In-place feeding detection method and system based on current multi-order fuzzy recognition

Abstract

The invention provides a feeding in-place detection method and system based on current multi-order fuzzy recognition, and the method and system replace a physical sensor with current analysis, and achieve the in-place detection of the feeding in-place through the innovative combination of multi-state hierarchical control, fuzzy recognition and a dual-threshold strategy. The three technical bottlenecks of poor anti-interference performance, high transient misjudgment and weak adaptability in industrial feeding detection are overcome, and the cooperative rise of the detection precision, the system robustness and the economic benefit is realized at the same time.

Description

Technical Field

[0001] This invention relates to the field of industrial automation control technology, and in particular to a material loading position detection method based on current multi-order fuzzy recognition, as well as a material loading position detection system, electronic equipment, and computer-readable storage medium based on current multi-order fuzzy recognition. Background Technology

[0002] With the emergence and development of "Industry 4.0 and Intelligent Manufacturing," the reliance on human labor in industrial production will gradually decrease, and the use of robotic arms to replace manual labor on production lines has become an inevitable trend. As the level of automation in production workshops continues to increase, machine tool loading and unloading systems are developing towards higher precision and higher reliability. Therefore, many companies are adopting a combination of gantry robots and CNC machine tools to form automated production lines.

[0003] In industrial automated material handling processes, when using robotic arms for automatic loading, the problem of workpieces not being loaded in the correct position often occurs. This leads to a cascading failure in the production system: abnormal processing quality, decreased overall equipment efficiency (OEE), and soaring maintenance costs. Currently, the detection of loading position mainly relies on photoelectric sensors, limit switches, or vision systems, which are susceptible to interference from dust and oil, and have limitations in installation location. Summary of the Invention

[0004] To address the technical problems existing in the prior art, the present invention provides the following technical solution: On the one hand, a feeding position detection method based on multi-order fuzzy current recognition is provided. This method is implemented by electronic equipment and includes: The current and speed values ​​of the drive motor are collected in real time, and the collected current data are processed by wavelet filtering and moving average. Wavelet packet decomposition is performed on the smoothed current data to extract high-frequency characteristic energy, and the time-domain signal and frequency-domain signal are fused. The feeding process is divided into three stages based on the rotation speed change: start-up, normal operation, and stop. During the start-up and stop phases, multiple state phases of the current are identified using a fuzzy control algorithm. In the preset state stages of the start-up and stop phases, the Canopy-K-means clustering algorithm based on dynamic time warping is used to extract cluster centers as benchmark features, and the material loading status is determined based on these features. During the normal operation phase, a static threshold strategy is used to determine when the material is in place.

[0005] Furthermore, the wavelet filtering of the acquired current data specifically includes: Wavelet denoising is performed using a threshold function that incorporates sample entropy. The wavelet threshold λj is adaptively set according to the formula λj = σj × √(2×log(Nj)), where σj is the mean square error of the wavelet coefficients of the j-th layer and Nj is the length of the wavelet coefficients of the j-th layer decomposition.

[0006] Furthermore, the feature fusion of the time-domain signal and the frequency-domain signal is specifically achieved through the following formula: , in, These are the fused feature values. To smooth the time-domain characteristic energy of current data, α represents the frequency domain characteristic energy of the high-frequency band after wavelet packet decomposition, and α is the fusion coefficient.

[0007] Furthermore, the identification of multiple state stages of the current using the fuzzy control algorithm specifically includes: The actual current ratio, current change rate, speed change rate, and position error are used as inputs to the fuzzy control algorithm. The Takagi-Sugeno-Kang fuzzy inference method and the weighted average defuzzification method are used to output the state variables of the current.

[0008] Furthermore, the adoption of the Canopy-K-means clustering algorithm based on dynamic time warping specifically includes: Pre-clustering stage: Based on the mean vector of the current feature dataset, by setting initial thresholds T1 and T2, the data points are divided into different Canopy sets using the DTW distance; Precise clustering stage: Using the Canopy centers obtained in the pre-clustering stage as the initial cluster centers, the DTW distance from data points to the cluster centers is calculated iteratively and clusters are reassigned to update the cluster centers; Based on the Laida criterion, a distance threshold T = μ + 3σ is set for each cluster, where μ is the mean of all DTW distances and σ is the corresponding standard deviation.

