Intelligent fault diagnosis method for wrapping robot

Abstract

The invention provides an intelligent fault diagnosis method for a wrapping robot, and the method comprises the following steps: collecting the winding speed, tension and angle of a cable through an information collection assembly, so as to form basic data information; after the basic data information is preprocessed, first-level diagnosis information is obtained through calculation, and first-level early warning is formed; when the first-level diagnosis information is greater than a set threshold value, triggering and starting second-level early warning, and judging whether specific class faults exist or not through a machine learning algorithm; if class faults exist, three-level early warning is started, the severity and urgency of the faults are divided according to a formula (1), and alarm levels are formed. When the winding quality problem is detected, real-time alarming can be performed, so that the production loss is reduced, the quality of a finished cable is improved, and meanwhile, the dependence on manual processing is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of wrapping fault analysis, and in particular to an intelligent fault diagnosis method for a wrapping robot. Background Art

[0002] In the cable production process, cable winding is a key step that directly affects the quality of the finished cable. At present, although automated equipment such as winding robots have partially replaced manual winding, there are still some problems. On the one hand, the automated equipment lacks effective real-time monitoring means, resulting in problems that cannot be discovered and adjusted in time during the winding process, thus affecting the quality of the cable; on the other hand, once the equipment fails or the winding effect is poor, there is a lack of effective grading processing mechanism and real-time alarm system, and measures cannot be taken in time, resulting in production losses. The current frontier direction in the field is mainly focused on improving the accuracy of monitoring and the intelligent level of fault diagnosis. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide an intelligent fault diagnosis method for a cable winding robot, which aims to monitor and process data in real time and in a graded manner to improve the monitoring accuracy of the cable winding process, and to give a real-time alarm when winding quality problems are detected, thereby reducing production losses, improving the quality of finished cable products, and reducing dependence on manual processing.

[0004] In order to solve the above technical problems, the technical solutions provided by the present invention are as follows:

[0005] The intelligent fault diagnosis method of the wrapping robot comprises the following steps:

[0006] Using the information collection component, the winding speed, tension and angle of the cable are collected to form basic data information;

[0007] After preprocessing the basic data information, the first-level diagnostic information is calculated to form a first-level warning;

[0008] When the first-level diagnostic information is greater than the set threshold, the second-level warning is triggered and the machine learning algorithm is used to determine whether there is a specific type of fault;

[0009] If there is a class fault, the third-level warning is activated, which divides the severity and urgency of the fault according to formula (1) and forms an alarm level;

[0010] L alarm =f(S status， T response , C severity )L alarm =f(S status , T response , C severity ) (1)

[0011] Among them, L alarm is the alarm level; T response is the fault response time; C severity is the severity of the fault.

[0012] Furthermore, the information collection component specifically includes a rotation speed sensor, a strain force sensor and an angle sensor. The rotation speed sensor is mounted on the rotating shaft of the wrapping device, and the strain force sensor is arranged closely between the cable and the wrapping wheel.

[0013] Furthermore, the cable winding state is specifically calculated by formula (2):

[0014]

[0015] Among them, S status Indicates the comprehensive score of the cable winding status, w i is the weight of the i-th parameter, x i is the actual measured value of the i-th parameter, n is the total number of parameters, at least 3, including winding speed parameter, tension parameter and speed angle parameter respectively.

[0016] Furthermore, the secondary warning classifies the processed basic data information through a machine learning algorithm, and analyzes each type of basic data information to determine whether there is a fault in this type of basic data information.

[0017] Furthermore, the machine learning algorithm specifically classifies basic data information through a decision tree, and compares the classified data with set data to determine whether there is a class fault.

[0018] Furthermore, the machine learning algorithm also includes in-depth analysis of winding data to identify potential fault modes. Specifically, a preliminary fault identification model composed of a convolutional neural network is trained and verified through a database constructed through historical data to obtain a fault identification model.

[0019] Furthermore, the fault identification model is specifically trained by a loss function, as shown in formula (3):

[0020]

[0021] Among them, L(θ) is the loss function, m is the number of samples, and y i is the true label, is the probability predicted by the model.

[0022] Furthermore, it also includes the optimization and updating of the fault identification model, specifically, regularly using the latest basic data information to train and optimize the model, and using transfer learning to add new fault modes to the existing fault identification model.

