A method for detecting wire rope damage based on sound signature and vibration fusion
By using a detection method that combines sound patterns and vibrations, a load-adaptive signal prediction model is constructed to obtain the deviation and breakage probability of the wire rope. This solves the problems of low efficiency and insufficient early warning in traditional detection methods, and achieves high-precision and reliable early diagnosis of wire ropes.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- LEVI INTELLIGENT (SHENZHEN) CO LTD
- Filing Date
- 2025-09-03
- Publication Date
- 2026-07-17
AI Technical Summary
Existing wire rope inspection methods rely on manual inspection, which is inefficient and makes it difficult to detect internal damage. They also lack perception and analysis of the operating status, making it impossible to provide early warnings and resulting in insufficient inspection accuracy and safety.
A detection method based on sound pattern and vibration fusion is adopted. By constructing a load-adaptive signal prediction model and a multi-modal deviation detection mechanism, the deviation between the predicted and actual signals is obtained. Combined with historical similar samples, the failure probability is calculated to achieve early and accurate warning and forced shutdown of wire rope.
It significantly improves the accuracy of wire rope condition monitoring and early warning capabilities, enabling precise diagnosis under complex working conditions, enhancing detection accuracy and the accuracy of damage risk prediction, and ensuring safety.
Smart Images
Figure CN120927810B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wire rope testing technology, specifically to a wire rope damage detection method based on the fusion of sound patterns and vibration. Background Technology
[0002] In the field of wire rope inspection, traditional methods mainly rely on periodic manual visual inspections. However, manual inspections are highly subjective, inefficient, and struggle to detect internal damage, lacking the ability to perceive and analyze the operational status of the wire rope. Throughout its lifespan, the operational characteristics of a wire rope naturally change over time, and existing inspection methods fail to consider this time-varying characteristic, further reducing inspection accuracy. Furthermore, traditional methods cannot achieve early prediction and warning of breakage risks, rendering current wire rope damage detection methods unsuitable for the high safety requirements of current application scenarios. Summary of the Invention
[0003] This application provides a method for detecting wire rope damage based on the fusion of sound patterns and vibrations, which addresses the technical problems of low accuracy and lack of early warning capabilities in existing wire rope damage detection technologies.
[0004] In view of the above problems, this application provides a method for detecting wire rope damage based on the fusion of acoustic signature and vibration. The method includes:
[0005] Obtain the current operating load of the target wire rope, and obtain the predicted sound pattern signal and predicted vibration signal based on the current operating load;
[0006] Collect the actual acoustic and vibration signals of the target wire rope under the current operating load;
[0007] A first deviation is obtained based on the predicted voiceprint signal and the actual voiceprint signal, and a second deviation is obtained based on the predicted vibration signal and the actual vibration signal.
[0008] When the first deviation is greater than or equal to the first deviation threshold, or the second deviation is greater than or equal to the second deviation threshold, damage prediction is performed based on the actual sound pattern signal and the actual vibration signal to obtain the wire rope damage probability.
[0009] When the probability of the steel wire rope breaking is greater than or equal to the probability threshold, the target steel wire rope is forcibly deactivated.
[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:
[0011] This application proposes a wire rope damage detection method based on the fusion of sound patterns and vibration. By constructing a load-adaptive signal prediction model and a multi-modal deviation detection mechanism, the accuracy of wire rope condition monitoring and early warning capabilities are significantly improved. Compared with traditional methods, the technical solution provided in this application significantly overcomes the problem of poor detection accuracy caused by neglecting factors such as load changes and service time in traditional methods. It achieves accurate perception of the health status of wire ropes under different working conditions, and realizes early, accurate, and reliable diagnosis of wire rope damage under complex working conditions, significantly improving the technical effect of wire rope detection accuracy and wire rope damage risk prediction accuracy. Attached Figure Description
[0012] 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.
[0013] Figure 1 This is a flowchart illustrating the wire rope damage detection method based on the fusion of sound patterns and vibration, provided in an embodiment of this application.
[0014] Figure 2 This is a schematic diagram of the process for constructing a voiceprint signal predictor provided in an embodiment of this application. Detailed Implementation
[0015] This application provides a wire rope damage detection method based on the fusion of sound patterns and vibrations, which is used to address the technical problem in the prior art that switchgear monitoring cannot integrate historical aging trends with real-time monitoring data to achieve accurate monitoring.