[0009] Furthermore, it also includes a parameter self-learning step: When the conditions of continuous feeding success, load change rate less than preset value, and current fluctuation less than preset value are met, the parameter self-learning process is triggered. Update the static threshold and cluster baseline centers, and verify the rationality of the update results through boundary conditions.

[0010] According to the method of claim 6, the clustering baseline center update in the parameter self-learning step includes: For newly collected valid samples, the corresponding cluster centers are updated using an incremental update method; The updated system implements a dual protection mechanism, including center offset checks and threshold mutation checks.

[0011] Furthermore, it also includes security alarm procedures: The threshold over-limit alarm and stall protection alarm are distinguished by analyzing the motor current and its rate of change. When the motor current continuously exceeds the threshold and the rate of change increases sharply, the stall protection alarm is triggered. When the current momentarily exceeds the limit but the rate of change is gradual, the threshold over-limit alarm is triggered.

[0012] On the other hand, a material loading detection system based on multi-order fuzzy current recognition is provided to implement the above-mentioned method, including: The online data acquisition and preprocessing module is used to acquire motor current and speed signals in real time, and perform wavelet filtering and moving average processing. The signal analysis module is used to perform wavelet packet decomposition on the preprocessed signal, extract high-frequency characteristic energy, and realize feature fusion of time-domain and frequency-domain signals. The multi-state recognition and judgment module is used to divide the feeding stage according to the speed change, and to identify multiple states of the current through fuzzy rules in the start-up and stop stages. The self-learning optimization module is used to dynamically update the threshold and clustering baseline center based on historical data when the conditions are met. The safety alarm module is used to implement graded responses for threshold exceeding, stall protection, and system failures.

[0013] Furthermore, it also includes general modules for user login authentication, system parameter configuration, and data storage management.

[0014] On the other hand, an electronic device is provided, comprising: a processor; and a memory storing computer-readable instructions, wherein when executed by the processor, the computer-readable instructions implement any of the above-described methods for material loading detection based on multi-level fuzzy current recognition.

[0015] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, the at least one instruction being loaded and executed by a processor to implement any of the above-described feeding position detection methods based on current multi-order fuzzy recognition.

[0016] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: This invention relates to a feeding position detection method and system based on multi-level fuzzy current recognition. It replaces physical sensors with current analysis and overcomes three major technical bottlenecks in industrial feeding detection—poor anti-interference, high transient misjudgment, and weak adaptability—through an innovative combination of multi-state hierarchical control, fuzzy recognition, and dual threshold strategy. At the same time, it achieves a synergistic leap in detection accuracy, system robustness, and economic benefits. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the feeding position detection method of the present invention.

[0019] Figure 2 This is a schematic diagram of the smoothing process of the present invention.

[0020] Figure 3 This is a diagram of the wavelet packet decomposition structure of the present invention. Figure 4 This is a flowchart of the material loading and detection strategy of the present invention.

[0021] Figure 5 This is a schematic diagram of the DTW matching results.

[0022] Figure 6 This is a flowchart of the Canopy-K-means algorithm.

[0023] Figure 7 This is a flowchart of the self-learning process of the present invention.

[0024] Figure 8 This is a schematic diagram of the material loading detection system module of the present invention. Detailed Implementation

[0025] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0026] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0027] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0028] In this embodiment of the invention, sometimes a subscript such as W1 may be mistakenly written as a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0029] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0030] This invention provides a method for detecting material loading in place based on multi-order fuzzy current recognition. This method can be implemented by an electronic device, which can be a terminal or a server. Figure 1 The flowchart shown is for a material loading position detection method based on multi-order fuzzy current recognition. The processing flow of this method may include the following steps: S1: Real-time acquisition of the current value IM and speed RM of the drive motor, and elimination of noise interference through wavelet filtering. Wavelet threshold denoising can be divided into hard thresholding and soft thresholding depending on the thresholding method.