[0023] Furthermore, it also includes building a fault database to facilitate comparison of real-time monitoring data with the fault database to achieve real-time fault diagnosis.

[0024] Furthermore, the alarm level also includes the integrated alarm system notifying the operator in real time, and performing graded processing according to the severity of the problem, and forming different graded response measures according to different graded alarms.

[0025] The present invention has the following beneficial effects:

[0026] The intelligent fault diagnosis method for the winding robot in the present invention improves the monitoring accuracy of the cable winding process through real-time monitoring and hierarchical data processing, and can give a real-time alarm when winding quality problems are detected, thereby reducing production losses, improving the quality of finished cable products, and reducing dependence on manual processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The present invention provides a flowchart of the intelligent fault diagnosis method for a wrapping robot. DETAILED DESCRIPTION

[0028] The present invention is described in detail below in conjunction with the various embodiments shown in the accompanying drawings, but it should be noted that these embodiments are not limitations of the present invention, and any equivalent transformations or substitutions in functions, methods, or structures made by ordinary technicians in the field based on these embodiments are all within the scope of protection of the present invention.

[0029] Refer to the attached Figure 1 As shown, the solution in the present invention is mainly used in the winding process in cable production, and the specific steps are as follows:

[0030] The first step is to use high-precision sensors to monitor the cable winding process in real time and collect key parameters such as winding speed, tension, and angle.

[0031] In this embodiment, in order to improve the monitoring efficiency of the cable winding process, a multi-level monitoring system is developed, in which the first level involves the use of high-precision sensors to achieve real-time acquisition of key parameters of cable winding. The sensors are specially configured to measure the following core parameters:

[0032] Winding speed: By installing a speed sensor, it can accurately capture the rotation speed of the cable on the wrapping machine, thereby ensuring the smoothness of the winding process.

[0033] Tension monitoring: Using strain gauge sensors, these sensors are closely attached between the cable and the wrapping wheel, and real-time feedback of cable tension changes during the winding process is provided to maintain the optimal tension state.

[0034] Winding angle: Use a photoelectric encoder or other angle sensor to monitor the cable winding angle to ensure that the cable is accurately wound according to the established pattern.

[0035] First, in the data acquisition stage, each sensor is equipped with a signal conditioning circuit to convert the raw signal into a digital signal for subsequent processing. The following is a mathematical description of sensor data acquisition:

[0036] Where v(t) represents the winding speed at time t, and θ represents the winding angle.

[0037] In the data processing unit, the collected data is processed at multiple levels using the following formula:

[0038] Here T is the tension in the cable, F is the force measured by the strain gauge sensor, A is the cross-sectional area of ​​the cable, and k is the proportional constant.

[0039] When the monitoring winding parameters exceed the preset range, the system will trigger the alarm mechanism. The specific implementation is as follows:

[0040] Install and calibrate the sensor to ensure it fits snugly with the cable and taping equipment.

[0041] Sensor data is monitored in real time and analyzed by the data processing unit.

[0042] When it detects that the winding speed, tension or angle deviates from the normal range, the system immediately activates the alarm system.

[0043] The alarm information is displayed through the user interface and accompanied by sound prompts. At the same time, the system automatically adjusts relevant parameters to restore the winding quality.

[0044] Record fault data to facilitate subsequent analysis and continuously optimize the winding process.

[0045] Through the above solution, the monitoring accuracy of cable winding can be significantly improved, and timely alarms can be issued when quality problems occur, thereby reducing production losses.

[0046] The second step is to process the collected data through a multi-level data processing algorithm to achieve preliminary diagnosis and early warning of faults.

[0047] In the cable winding monitoring system, the collected data needs to go through a sophisticated multi-level data processing algorithm to achieve the initial diagnosis and early warning function of the fault. The following are the detailed implementation steps of the processing algorithm:

[0048] First, the multi-level data processing algorithm

[0049] Preprocessing stage: Filter the collected raw data to eliminate sensor noise and random errors. Apply linear interpolation or more complex signal processing techniques to fill in missing values ​​or outliers in the data.

[0050] Primary diagnosis: Use the following formula to make a preliminary assessment of the cable's winding condition:

[0051]

[0052] Among them, S status Indicates the comprehensive score of the cable winding status, w i is the weight of the i-th parameter, x i is the actual measured value of the ith parameter, n is the total number of parameters, at least 3, including winding speed parameter, tension parameter and speed angle parameter. If S status exceeds the predetermined threshold, the next level of diagnosis is triggered.