[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.
[0017] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.
[0018] Example 1, as Figure 1As shown, this application provides a method for detecting wire rope damage based on the fusion of sound patterns and vibration, wherein the method includes:
[0019] S10: Obtain the current operating load of the target wire rope, and obtain the predicted sound pattern signal and the predicted vibration signal based on the current operating load.
[0020] In the process of monitoring the condition of wire ropes, traditional methods usually use fixed thresholds or historical data under a single working condition as a reference, which cannot adapt to the dynamic changes in the vibration and acoustic characteristics of wire ropes under different loads, resulting in a disconnect between the reference signal and the actual operating condition.
[0021] Step S10 in the method provided in this application embodiment includes:
[0022] Obtain the current operating load of the target wire rope and retrieve the sound signature predictor and vibration signal predictor bound to the target wire rope;
[0023] The current operating load is input into the phonogram signal predictor and the vibration signal predictor respectively to obtain the predicted phonogram signal and the predicted vibration signal;
[0024] The construction steps of the voiceprint signal predictor and the vibration signal predictor include:
[0025] The application information attributes of the target wire rope are obtained, and the operation records of historical wire ropes of the same type that are running normally within a preset service period are retrieved based on the application information attributes to construct a sample wire rope load set.
[0026] Acoustic and vibration signals from historical operating records of the same type of wire rope in the sample wire rope load set were collected to construct a sample acoustic signal set and a sample vibration signal set.
[0027] Using the sample steel wire rope load set as input features and the sample sound pattern signal set as supervision features, a sound pattern signal predictor is constructed; and using the sample steel wire rope load set as input features and the sample vibration signal set as supervision features, a vibration signal predictor is constructed.
[0028] The voiceprint signal predictor is constructed using the sample wire rope load set as input features and the sample voiceprint signal set as supervision features, as follows: Figure 2 As shown, it includes:
[0029] Obtain the service time of each historical wire rope of the same type corresponding to the sample wire rope load set, and construct the sample service time set;
[0030] Based on the variation characteristics of the voiceprint signals in the sample voiceprint signal set with service time, a first cluster analysis is performed on the sample service time set to obtain multiple voiceprint service time intervals.
[0031] Based on the multiple service time intervals of the sound patterns, the sample wire rope load set and the sample sound pattern signal set are divided to obtain multiple sound pattern grouped load subsets and multiple sample sound pattern signal subsets;
[0032] Each of the aforementioned voiceprint group load subsets is used as input features, and the corresponding sample voiceprint signal subsets are used as supervision features to construct multiple voiceprint signal prediction branches, thus forming the voiceprint signal predictor.
[0033] The vibration signal predictor is constructed using the sample wire rope load set as input features and the sample vibration signal set as supervision features, including:
[0034] Based on the variation characteristics of vibration signals with service time in the sample vibration signal set, a second cluster analysis is performed on the sample service time set to obtain multiple vibration service time intervals.
[0035] Based on the multiple vibration service time intervals, the sample wire rope load set and the sample vibration signal set are divided to obtain multiple vibration group load subsets and multiple sample vibration signal subsets;
[0036] The vibration signal predictor is formed by constructing multiple vibration signal prediction branches using each vibration group load subset as input features and the corresponding sample vibration signal subset as supervision features.
[0037] Specifically, the current operating load is input into the voiceprint signal predictor and the vibration signal predictor respectively to obtain the predicted voiceprint signal and the predicted vibration signal, including:
[0038] Obtain the current service time of the target wire rope;
[0039] Based on the current service time, a target voiceprint service time interval is determined from the plurality of voiceprint service time intervals, and a corresponding target voiceprint signal prediction branch is selected in the voiceprint signal predictor based on the target voiceprint service time interval.
[0040] The current operating load is input into the target voiceprint signal prediction branch to obtain the predicted voiceprint signal;
[0041] Based on the current service time, a target vibration service time interval is determined from the plurality of vibration service time intervals, and a corresponding target vibration signal prediction branch is selected in the vibration signal predictor based on the target vibration service time interval.