[0031] Hard thresholding method: , Soft thresholding method: , Hard thresholding functions exhibit abrupt transitions at the threshold λ (wavelet threshold), resulting in discontinuities and additional oscillations in the reconstructed signal, thus reducing signal smoothness. Conversely, soft thresholding functions consistently produce wavelet coefficients smaller than the original coefficients, violating the principle that wavelet coefficients greater than the threshold should be preserved as much as possible, making the reconstructed signal prone to losing detail and resulting in larger reconstruction errors. Therefore, a new function is constructed by introducing sample entropy into the thresholding function: ,in: : The k-th input wavelet coefficient of the j-th layer (in the hard thresholding method, if | If |>λ, retain the original value; otherwise, set it to 0. In the soft thresholding method, if | |>λ then take λ: Wavelet threshold (or 0 otherwise); W: Original wavelet coefficients to be decomposed (original signal coefficients input to wavelet transform); λ: Wavelet threshold (critical value controlling denoising intensity); S: Sample entropy at each scale normalized to the maximum component sample entropy (reflects signal complexity and is used to dynamically adjust threshold weights).

[0032] The data acquisition module acquires real-time signal data of the current feeding process. The raw current signal is first filtered by wavelet filtering to eliminate electromagnetic noise and impact interference. Wavelet threshold denoising can be divided into hard thresholding and soft thresholding depending on the threshold's application method. Hard thresholding functions exhibit abrupt changes at the threshold λ (wavelet threshold), resulting in discontinuities and additional oscillations in the reconstructed signal, reducing signal smoothness. Soft thresholding, on the other hand, results in wavelet coefficients that are consistently smaller than the original coefficients. This contradicts the principle that wavelet coefficients greater than the threshold should be considered valid and preserved as much as possible, making the reconstructed signal prone to losing detail and resulting in larger reconstruction errors. Therefore, sample entropy is introduced into the thresholding function to construct a new function: Where Wj,k is the k-th wavelet coefficient of the j-th layer, λ is the wavelet threshold, and S is the sample entropy at each scale normalized to the maximum component sample entropy.

[0033] After determining the threshold function, to address the issue that selecting a fixed threshold for different decomposition scales leads to a deterioration in denoising performance, the threshold is set using the following calculation formula: λj=σj×sqrt(2×log(Nj)), Where λj is the wavelet threshold of the j-th layer (the critical value used to filter noise in this layer); σj is the mean square error of the wavelet coefficients of the j-th layer (reflecting the noise level of the coefficients in this layer); Nj is the length of the wavelet coefficients of the j-th layer decomposition (the number of data points in this layer); log is the natural logarithm function, and sqrt is the square root function.

[0034] After wavelet denoising, the current data is smoothed using a moving average method. The calculation formula is as follows:

[0035] N is the size of the moving window (the number of consecutive sampling points used to balance smoothness and real-time performance). The original current data collected at time i; The current data at time M is the average of the original data within the window. S2: Perform wavelet packet decomposition on the smoothed current data to extract high-frequency characteristic energy, and fuse the time-domain signal with the frequency-domain signal. The calculation formula is as follows:

[0036] Where α is the fusion coefficient (0≤α≤1, balancing the weights of time-domain and frequency-domain features); Etime is the time-domain feature energy of the smoothed current data (such as statistical features like variance and peak value); and Efreq is the frequency-domain feature energy of the high-frequency band after wavelet packet decomposition (reflecting transient changes in current). S3: The feeding process is divided into a start-up stage, a normal operation stage, and a stop stage based on the change in rotation speed; S4: During the start-up and stop phases, the five state phases of the current are identified through the fuzzy control algorithm. The input quantities are the actual current ratio IRATIO, the current change rate DI, the speed change rate DW, and the position error E. The output quantities are the current state quantities. 1. IRATIO's {Z,S,M,L,SAT}; 2. The fuzzy set of DI is {NB, NS, Z, PS, PB}; 3. The fuzzy set of DW is {Z,L,M,H,VH}; 4. The fuzzy set of E is {NB, NS, Z, PS, PB}; S5: The judgment strategy is the same in the start-up and stop phases. Taking the start-up phase as an example, in the start-up phase, the peak current in phases 1-2 is very large, at which point the position detection is turned off; in phase 3, the current oscillates violently and there is a step jump; in phase 4, the current fluctuates dynamically at high frequency, and the speed command is responded to quickly; in phase 5, the current changes smoothly and stably.