[0053] Secondary warning: A more in-depth analysis of the potential problem parameters identified in the initial diagnosis. Apply machine learning algorithms, such as support vector machines (SVM) or decision trees, to classify the data and determine whether a fault exists.

[0054] Three-level alarm mechanism:

[0055] When the second-level warning confirms that there is a fault, the system will activate the third-level alarm mechanism.

[0056] The alarm level is divided according to the severity and urgency of the fault and is determined by the following formula: alarm =f(S status , T response , C severity )L alarm =f(S status , T response , C severity ) (1)

[0057] Among them, L alam is the alarm level; T response is the fault response time; C severity is the severity of the fault.

[0058] The specific process is as follows:

[0059] In the preprocessing stage, a low-pass filter is used to remove high-frequency noise and retain the signals related to the winding process.

[0060] In primary diagnosis, weights are assigned to each parameter based on historical data and expert experience, and a comprehensive score is calculated.

[0061] If the score is outside the normal range, it will enter the second-level warning and classify the data through the trained model.

[0062] After the fault is confirmed, the alarm level is determined by the above formula according to the urgency of the fault and the possible losses.

[0063] The alarm information will be displayed through the human-machine interface (HMI) of the control system, and corresponding measures will be taken according to the alarm level, such as issuing an audible warning, suspending the production line, notifying maintenance personnel, etc.

[0064] Through this multi-level data processing algorithm, the system can promptly and accurately diagnose potential problems in the cable winding process and take corresponding early warning measures, greatly improving production efficiency and the quality of finished cable products.

[0065] Step 3. Introduce artificial intelligence algorithms to conduct in-depth analysis of winding data to identify potential failure modes.

[0066] In-depth analysis and fault identification module

[0067] In order to further improve the fault identification capability during cable winding, a deep learning algorithm based on artificial intelligence can also be introduced. The details are as follows:

[0068] First, data preparation:

[0069] In this embodiment, the collected sensor data is normalized to eliminate the influence of different dimensions:

[0070] In the above formula, X norm is the normalized data, X is the original data, and X min and X max are the minimum and maximum values ​​of the data respectively.

[0071] The processed data is divided into training set and test set for subsequent training and verification of deep learning models.

[0072] Second, deep learning model selection and training

[0073] Convolutional neural network (CNN) is used as the main fault identification model because of its strong feature extraction capability and is suitable for processing unstructured data.

[0074] The CNN model structure is designed as follows:

[0075] Input layer: receives normalized sensor data.

[0076] Convolutional layer: Use multiple convolution kernels to extract data features.

[0077] Pooling layer: reduces the data dimension and retains important information.

[0078] Fully connected layer: maps the extracted features to fault categories.

[0079] Output layer: outputs fault identification results.

[0080] The model is trained using the following loss function:

[0081]

[0082] Among them, L(θ) is the loss function, m is the number of samples, and y i is the true label, is the probability predicted by the model.

[0083] Again, fault identification and alarm.

[0084] The trained model is applied to the sensor data collected in real time to identify the fault mode.

[0085] When the failure probability predicted by the model exceeds the preset threshold, the alarm mechanism is triggered:

[0086] P fault >θ alarm ; Among them, P fault The failure probability θ predicted by the model alarm is the alarm threshold.

[0087] The alarm system takes appropriate measures according to the type and severity of the fault, such as adjusting equipment parameters, notifying operators or automatically shutting down.

[0088] Finally, the model is optimized and updated.

[0089] New data is collected regularly to retrain and optimize the model to adapt to changes in the production process.

[0090] Use techniques such as transfer learning to leverage the infrastructure of existing models to quickly adapt to new failure modes.

[0091] Assume that after deep learning model analysis, it is found that a cable has a potential fault of uneven tension during the production process. The system will monitor the tension sensor data in real time and perform feature extraction and fault identification through the CNN model. Once the model identifies that the probability of failure exceeds the threshold, an alarm will be immediately issued through the HMI system, and relevant equipment parameters will be automatically adjusted to optimize the winding process. Fault data will be recorded for subsequent model training and fault prevention. Through the implementation of the above deep analysis and fault identification modules, this solution can efficiently identify various fault modes in the cable winding process, significantly improving production efficiency and cable product quality.