[0042] The current operating load is input into the target vibration signal prediction branch to obtain the predicted vibration signal.
[0043] In this embodiment, the current operating load of the target wire rope is obtained, and the acoustic signal predictor and vibration signal predictor bound to the target wire rope are retrieved. The current operating load refers to the weight of the current load on the wire rope, measured in N.
[0044] Construct a voiceprint signal predictor and a vibration signal predictor.
[0045] Specifically, the application information attributes of the target wire rope are obtained. These attributes include the application scenario (e.g., elevator system, cableway), the equipment model, and the function (e.g., traction, hoisting), ensuring that usable historical samples are retrieved. Based on these application information attributes, historical operating records of the same type of wire rope that have been operating normally within a preset service period are retrieved to construct a sample wire rope load set. The preset service period refers to the service time within the wire rope's lifespan, such as 1 to 5 years.
[0046] Acoustic and vibration signals from historical operating records of the same type of wire rope were collected from the sample wire rope load set to construct sample acoustic and vibration signal sets. The collected acoustic and vibration signals came from historical wire ropes of the same type under different loads. Collecting acoustic and vibration signals under multiple loads enables subsequent models trained based on sample data to achieve a more accurate recognition rate when facing various loads.
[0047] Obtain the service time of each historical wire rope of the same type from the sample wire rope load set, and construct the sample service time set.
[0048] Based on the characteristics of voiceprint signal variation with service time within the sample voiceprint signal set, a first cluster analysis is performed on the sample service time set to obtain multiple voiceprint service time intervals. For example, the K-Means clustering algorithm can be used to cluster similar service times, thus obtaining multiple voiceprint service time intervals.
[0049] Based on multiple service time intervals of the acoustic signatures, the sample wire rope load set and the sample acoustic signature signal set are divided to obtain multiple acoustic signature grouped load subsets and multiple sample acoustic signature signal subsets. Each subset contains sample wire rope acoustic signature signals and load signals within one acoustic signature service time interval.
[0050] Multiple voiceprint prediction branches are constructed using each voiceprint group load subset as input features and the corresponding sample voiceprint signal subset as supervised features, forming a voiceprint prediction engine. Specifically, voiceprint prediction branches are constructed based on machine learning. Each prediction branch has a three-layer structure: an input layer to receive the load signal, a hidden layer with 64 nodes activated using the ReLU function, and an output layer to output the predicted voiceprint signal. Weights are randomly assigned to the constructed voiceprint prediction branches. Using the load signal from the voiceprint group load subset as input, the predicted output voiceprint signal is compared with the corresponding sample voiceprint signal. The deviation between the predicted and sample voiceprint signals is used to adjust the weights through gradient descent and backpropagation. The Adam optimizer is used to reduce the deviation and adjust the weights. This process is continuously optimized until the voiceprint prediction branch converges. For example, if the accuracy of the output predicted voiceprint signal is above 90% when inputting the load signal, the voiceprint prediction branch is considered successfully trained. The multiple trained voiceprint prediction branches are then integrated into a single voiceprint prediction engine.
[0051] Based on the variation characteristics of vibration signals with service time in the sample vibration signal set, a second cluster analysis is performed on the sample service time set to obtain multiple vibration service time intervals. Similarly, clustering is performed using an algorithm such as K-Means to cluster similar service times, resulting in multiple vibration service time intervals.
[0052] Based on multiple vibration service time intervals, the sample wire rope load set and sample vibration signal set are divided into multiple vibration group load subsets and multiple sample vibration signal subsets. Each subset contains the sample wire rope vibration signal and load signal within one vibration service time interval.
[0053] Multiple vibration signal prediction branches are constructed using different vibration group load subsets as input features and corresponding sample vibration signal subsets as supervision features to form a vibration signal predictor. For example, using the same construction and training method as the phonogram signal predictor, a three-layer machine learning-based vibration signal prediction branch is constructed. Each prediction branch has three layers: an input layer to receive the load signal, a hidden layer with 64 nodes activated using the ReLU function, and an output layer to output the predicted vibration signal. The vibration signal predictor is trained using the load signal from the vibration group load subset as input and the corresponding sample vibration signal as supervision until convergence. For example, if the accuracy of the output predicted vibration signal is above 90% when inputting the load signal, the vibration signal prediction branch training is complete. Multiple vibration signal prediction branches are constructed and trained using the same method and then integrated to obtain the vibration signal predictor.