[0037] Each time a new smoothing value is obtained, it is Ismooth_new (the smoothed current value calculated at the current time, i.e., the value at the next time). After obtaining the smoothed current data, the current data within the sliding window is updated in real time. The Ismooth_trace is generated as follows: Figure 2 As shown in the figure. Wavelet transform theory can decompose a signal into low-frequency and high-frequency components. However, because its scale is based on binary changes, wavelet transform theory has insufficient analytical capabilities for high-frequency signals. Therefore, wavelet packet decomposition is used to extract high-frequency characteristic energy from the denoised current data. The wavelet packet decomposition structure diagram is shown in the figure. Figure 3 As shown. Then, the time-domain and frequency-domain signals are fused using feature fusion. This leverages the anti-interference capabilities of wavelet denoising and the feature enhancement depth of wavelet packet decomposition to reduce the false positive rate. The calculation formula is as follows: Efused = α×Etime + (1-α)×Efreq (α is the fusion coefficient, 0≤α≤1, which balances the weights of time-domain and frequency-domain features; Etime is the time-domain feature energy of the smoothed current data, such as variance, peak value and other statistical features; Efreq is the frequency-domain feature energy of the high-frequency band after wavelet packet decomposition, which reflects the transient changes of the current) During the motor start-up and stop phases, the current undergoes complex transient processes, making detection strategies prone to misjudgment. Based on changes in rotational speed and current, the workpiece loading process is divided into three phases: start-up, normal operation, and stop. Segmented control judgments are then performed based on the state characteristics of each phase.

[0038] The judgment strategy is the same in both the start-up and stop phases; taking the start-up phase as an example, the current is divided into 5 stages for the start-up phase, and the fuzzy control algorithm is used to identify and judge the state. The actual current ratio, current change rate, speed change rate, and position error are used as inputs, and the final output is the current state of the current. The definition of the fuzzy language is shown in Table 1:

[0039] Table 1 The actual current ratio parameter is represented by IRATIO, and the fuzzy set is defined as {Z,S,M,L,SAT}; the current change rate parameter is represented by DI, and the fuzzy set is defined as {NB,NS,Z,PS,PB}; the speed change rate parameter is represented by DW, and the fuzzy set is defined as {Z,L,M,H,VH}; the position error parameter is represented by E, and the fuzzy set is defined as {NB,NS,Z,PS,PB}. The fuzzy control structure consists of four inputs and one output. It employs the "Takagi-Sugeno-Kang" fuzzy inference method and a weighted average defuzzification method to determine the current state of the motor based on the current inputs. The established fuzzy recognizer inference rules are shown in Table 2.

[0040] Table 2 During the startup and shutdown phases, the Canopy-K-means algorithm is used to extract the cluster centers from phases 3 and 4 as baseline features to determine the material loading status. This method uses the initial cluster centers obtained from Canopy clustering as input for K-means clustering, avoiding the blind selection of the K value in the K-means algorithm.

[0041] The flowchart of the material loading and detection strategy is as follows: Figure 4 As shown in the diagram, during the startup phase, the peak current in stages 1-2 is very high, at which point the position detection is disabled; in stage 3, the current oscillates violently with step jumps; in stage 4, the current fluctuates dynamically at high frequencies, responding quickly to speed commands; and in stage 5, the current changes smoothly and stably. During the startup and shutdown phases, the Canopy-K-means algorithm is used to extract the cluster centers of stages 3 and 4 as baseline features to determine the material loading status. This method uses the initial cluster centers obtained from Canopy clustering as input for K-means clustering, avoiding the blind selection of the K value in the K-means algorithm.

[0042] The original k-means algorithm, which uses Euclidean distance as a similarity measure, cannot handle issues such as missing data or inconsistent data lengths caused by various reasons. Therefore, the Canopy-K-means algorithm employs Dynamic Time Warping (DTW) to measure data similarity. DTW first performs optimal alignment or coupling between two numerical sequences. After alignment, it flexibly calculates the similarity between the two time sequences by calculating the coordinates of the alignment or coupling points. The DTW distance algorithm first extends and shortens the time sequences before calculating the similarity between them.