[0092] The fourth step is to establish a fault database, compare the real-time monitoring data with the fault database, and achieve accurate diagnosis of the fault.

[0093] In order to improve the accuracy of fault diagnosis, a fault database can be designed to collect and store parameter characteristics of various known cable winding faults. By dynamically comparing the real-time monitoring data with the records in the fault database, the system can quickly and accurately identify the specific fault type.

[0094] The specific fault database construction process is as follows:

[0095] 1) Data collection:

[0096] Collect case data of various cable entanglement failures that have occurred in history, including but not limited to sensor measurement data, environmental parameters, operating conditions, etc.

[0097] These data are sorted and classified to form a standard fault data set.

[0098] 2) Feature extraction:

[0099] Feature extraction is performed on each fault case, including statistical features (such as mean, variance, frequency domain features, etc.) and feature representation based on deep learning.

[0100] The feature extraction process can be expressed using the following formula: F = φ(X), where F is the extracted feature vector, φ() is the feature extraction function, and X is the original data.

[0101] 3) Database construction:

[0102] The extracted feature vectors and their corresponding fault labels are stored in the database to form a fault knowledge base.

[0103] The database structure design should facilitate rapid retrieval and comparison to improve diagnostic efficiency. With the development of subsequent technologies, the types of faults will also increase, and a database can be established to continuously update the judgment.

[0104] 4) Use the database to conduct real-time monitoring and fault comparison;

[0105] Real-time data stream processing: Collect sensor data in real time and preprocess it, including filtering, normalization, etc. The processed data is converted into feature vectors and compared with the features in the fault database.

[0106] Fault diagnosis algorithm: Use similarity calculation methods, such as Euclidean distance or cosine similarity, to evaluate the similarity between real-time features and fault features in the database.

[0107] The similarity calculation formula is as follows:

[0108] Among them, S is the similarity, F real-time is the real-time feature vector, F db is the feature vector in the database.

[0109] 5) Fault determination and alarm:

[0110] When the similarity exceeds the preset threshold, it is judged as a matching failure and the system triggers an alarm.

[0111] The alarm system provides detailed fault description and response suggestions based on the fault type.

[0112] In the actual production process, the sensor data of cable winding is collected in real time and preprocessed. The preprocessed data is converted into a feature vector through a feature extraction function. The real-time feature vector is compared with the feature vector in the fault database using a similarity calculation method. When the similarity exceeds the threshold, the system immediately displays the fault diagnosis results through the HMI interface and issues an alarm. The operator takes corresponding measures based on the fault description and response suggestions provided by the system, such as adjusting equipment parameters or performing repairs.

[0113] Through the implementation of the above scheme, rapid and accurate diagnosis of faults in the cable winding process can be achieved, effectively reducing fault diagnosis time, improving production efficiency, and reducing potential losses.

[0114] Step 5: When a winding quality problem is detected, the operator is notified in real time through the integrated alarm system and graded according to the severity of the problem.

[0115] When the cable winding quality is detected to be not in compliance with the preset standards, the system will activate a hierarchical real-time alarm mechanism, which dynamically adjusts the alarm level according to the severity and urgency of the winding defect, ensuring that operators can respond quickly.

[0116] 1) Alarm level setting:

[0117] Set multiple alarm levels, for example, level 1 is a warning, level 2 is a minor fault, and level 3 is a major fault.

[0118] Each level is associated with specific parameter thresholds and response measures.

[0119] 2) Real-time data monitoring:

[0120] The data processing unit is used to analyze the sensor data in real time, and once it is detected that the parameters are out of the normal range, the alarm process is immediately triggered.

[0121] 3) Alarm triggering conditions:

[0122] Set the trigger condition as shown in the following formula:

[0123] If T>T threshold or v<v threshold orθ≠θ target , then trigger alarm,

[0124] T is the tension measured in real time, v is the speed, θ is the angle, T threshold 、v threshol d and θ target are the corresponding thresholds respectively.

[0125] 4) Alarm notification process:

[0126] The system sends out alarms through the user interface (UI) and sound signals at the same time, and is integrated with the entire production system control system, such as PLC.

[0127] The alarm information includes fault level, fault description, possible cause and recommended countermeasures.

[0128] Tiered Response Measures:

[0129] Level 1 (warning): The system automatically adjusts relevant parameters and notifies the operator to pay attention.