[0054] Obtain the current service life of the target wire rope that needs to be inspected.
[0055] Based on the current service time, a target voiceprint service time interval is determined from multiple voiceprint service time intervals. Then, the corresponding target voiceprint signal prediction branch is selected in the voiceprint signal predictor based on this target voiceprint service time interval. For example, if the current service time is one year and six months, the determined target voiceprint service time interval can be one to two years. The corresponding target voiceprint signal prediction branch is then selected based on this target voiceprint service time interval. The current operating load is input into the target voiceprint signal prediction branch to obtain the predicted voiceprint signal.
[0056] Similarly, based on the current service time, a target vibration service time interval is determined from multiple vibration service time intervals, and the corresponding target vibration signal prediction branch is selected in the vibration signal predictor according to the target vibration service time interval. The current operating load is input into the target vibration signal prediction branch to obtain the predicted vibration signal.
[0057] By acquiring the current operating load in real time and generating corresponding predicted sound patterns and vibration signals based on a pre-trained signal predictor, a dynamic adaptive signal benchmark was successfully constructed. This benchmark can accurately reflect the normal signal characteristics that a wire rope should possess under a specific load, providing a scientific and reliable comparison benchmark for subsequent deviation analysis.
[0058] S20: Collect the actual acoustic and vibration signals of the target wire rope under the current operating load.
[0059] In this embodiment, the actual acoustic and vibration signals of the target wire rope under the current operating load are collected. For example, an industrial microphone can be used to collect the acoustic signal, and a vibration sensor can be used to collect the vibration signal. By deploying sensors to synchronously collect the actual acoustic and vibration signals under the current load, the real-time and accurate data acquisition is ensured. This provides first-hand data reflecting the true state of the wire rope for subsequent analysis, guaranteeing the quality of the data source for subsequent deviation calculations and damage prediction.
[0060] S30: Obtain a first deviation degree based on the predicted voiceprint signal and the actual voiceprint signal, and obtain a second deviation degree based on the predicted vibration signal and the actual vibration signal.
[0061] After obtaining the actual and predicted signals, quantifying the differences between them and accurately identifying anomalies is a major challenge. Simple amplitude comparisons cannot capture the complex characteristics of signal changes in the time and frequency domains, while traditional difference measurement methods are ill-suited to the non-stationary nature of wire rope signals, resulting in insensitivity to early and subtle damage characteristics and an inability to distinguish between normal operating condition fluctuations and abnormal damage signs.
[0062] Step S30 in the method provided in this application embodiment includes:
[0063] Feature extraction is performed on the predicted voiceprint signal and the actual voiceprint signal to obtain the predicted voiceprint feature vector and the actual voiceprint feature vector;
[0064] The first deviation is obtained based on the predicted voiceprint feature vector and the actual voiceprint feature vector;
[0065] Feature extraction is performed on the predicted vibration signal and the actual vibration signal to obtain the predicted vibration feature vector and the actual vibration feature vector;
[0066] The second deviation is obtained based on the predicted vibration feature vector and the actual vibration feature vector.
[0067] In this embodiment, an open-source signal processing library, such as Python's librosa library, is used to extract features from the Mel-frequency cepstral coefficients of both the predicted and actual voiceprint signals. Mel-frequency cepstral coefficients effectively capture the spectral shape details of sound. The signal processing library calculates a set of, for example, 12 Mel-frequency cepstral coefficients for each voiceprint signal; this set of coefficients constitutes a feature vector representing the core features of the signal. Feature extraction is then performed using the same method to obtain the predicted voiceprint feature vector and the actual voiceprint feature vector.
[0068] The first deviation is obtained by comparing the predicted and actual voiceprint feature vectors. For example, the cosine similarity between the predicted and actual voiceprint feature vectors is calculated as follows: Cosine similarity = (Predicted voiceprint feature vector × Actual voiceprint feature vector) ÷ (Modulus of predicted voiceprint feature vector × Modulus of actual voiceprint feature vector). The closer the cosine similarity is to 1, the more similar the two feature vectors are. The first deviation is calculated as 1 - Cosine similarity; a larger first deviation indicates a greater deviation between the predicted and actual voiceprint signals.