[0043] The Canopy-K-means algorithm consists of a pre-clustering stage and a precise clustering stage. The specific steps of the pre-clustering stage are as follows: First, calculate the mean vector M of the current feature dataset D, and then calculate the Euclidean distance d between this mean vector and all remaining data points.

[0044] Let dmax be the furthest distance from the mean and dmin be the closest distance from the mean. Let the initial thresholds be T1 = 0.6 * dmax and T2 = 0.4 * dmax. Take the point closest to the mean as the initial cluster center.

[0045] Calculate the DTW distance di between the cluster center and all remaining data points, and assign data points with a distance di less than T1 to a Canopy set.

[0046] Remove points whose distance di is less than T2 from the dataset.

[0047] Repeat steps 2 to 4 until the residual current feature dataset D is empty, then end the loop.

[0048] The specific steps of the precise clustering stage are as follows: Input the initial set of cluster centers C = {c1,c2,...,ck} output by Canopy, and the number of clusters k.

[0049] For each data point X, calculate its distance to all cluster centers, and assign X to the nearest cluster center to form k temporary clusters C_i (i=1,2,...,k).

[0050] (3) Calculate the average value of the n data points X_i within each temporary cluster, and use it as the new cluster center C_new for that cluster: C_new = (1 / n)ΣX_i, Where n is the number of samples in the cluster, and X_i is the i-th data point in the cluster.

[0051] (4) Determine whether the new cluster centers satisfy the clustering convergence function (the sum of the Euclidean distances between the old and new cluster centers is less than a preset threshold ε). The clustering convergence function is as follows: Σ||C_new - C_old|| ≤ ε (where C_new is the new cluster center, C_old is the old cluster center, and ε is the convergence threshold); If any of the following conditions are met, stop the iteration; otherwise, repeat steps 2 to 4.

[0052] Σ||C_new - C_old|| ≤ ε; The maximum number of iterations, max_iter, has been reached. ε is a preset convergence threshold, which measures the degree of change between the new cluster centers in the Kth iteration and the old cluster centers in the (K-1)th iteration; max_iter is the maximum number of iterations, to prevent the algorithm from looping infinitely or converging too slowly. ε is set to 0.0001, and max_iter is set to 300.

[0053] Five state stages of the current during the startup phase are identified using a fuzzy control algorithm. Clustering of the current data in stages 3 and 4 yields the cluster centers for stages 3 and 4 under normal conditions, reflecting the current variation characteristics under normal feeding conditions. Multiple sets of data under normal conditions are acquired to form a training dataset, thus obtaining the baseline features of the cluster centers for stages 3 and 4. The detection logic is as follows: Training phase (establishing benchmarks) Collect feature vector data for stages 3 and 4 under normal material feeding conditions, denoted as D3 (stage 3 feature vector dataset) and D4 (stage 4 feature vector dataset).

[0054] Canopy-K-means clustering was applied to D3 and D4 respectively.

[0055] Canopy coarse clustering: First, calculate the mean point of datasets D3 and D4 respectively. Then, calculate the distances d3_i (in D3) and d4_i (in D4) between the mean point and all remaining data points. Let the initial thresholds be T1_3 = 0.6 * d3_max, T2_3 = 0.4 * d3_max; T1_4 = 0.6 * d4_max, T2_4 = 0.4 * d4_max (where d3_max and d4_max are the farthest distances from data points in D3 and D4 to the mean point, respectively, and d3_min and d4_min are the closest distances). Data points with a distance less than T1_3 are assigned to a Canopy set in D3, and points with a distance less than T2_3 are removed from the D3 dataset. The same process is applied to D4 until the data list is empty, resulting in the Canopy sets for D3 and D4 and the data points within each Canopy.

[0056] K-means fine clustering: Uses the K value obtained from the Canopy step, and uses the center point of each Canopy step as the initial cluster center for K-means. Assignment step: Assigns each data point to the nearest cluster center. Update step: Recalculates the cluster center for each cluster (usually the mean of all points in that cluster). Repeats these two steps until the cluster centers no longer change significantly or the maximum number of iterations is reached.

[0057] The cluster centers C3 for stage 3 and C4 for stage 4 are obtained.

