energy entropy of at least two frequency bands is extracted to generate a first feature vector characterizing the disorder of energy distribution in each frequency band. Based on the angular domain vibration data, the degree of angular clustering of abnormal vibrations in each frequency band in different rotation angle intervals is determined, and a second feature vector characterizing the correlation between vibration events in each frequency band and the drum rotation angle is generated. Based on the historical second feature vector obtained in the preset period, the weight coefficients of each frequency band in the first feature vector of the current period are dynamically adjusted, and the first feature vector is weighted to obtain the weighted feature vector of the current period. The weighted feature vector is matched with the standard template vectors corresponding to at least two types of tangled ropes, and the tangled rope pre-identification type and corresponding identification confidence level for the current period are determined based on the matching results. Based on the pre-identification type of the tangled rope and the identification confidence level, the corresponding control command is determined and executed from at least two preset control strategies of different levels.

2. The intelligent control method for an electric hoist according to claim 1, characterized in that, The vibration data specifically refers to triaxial vibration signals collected by an accelerometer installed on the drum bearing housing; the rotational position data specifically refers to angular pulse signals collected by an encoder installed on the drum shaft end.

3. The intelligent control method for an electric hoist according to claim 1, characterized in that, The specific extraction process of energy entropy for at least two frequency bands includes: performing wavelet packet decomposition on the angular domain vibration data to obtain wavelet packet coefficients for multiple frequency bands; for each frequency band, calculating the energy entropy of that frequency band based on the energy proportion of each wavelet packet coefficient within that frequency band, wherein the energy entropy is negatively correlated with the uniformity of energy distribution within that frequency band.

4. The intelligent control method for an electric hoist according to claim 1, characterized in that, The specific process for determining the degree of angular clustering of abnormal vibrations in different rotation angle intervals of each frequency band includes: dividing the angular domain vibration data of each revolution into multiple continuous angular intervals; for each frequency band, detecting whether the local vibration energy in each angular interval meets the preset anomaly judgment criteria, marking the angular intervals that meet the criteria as abnormal events, and determining the center angle of the interval as the angular position corresponding to the abnormal event; based on the angular positions of the marked abnormal events within a preset number of historical revolutions, calculating the angular correlation coefficient that characterizes the degree of clustering of these angular positions, and using it as the second feature vector element of the frequency band.

5. The intelligent control method for an electric hoist according to claim 1, characterized in that, The specific adjustment process of the weight coefficients of each frequency band in the first feature vector includes: obtaining the angular correlation coefficients of each frequency band in the second feature vector of the previous period; weighting and combining the angular correlation coefficients of each frequency band with the reference weight coefficients of each frequency band to obtain the initial weights of each frequency band; normalizing the initial weights of each frequency band to generate the weight coefficients of each frequency band in the current period; wherein, the reference weight coefficients are preset to the same value when the system starts and are dynamically updated according to the execution effect of subsequent control decisions.

6. The intelligent control method for an electric hoist according to claim 1, characterized in that, The specific process for determining the pre-identification type of tangled rope in the current period and the corresponding identification confidence level includes: calculating the similarity between the weighted feature vector and the preset spiral tangled rope template, biting rope template, and knot template respectively; if the highest similarity is lower than the preset benchmark threshold, the current period is determined to be in a normal state; otherwise, the tangled rope type corresponding to the highest similarity is determined as the pre-identification type of tangled rope in the current period; the identification confidence level is determined based on the difference between the highest similarity and the second highest similarity and the mean value of the frequency band angle correlation coefficients in the second feature vector of the current period.

7. The intelligent control method for an electric hoist according to claim 1, characterized in that, If the pre-identified type of the tangled rope is the first type, then an emergency stop control command is executed, where the first type is the kink type; If the pre-identified type of the tangled rope is the second type, then a continuous timeout judgment is initiated. The length of the time window is dynamically determined based on the average torque of the current cycle and the identification confidence level. When the second type is continuously identified within the time window, a load reduction operation control command is executed. The second type is the rope biting type. If the pre-identified type of the tangled rope is the third type, then a periodic anomaly index is constructed that integrates the first feature vector and the second feature vector. The periodic anomaly index is the weighted sum of the energy entropy and angle correlation coefficient of each frequency band. When the periodic anomaly index exceeds the dynamic threshold updated in real time based on the statistical quantile of historical data, an automatic correction control command is executed. The third type is the spiral tangled rope type. The method further includes a parameter self-learning step: determining the feedback type of the current control decision according to preset judgment rules, and updating the benchmark weight coefficients. If the feedback type is a positive feedback type, then increase the baseline weight coefficient of at least one frequency band that contributed the most when triggering this decision. If the feedback type is a negative feedback type, then reduce the baseline weight coefficient of at least one frequency band that contributed the most when triggering this decision. The positive feedback type and negative feedback type are determined based on at least one factor among the control command type, vibration data characteristics within a preset time period after execution, and operator reset operation.

8. The intelligent control method for an electric hoist according to claim 6, characterized in that, After determining the type of tangled ropes to be identified in the current period and the corresponding identification confidence level, the following steps are also included: Based on the identification confidence level of each period within a preset historical window, the fluctuation statistic of the identification confidence level is calculated. The fluctuation statistic is used to characterize the dispersion of the identification confidence level in each period. The data collection status for the current period is determined based on the identification confidence level and the confidence level fluctuation statistics. Based on the data acquisition status, adjust the vibration data sampling rate and / or the number of wavelet packet decomposition layers for subsequent cycles.

9. The intelligent control method for an electric hoist according to claim 8, characterized in that, After adjusting the sampling rate of vibration data in subsequent cycles and / or the number of wavelet packet decomposition levels, the following is also included: The adjusted sampling rate and / or wavelet packet decomposition level will be applied to the vibration data acquisition and wavelet packet decomposition processing for the next cycle.

10. An intelligent control system for an electric hoist, characterized in that, The intelligent control method for the electric hoist as described in any one of claims 1-9 includes an operation data acquisition module, a data preprocessing module, a weighted analysis module, a confidence analysis module, and a control decision module; The operation data acquisition module is used to acquire vibration data and drum rotation position data collected during the operation of the electric hoist. The data preprocessing module is used to perform equal-angle resampling on the vibration data based on the rotational position data and generate a first feature vector characterizing the disorder of energy distribution in each frequency band; and to generate a second feature vector characterizing the correlation between vibration events in each frequency band and the drum rotation angle based on the angular domain vibration data. The weighted analysis module is used to dynamically adjust the weight coefficients of each frequency band in the first feature vector of the current period according to the historical second feature vector obtained in the preset period, and to perform weighted processing on the first feature vector to obtain the weighted feature vector of the current period. The confidence analysis module is used to match the weighted feature vector with the standard template vectors corresponding to at least two preset types of tangled ropes, and determine the tangled rope pre-identification type and the corresponding identification confidence level for the current period based on the matching results. The control decision module is used to determine and execute corresponding control instructions from at least two preset control strategies of different levels based on the pre-identification type of the tangled rope and the identification confidence level.

Images