[0069] Feature extraction is performed on the predicted vibration signal and the actual vibration signal to obtain the predicted vibration feature vector and the actual vibration feature vector. For example, a fast Fourier transform can be used to convert the vibration signal to obtain the spectrum, and then obtain the predicted vibration feature vector and the actual vibration feature vector composed of the spectrum, amplitude, and phase.
[0070] The second deviation is obtained based on the predicted vibration feature vector and the actual vibration feature vector. The cosine similarity between the predicted and actual vibration feature vectors is calculated using the same method as the first deviation: Cosine similarity = (Predicted vibration feature vector · Actual vibration feature vector) ÷ (Magnitude of predicted vibration feature vector × Magnitude of actual vibration feature vector). The closer the cosine similarity is to 1, the more similar the two feature vectors are. The first deviation = 1 - Cosine similarity; a larger first deviation indicates a greater deviation between the predicted and actual vibration signals.
[0071] By extracting high-dimensional feature vectors from predicted and actual signals and calculating their deviation, a multi-dimensional and refined measurement of signal differences is achieved. This enables sensitive capture of abnormal patterns contained in signals, significantly improves the ability to perceive minute damage features, and provides accurate and reliable criteria for subsequent predictions.
[0072] S40: When the first deviation is greater than or equal to the first deviation threshold, or the second deviation is greater than or equal to the second deviation threshold, damage prediction is performed based on the actual sound pattern signal and the actual vibration signal to obtain the wire rope damage probability.
[0073] Traditional methods often rely on simple rules, such as issuing warnings when parameters exceed limits. They lack the ability to integrate and deeply analyze information from multiple sources, and cannot quantify the probability of damage. This results in low reliability of warning decisions and makes it difficult to meet the risk warning needs under complex working conditions.
[0074] Step S40 in the method provided in this application embodiment includes:
[0075] Based on the actual acoustic signature signal and the actual vibration signal, damage prediction is performed to obtain the probability of wire rope breakage, including:
[0076] The actual sound signature feature vector and the actual vibration feature vector are merged to construct the actual signal feature vector;
[0077] Using the actual signal feature vector as the retrieval condition, similar feature wire rope samples with feature vector similarity greater than or equal to a preset similarity threshold are collected to construct a similar sample set, and the total number of similar samples in the similar sample set is obtained.
[0078] The number of damaged samples is obtained by counting the number of steel wire ropes that broke within a preset service period of the similar samples.
[0079] The probability of wire rope breakage is obtained by the ratio of the number of damaged samples to the total number of similar samples.
[0080] When the first deviation is less than the first deviation threshold and the second deviation is less than the second deviation threshold, the target wire rope is marked as operating normally, and wire rope monitoring continues.
[0081] In this embodiment, when the first deviation is greater than or equal to the first deviation threshold, or the second deviation is greater than or equal to the second deviation threshold, damage prediction is performed based on the actual acoustic signal and the actual vibration signal to obtain the probability of wire rope breakage. For example, the first deviation threshold can be set to 0.4, and the second deviation threshold can be set to 0.5. Signals that are more sensitive to changes in wire rope breakage can be set to smaller deviation thresholds. The larger the deviation, the greater the deviation between the actual acoustic signal and vibration signal and the predicted signal, the more likely it is that the large signal deviation is caused by breakage, and the more necessary it is to perform breakage prediction.
[0082] Specifically, the actual sound signature feature vector and the actual vibration feature vector are merged to construct the actual signal feature vector.
[0083] Using the actual signal feature vector as the retrieval condition, similar feature wire rope samples with feature vector similarity greater than or equal to a preset similarity threshold are collected to construct a similar sample set, and the total number of similar samples in the similar sample set is obtained. For example, the cosine similarity between the actual signal feature vector and the feature vectors of other wire rope samples is calculated, and similar feature wire rope samples with feature vector similarity greater than or equal to the preset similarity threshold are added to the similar sample set. The preset similarity threshold is a pre-set threshold representing the similarity of signal features. Setting it too low may allow wire rope samples with insufficient similarity to enter the similar sample set, diluting the representativeness of the sample set. Setting it too high may result in too small a number of selected samples. For example, the preset similarity threshold can be set to 0.7.