[0058] For each cluster, calculate the DTW distance d_i from all sample points within that cluster to its cluster center. Calculate the mean μ and standard deviation σ for all d_i within that cluster.

[0059] According to the Laida criterion, the distance threshold for this cluster is set as T = μ + 3σ.

[0060] Testing phase The fuzzy control algorithm identifies whether the current stage is stage 3 or stage 4; Extract the data feature vector X_current of the current time window; Calculate the DTW distance d_j from X_current to all baseline cluster centers in the corresponding stage. Find the minimum distance d_min and its corresponding cluster C_min.

[0061] The Laida criterion is applied to determine whether the material is properly loaded. If d_min < T_min (T_min is the distance threshold corresponding to C_min), the material is considered to be properly loaded; otherwise, the material is considered to be improperly loaded.

[0062] S6: After stage 5, the current fluctuation amplitude is very small, and the system has completely entered a steady state without any violent transient processes. Therefore, the static threshold strategy is used for judgment during the normal operation phase. The calculation formula is as follows: T_static = K_static × I_avg_normal In the formula: T_static is the static threshold, K_static is the static coefficient, and I_avg_normal is the moving average current during the normal operation phase after 5 stages; S7: When the following conditions are met: N consecutive successful material feedings, load change rate <15%, and current fluctuation <0.1 times the rated current, parameter self-learning is triggered: A self-learning mechanism is established to optimize parameters. This mechanism dynamically optimizes parameters through historical data analysis and includes strict boundary condition verification, thereby ensuring that the function can be effectively implemented under different operating conditions, improving the reliability and stability of the function. The self-learning process is shown in Figure 7.

[0063] The system will automatically trigger the parameter self-learning process when the following conditions are met simultaneously.

[0064] 1. Successful feeding N times in a row; 2. Load change rate is less than 15%; 3. Current fluctuation is less than 0.1 times the rated current.

[0065] Static threshold update strategy: 1. Update the historical maximum average current I_max_hist (the maximum value of the moving average current during normal operation in history); 2. Calculate the new dynamic threshold and static threshold, and update the calculation formula as follows: T_static_new = K_static × I_max_hist_new (T_static_new is the updated static threshold; K_static is the static coefficient; I_max_hist_new is the updated historical maximum average current).

[0066] 3. Verify the reasonableness of the threshold through boundary conditions. The boundary verification conditions are as follows: 1) Mutation constraint: |T_new - T_old| ≤ 0.2*T_old (T_new is the new threshold, T_old is the old threshold); 2) Historical range: 0.7*T_hist_min ≤ T_new ≤1.3*T_hist_max (T_hist_min is the historical minimum threshold, T_hist_max is the historical maximum threshold); Clustering benchmark center update strategy: Once the triggering condition is met, capture a time window of data for a complete material feeding cycle.

[0067] For each newly acquired feature vector X_new, calculate its DTW distance d_new to the current baseline cluster center. Check if d_new is less than the distance threshold T corresponding to the cluster. If d_new < T, the sample is determined to belong to the cluster and is a valid sample; otherwise, the sample is considered an anomaly, discarded, and a "feature anomaly" log is triggered.

[0068] For valid samples, the corresponding cluster centers are updated incrementally (assuming the sample belongs to the k-th cluster): C_new=(n*C_old + X_new) / (n+1) (n is the original number of cluster samples, C_old is the original cluster center, and X_new is the newly added valid sample vector) Where C_old is the cluster center vector before the update; n is the number of samples currently contained in the cluster; X_new is the feature vector of the newly added sample; and C_new is the cluster center vector after the update.

[0069] Synchronously update the number of samples in the target cluster: n_new = n_old +1 Merge all historical samples of the cluster with the new samples, and recalculate the distance from all historical samples (including newly added samples) within the cluster to the new cluster center. Calculate the mean μ_new (the average DTW distance from all sample points in the new cluster to the new center) and the standard deviation σ_new (the standard deviation of the DTW distance within the new cluster).