[0130] Level 2 (minor fault): The system pauses the production line and waits for the operator to confirm and take appropriate measures.

[0131] Level 3 (Major Failure): The system immediately stops the production line and notifies the maintenance team through emergency communication channels.

[0132] The specific implementation process is as follows:

[0133] When the monitoring system detects abnormal winding parameters, an alarm of the corresponding level is triggered according to the preset alarm conditions.

[0134] The system displays alarm information through the UI and issues sound prompts through the speaker.

[0135] The operator takes action based on the alarm level:

[0136] For early warning, monitor parameter changes and adjust actions based on system recommendations.

[0137] For minor faults, check the equipment status and make necessary adjustments or maintenance.

[0138] For major faults, immediately notify the maintenance team for emergency processing.

[0139] Record all alarm events and measures taken for subsequent analysis and improvement.

[0140] This graded alarm system ensures that when quality problems occur during the cable winding process, operators are quickly notified and can take appropriate measures based on the severity of the problem, thereby reducing losses and potential defective rates in the production process.

[0141] In this embodiment, through multi-level early warning and subsequent multi-level response, etc., not only the safety performance is improved, but also the alarm or fault is made clearer.

[0142] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0143] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. Intelligent fault diagnosis method for wrapping robot, characterized in that: The following steps are included: Using the information collection component, the winding speed, tension and angle of the cable are collected to form basic data information; After preprocessing the basic data information, the first-level diagnostic information is calculated to form a first-level warning; When the first-level diagnostic information is greater than the set threshold, the second-level warning is triggered and the machine learning algorithm is used to determine whether there is a specific type of fault; If there is a class fault, the third-level warning is activated, which divides the severity and urgency of the fault according to formula (1) and forms an alarm level; L alarm =f(S status ,T response ,C severity ) (1) Among them, L alarm is the alarm level; T response is the fault response time; C severity is the severity of the fault.

2. The intelligent fault diagnosis method for a wrapping robot according to claim 1, characterized in that: The information collection component specifically includes a rotation speed sensor, a strain force sensor and an angle sensor. The rotation speed sensor is mounted on the rotating shaft of the wrapping device, and the strain force sensor is arranged closely between the cable and the wrapping wheel.

3. The intelligent fault diagnosis method for a wrapping robot according to claim 1, characterized in that: The cable winding state is specifically calculated by formula (2): Among them, S status Indicates the comprehensive score of the cable winding status, w i is the weight of the i-th parameter, x i is the actual measured value of the i-th parameter, n is the total number of parameters, at least 3, including winding speed parameter, tension parameter and speed angle parameter respectively.

4. The intelligent fault diagnosis method for a wrapping robot according to claim 1, characterized in that: The secondary warning classifies the processed basic data information through a machine learning algorithm, and analyzes each type of basic data information to determine whether there is a fault in that type of basic data information.

5. The intelligent fault diagnosis method for a wrapping robot according to claim 4 is characterized in that: The machine learning algorithm specifically classifies basic data information through a decision tree, and compares the classified data with the set data to determine whether there is a class fault.

6. The intelligent fault diagnosis method for a wrapping robot according to claim 1, characterized in that: The machine learning algorithm also includes in-depth analysis of winding data to identify potential failure modes. Specifically, a preliminary fault identification model composed of a convolutional neural network is trained and verified through a database constructed through historical data to obtain a fault identification model.

7. The intelligent fault diagnosis method for a wrapping robot according to claim 6, characterized in that: The fault identification model is specifically trained by a loss function, as shown in formula (3): Among them, L(θ) is the loss function, m is the number of samples, and y i is the true label, is the probability predicted by the model.

8. The intelligent fault diagnosis method for a wrapping robot according to claim 6, characterized in that: It also includes the optimization and updating of the fault identification model, specifically, regularly using the latest basic data information to train and optimize the model, and using transfer learning to add new fault modes to the existing fault identification model.

9. The intelligent fault diagnosis method for a wrapping robot according to claim 1, characterized in that: It also includes building a fault database to facilitate the comparison of real-time monitoring data with the fault database and realize real-time fault diagnosis.

10. The intelligent fault diagnosis method for a wrapping robot according to claim 1, characterized in that: The alarm level also includes the integrated alarm system notifying the operator in real time, and performing graded processing according to the severity of the problem, and forming different graded response measures according to different graded alarms.

Images