[0084] Obtain the total number of similar samples in the similar sample set, and count the number of wire ropes that broke within the preset service period in the similar sample set to obtain the number of broken samples.
[0085] The probability of wire rope breakage is obtained by the ratio of the number of broken samples to the total number of similar samples. The probability of wire rope breakage = number of broken samples ÷ total number of similar samples. For example, if the total number of similar samples is 200 and the number of broken samples is 70, then the probability of wire rope breakage is 70 ÷ 200 = 0.35.
[0086] When the first deviation is less than the first deviation threshold and the second deviation is less than the second deviation threshold, the target wire rope is marked as operating normally, and the wire rope monitoring is continued as described above.
[0087] By integrating actual sound patterns and vibration signal characteristics and performing probability calculations based on historical similar samples, a precise quantitative assessment of the risk of wire rope breakage was achieved. This elevates the assessment from simply "whether to issue a warning" to "what is the probability of the risk," providing richer and more accurate data for safety decision-making.
[0088] S50: When the probability of the steel wire rope breaking is greater than or equal to the probability of breaking threshold, the target steel wire rope is forcibly deactivated.
[0089] In this embodiment, when the probability of wire rope breakage exceeds a breakage probability threshold, the target wire rope is forcibly deactivated to prevent safety risks caused by wire rope breakage. For example, equipment using the target wire rope may be directly deactivated, such as elevators using the target wire rope. To ensure safety and reduce false alarms, the breakage probability threshold can be set to 0.7, for example.
[0090] When the probability of wire rope breakage is less than the breakage probability threshold, a maintenance reminder message is output, such as sending a text message to the maintenance personnel's client saying "The wire rope may be at risk of breakage, please maintain it in time".
[0091] By setting a damage probability threshold and automatically triggering a forced shutdown mechanism, the risk control measures are executed automatically. This ensures that when a high risk is confirmed, the equipment can be stopped immediately without human intervention, greatly shortening the response time, effectively preventing accidents, and ensuring the safety of wire rope application scenarios.
[0092] In summary, the embodiments of this application have at least the following technical effects:
[0093] This application proposes a wire rope damage detection method based on the fusion of acoustic and vibration signals. By constructing a load-adaptive signal prediction model and a multimodal deviation detection mechanism, the accuracy and early warning capability of wire rope condition monitoring are significantly improved. Compared with traditional methods, the technical solution provided in this application significantly overcomes the problem of poor detection accuracy caused by neglecting factors such as load changes and service time in traditional methods. Specifically, by fusing the characteristics of acoustic and vibration signals for analysis and introducing a predictor branch that considers service time segments, this method can effectively distinguish between normal aging characteristics and abnormal damage signals, identify potential damage risks, and enhance the anti-interference capability and reliability of wire rope damage detection. At the same time, the calculation of damage probability based on historical similar samples provides an intuitive and reliable quantitative basis for safety decisions, achieving accurate and reliable prediction of wire rope damage risks.
[0094] This application enables precise perception of the health status of wire ropes under different working conditions, achieving early, accurate, and reliable diagnosis of wire rope damage under complex working conditions, and significantly improving the technical effect of wire rope detection accuracy and wire rope damage risk prediction accuracy.
[0095] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.
[0096] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.
[0097] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.