[0070] Update distance threshold: T_new = μ_new + 3σ_new. After the update, a dual protection mechanism will be implemented: (1) Center offset check: If ||C_new - C_old|| (Euclidean distance between the new cluster center and the original center) ≤ 0.1*||C_old|| (modulus of the original center), it is considered an abnormal update. The update is revoked, the original cluster centers are kept unchanged, and an alarm is issued. (2) Threshold mutation check: If |T_new (updated distance threshold) - T_old (original distance threshold)| ≤ 0.2*T_old, then freeze the update and request manual intervention.

[0071] If the safety boundary check passes, the updated cluster centers, sample counts, and distance thresholds are used to overwrite the original values, thus completing the update.

[0072] The parameter self-learning mechanism can be divided into three state stages: Detection state: The detection state is the process of performing material loading detection, which monitors the equipment status in real time and determines whether the self-learning trigger conditions are met.

[0073] Adaptive state: Collect and validate valid samples, and build an update buffer.

[0074] Update state: Perform static threshold update, cluster baseline center update and verification. Material loading detection is a real-time process. Parameter updates cannot be performed during material loading detection; they must wait until the system is idle.

[0075] This three-state mechanism, through strict time-series isolation and resource allocation, ensures the real-time performance of material loading and detection while achieving safe and efficient execution of parameter self-learning.

[0076] In the above-mentioned feeding position detection method based on current multi-order fuzzy recognition, the fuzzy control algorithm adopts the Takagi-Sugeno-Kan fuzzy inference and weighted average defuzzification method.

[0077] In the aforementioned material loading detection method based on multi-level fuzzy current recognition, the five state recognitions in the start-up and stop phases specifically include: During states 1-3, the current fluctuates drastically, and the position detection function is disabled. In state phases 4-5, the current fluctuates at high frequency, and the clustering baseline center update strategy is activated.

[0078] In the above-mentioned feeding position detection method based on multi-order fuzzy current recognition, the boundary condition verification includes: Center offset check: If ||C_new - C_old|| (Euclidean distance between the new cluster center and the original center) ≤ 0.1*||C_old|| (modulus of the original center), it is considered an abnormal update and the update is cancelled; Threshold mutation check: If |T_new (updated distance threshold) - T_old (original distance threshold)| ≤ 0.2*T_old, then freeze the update and request manual intervention; Mutation constraint: The difference between the new threshold (T_new) and the old threshold (T_old) is ≤ 20% of the old threshold (T_old); Historical range constraint: New threshold (T_new) ∈ [70% of historical minimum (T_hist_min), 130% of historical maximum (T_hist_max).

[0079] A material loading detection system based on multi-order fuzzy current recognition, characterized in that it includes: The online data acquisition and preprocessing module is used to acquire motor current and speed signals in real time, and perform wavelet filtering and moving average processing. The signal analysis module performs power spectrum analysis, variable mode decomposition, short-time Fourier transform, and WVD time-frequency analysis on the preprocessed signal. The multi-state recognition and judgment module identifies five states of current through fuzzy rules during the startup phase and adaptively switches the detection strategy. The self-learning optimization module can dynamically update thresholds based on historical data and verify the results through boundary conditions. The safety alarm module enables tiered responses to threshold exceedances, stall protection, and system failures.

[0080] In the aforementioned material loading detection system based on multi-level fuzzy current recognition, the safety alarm module distinguishes alarm types in the following way: When the motor current continuously exceeds the threshold and the rate of change increases sharply, the stall protection alarm is triggered. When the current momentarily exceeds the limit but the rate of change is gradual, the threshold over-limit alarm is triggered.

[0081] The aforementioned material loading detection system based on multi-level fuzzy current recognition also includes a general module for user login authentication, system parameter configuration, and data storage management.

[0082] Please refer to the preceding steps for a detailed understanding of the system's specific methods.

[0083] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0084] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable system. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0085] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0086] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0087] It should be understood that, in various embodiments of the present invention, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0088] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0089] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, systems, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0090] In the embodiments provided by this invention, it should be understood that the disclosed devices, systems, and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between systems or units may be electrical, mechanical, or other forms.

[0091] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0092] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0093] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0094] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for detecting material loading position based on multi-order fuzzy current recognition, characterized in that, Includes the following steps: The current and speed values ​​of the drive motor are collected in real time, and the collected current data are processed by wavelet filtering and moving average. Wavelet packet decomposition is performed on the smoothed current data to extract high-frequency characteristic energy, and the time-domain signal and frequency-domain signal are fused. The feeding process is divided into three stages based on the rotation speed change: start-up, normal operation, and stop. During the start-up and stop phases, multiple state phases of the current are identified using a fuzzy control algorithm. In the preset state stages of the start-up and stop stages, the Canopy-K-means clustering algorithm based on dynamic time warping is used to extract cluster centers as benchmark features, and the material loading status is determined based on these features. During the normal operation phase, a static threshold strategy is used to determine when the material is in place.

2. The method according to claim 1, characterized in that, The wavelet filtering of the acquired current data specifically includes: Wavelet denoising is performed using a threshold function that incorporates sample entropy; The wavelet threshold λj is adaptively set according to the formula λj = σj × √(2×log(Nj)), where σj is the mean square error of the wavelet coefficients of the j-th layer and Nj is the length of the wavelet coefficients of the j-th layer decomposition.

3. The method according to claim 1, characterized in that, The feature fusion of the time-domain signal and the frequency-domain signal is specifically achieved through the following formula: ， in, These are the fused feature values. To smooth the time-domain characteristic energy of current data, α represents the frequency domain characteristic energy of the high-frequency band after wavelet packet decomposition, and α is the fusion coefficient.

4. The method according to claim 1, characterized in that, The identification of multiple state stages of current using a fuzzy control algorithm specifically includes: The actual current ratio, current change rate, speed change rate, and position error are used as inputs to the fuzzy control algorithm. The Takagi-Sugeno-Kang fuzzy inference method and the weighted average defuzzification method are used to output the state variables of the current.

5. The method according to claim 1, characterized in that, The aforementioned Canopy-K-means clustering algorithm based on dynamic time warping specifically includes: Pre-clustering stage: Based on the mean vector of the current feature dataset, by setting initial thresholds T1 and T2, the data points are divided into different Canopy sets using the DTW distance; Precise clustering stage: Using the Canopy centers obtained in the pre-clustering stage as the initial cluster centers, the DTW distance from data points to the cluster centers is calculated iteratively and clusters are reassigned to update the cluster centers; Based on the Laida criterion, a distance threshold T = μ + 3σ is set for each cluster, where μ is the mean of all DTW distances and σ is the corresponding standard deviation.

6. The method according to claim 1, characterized in that, It also includes parameter self-learning steps: When the conditions of continuous feeding success, load change rate less than preset value, and current fluctuation less than preset value are met, the parameter self-learning process is triggered. Update the static threshold and cluster baseline centers, and verify the rationality of the update results through boundary conditions. According to the method of claim 6, the clustering baseline center update in the parameter self-learning step includes: For newly collected valid samples, the corresponding cluster centers are updated using an incremental update method; The updated system implements a dual protection mechanism, including center offset checks and threshold mutation checks.

7. The method according to claim 1, characterized in that, It also includes security alarm procedures: The threshold over-limit alarm and stall protection alarm are distinguished by analyzing the motor current and its rate of change. When the motor current continuously exceeds the threshold and the rate of change increases sharply, the stall protection alarm is triggered. When the current momentarily exceeds the limit but the rate of change is gradual, the threshold over-limit alarm is triggered.

8. A material loading detection system based on multi-order fuzzy current recognition, used to implement the method according to any one of claims 1-7, characterized in that, include: The online data acquisition and preprocessing module is used to acquire motor current and speed signals in real time, and perform wavelet filtering and moving average processing. The signal analysis module is used to perform wavelet packet decomposition on the preprocessed signal, extract high-frequency characteristic energy, and realize feature fusion of time-domain and frequency-domain signals. The multi-state recognition and judgment module is used to divide the feeding stage according to the speed change, and to identify multiple states of the current through fuzzy rules in the start-up and stop stages. The self-learning optimization module is used to dynamically update the threshold and clustering baseline center based on historical data when the conditions are met. The safety alarm module is used to implement graded responses for threshold exceeding, stall protection, and system failures.

9. The system according to claim 8, characterized in that, It also includes general modules for user login authentication, system parameter configuration, and data storage management.

10. An electronic device, characterized in that, The electronic device includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 7.

Images