Claims
1. A method for detecting wire rope damage based on the fusion of sound patterns and vibration, characterized in that, The method includes: Obtain the current operating load and current service time of the target wire rope; The application information attributes of the target wire rope are obtained, and the operation records of historical wire ropes of the same type that are running normally within a preset service period are retrieved based on the application information attributes to construct a sample wire rope load set. Acoustic and vibration signals from historical operating records of the same type of wire rope in the sample wire rope load set were collected to construct a sample acoustic signal set and a sample vibration signal set. Using the sample steel wire rope load set as input features and the sample sound pattern signal set as supervision features, a sound pattern signal predictor is constructed; and using the sample steel wire rope load set as input features and the sample vibration signal set as supervision features, a vibration signal predictor is constructed. Among them, the service time of each historical wire rope of the same type corresponding to the sample wire rope load set is obtained to construct the sample service time set; Based on the variation characteristics of the voiceprint signals in the sample voiceprint signal set with service time, a first cluster analysis is performed on the sample service time set to obtain multiple voiceprint service time intervals. Based on the multiple service time intervals of the sound patterns, the sample wire rope load set and the sample sound pattern signal set are divided to obtain multiple sound pattern grouped load subsets and multiple sample sound pattern signal subsets; Each of the aforementioned voiceprint group load subsets is used as input features, and the corresponding sample voiceprint signal subsets are used as supervision features to construct multiple voiceprint signal prediction branches, thus forming the voiceprint signal predictor. Based on the variation characteristics of vibration signals with service time in the sample vibration signal set, a second cluster analysis is performed on the sample service time set to obtain multiple vibration service time intervals. Based on the multiple vibration service time intervals, the sample wire rope load set and the sample vibration signal set are divided to obtain multiple vibration group load subsets and multiple sample vibration signal subsets; The vibration signal predictor is formed by constructing multiple vibration signal prediction branches using each vibration group load subset as input features and the corresponding sample vibration signal subset as supervision features. Based on the current service time, a target voiceprint service time interval is determined from the plurality of voiceprint service time intervals, and a corresponding target voiceprint signal prediction branch is selected in the voiceprint signal predictor based on the target voiceprint service time interval. The current operating load is input into the target voiceprint signal prediction branch to obtain the predicted voiceprint signal; Based on the current service time, a target vibration service time interval is determined from the plurality of vibration service time intervals, and a corresponding target vibration signal prediction branch is selected in the vibration signal predictor based on the target vibration service time interval. The current operating load is input into the target vibration signal prediction branch to obtain the predicted vibration signal; Collect the actual acoustic and vibration signals of the target wire rope under the current operating load; A first deviation is obtained based on the predicted voiceprint signal and the actual voiceprint signal, and a second deviation is obtained based on the predicted vibration signal and the actual vibration signal. When the first deviation is greater than or equal to the first deviation threshold, or the second deviation is greater than or equal to the second deviation threshold, damage prediction is performed based on the actual sound pattern signal and the actual vibration signal to obtain the wire rope damage probability. When the probability of the steel wire rope breaking is greater than or equal to the probability threshold, the target steel wire rope is forcibly deactivated.
2. The method according to claim 1, characterized in that, A first deviation is obtained based on the predicted voiceprint signal and the actual voiceprint signal, and a second deviation is obtained based on the predicted vibration signal and the actual vibration signal, including: Feature extraction is performed on the predicted voiceprint signal and the actual voiceprint signal to obtain the predicted voiceprint feature vector and the actual voiceprint feature vector; The first deviation is obtained based on the predicted voiceprint feature vector and the actual voiceprint feature vector; Feature extraction is performed on the predicted vibration signal and the actual vibration signal to obtain the predicted vibration feature vector and the actual vibration feature vector; The second deviation is obtained based on the predicted vibration feature vector and the actual vibration feature vector.
3. The method according to claim 2, characterized in that, Based on the actual acoustic signature signal and the actual vibration signal, damage prediction is performed to obtain the probability of wire rope breakage, including: The actual sound signature feature vector and the actual vibration feature vector are merged to construct the actual signal feature vector; Using the actual signal feature vector as the retrieval condition, similar feature wire rope samples with feature vector similarity greater than or equal to a preset similarity threshold are collected to construct a similar sample set, and the total number of similar samples in the similar sample set is obtained. The number of damaged samples is obtained by counting the number of steel wire ropes that broke within a preset service period of the similar samples. The probability of wire rope breakage is obtained by the ratio of the number of damaged samples to the total number of similar samples.
4. The method according to claim 1, characterized in that, When the first deviation is less than the first deviation threshold and the second deviation is less than the second deviation threshold, the target wire rope is marked as operating normally, and wire rope monitoring continues.
5. The method according to claim 1, characterized in that, When the probability of the wire rope breaking is less than the probability threshold, a maintenance reminder message is output.