Adaptive Stranding Control Method for Cabling Machine Based on Multi-Dimensional Tension Feedback

Abstract

The present invention relates to the technical field of cable manufacturing, and particularly to an adaptive stranding control method for a stranding machine based on multi-dimensional tension feedback, including the following steps: Step 1: Configure a phase marking device on the stranding main shaft to generate a periodic reference signal; Step 2: Conduct a multi-wire dynamic coupling analysis on the pressure distribution characteristics; Step 3: Construct a multi-dimensional feature space including real-time process states and historical fault modes; Step 4: Establish a dynamic lay length prediction model based on the trend component of the stability coefficient; Step 5: Real-time monitor the pressure distribution mutation rate, the strain frequency domain migration rate, and the abnormality degree of the balance index. When the three cooperate to exceed the dynamic threshold, execute a hierarchical control strategy, including progressive speed reduction, pre-tension balance, and a safety shutdown sequence. By real-time monitoring and dynamically adjusting the tension and strain during the stranding process, the stability of the stranding process is effectively improved, and quality problems caused by uneven tension or excessive strain are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of cable manufacturing, and particularly to an adaptive stranding control method for a stranding machine based on multi-dimensional tension feedback. Background Art

[0002] In the field of precision stranding of special cables, certain high-end application scenarios have extremely demanding requirements for the stability of the core structure. For example, during the kilometer-level stranding process of composite cables for deep-sea exploration, traditional monitoring systems can only capture the macroscopic changes in axial tensile force, and lack effective means to sense the hidden stress distribution caused by torsional friction between the stranded cores (such as the microscopic shear effect between adjacent insulating layers). This results in local stress concentration in the stranded cable body under the deep-sea high-pressure environment, manifested as non-uniform wear of the armor layer (the service life is shortened by about 30%), and in severe cases, intermittent interruption of the signal transmission channel.

[0003] Moreover, in the complex working conditions of industrial sites, the dynamic interference problem of stranding equipment is particularly prominent. Taking the continuous production of aerospace wires as an example, the coupling effect of high-frequency electromagnetic fields and mechanical vibrations in the workshop makes the traditional tension feedback signal mixed with pseudo-wave components that have nothing to do with the true deformation of the wire. Existing filtering algorithms are difficult to distinguish such interference from true process disturbances (such as the tension transient caused by the sudden change in the inertia of the wire pay-off reel), resulting in the control system frequently performing ineffective adjustment actions, and ultimately causing the dispersion of the lay length accuracy to exceed the standard (the measured fluctuation range reaches ±0.15 mm), directly affecting the anti-fatigue performance of the wire in a high-frequency vibration environment. Summary of the Invention

[0004] Based on the above objectives, the present invention provides an adaptive stranding control method for a stranding machine based on multi-dimensional tension feedback, including the following steps:

[0005] Step 1: Configure a phase marking device on the stranding main shaft to generate a periodic reference signal, and set a composite tension detection structure on each wire transmission path. This structure includes circumferentially distributed pressure sensing units and axial strain sensing units, and synchronously obtains the radial pressure distribution characteristics, axial strain evolution characteristics of each wire, and the dynamic phase relationship between the two under the trigger of the reference signal;

[0006] Step 2: Perform multi-wire dynamic coupling analysis on the pressure distribution characteristics, extract the intensity of the tension interaction between the wires through time-shift correlation calculation, and generate a tension balance index in combination with the frequency-domain migration law and phase relationship of the strain evolution characteristics;

[0007] Step 3: Construct a multi-dimensional feature space including real-time process states and historical fault modes, and use an adaptive weighting mechanism to fuse the tension balance index with the multi-dimensional features to generate a stranding stability coefficient that reflects both instantaneous fluctuations and trend deterioration;

[0008] Step 4: Establish a pitch dynamic prediction model based on the trend component of the stability coefficient, generate compensation parameters in combination with the material deformation characteristics, and implement micro-step compensation based on the fluctuation component in a specific rotation phase interval;

[0009] Step 5: Real-time monitor the pressure distribution mutation rate, strain frequency domain migration rate, and equilibrium degree index anomaly. When the three cooperate to exceed the dynamic threshold, execute a hierarchical control strategy, including progressive speed reduction, pre-tension balance, and safety shutdown sequence.

[0010] Preferably, the configuration method of the composite tension detection structure includes:

[0011] Dynamically determine the density of the circumferential detection units according to the proportional relationship between the geometric characteristics of the stranding die and the physical properties of the wire, so that the overlapping monitoring areas formed by the adjacent detection unit coverage areas form a preset proportion;

[0012] Dynamically adjust the distance between the detection unit groups based on the real-time stranding process parameters, where the distance adjustment amount depends on the friction characteristics real-time feedback by the wire surface state;

[0013] The implantation depth of the strain sensing unit is dynamically adjusted according to the wire diameter and the material bending stiffness characteristics, and the bending stiffness characteristics are extracted from the pre-stored material characteristics library.

[0014] Preferably, the acquisition process of the real-time stranding process parameters includes:

[0015] Jointly calculate the actual stranding pitch through the spindle rotation phase signal and the wire displacement detection device;

[0016] Dynamically correct the friction characteristic parameters according to the analysis results of the wire surface optical characteristics;

[0017] Periodically update the reference parameter values, and the update trigger condition is associated with the change rate of the stranding stability coefficient.

[0018] Preferably, the specific implementation of the dynamic coupling analysis includes:

[0019] Normalize and preprocess the multi-wire pressure distribution data to eliminate the magnitude difference;

[0020] Establish a time-shift analysis window matching the current production process rhythm, and the window width is dynamically adjusted according to the spindle speed;

[0021] Calculate the correlation intensity of each wire pressure fluctuation sequence at different time-shift amounts, and extract the key indicators characterizing the tension coupling degree;

[0022] Perform dynamic smoothing processing on the key indicators, and the smoothing intensity is negatively correlated with the wire feeding speed.

[0023] Preferably, the implementation process of the adaptive weighting mechanism includes:

[0024] Construct a multi-dimensional influence factor set including the operating status of the equipment, environmental conditions, and material properties;

[0025] Perform dynamic normalization processing on each influence factor, and the normalization interval is automatically adjusted according to the distribution characteristics of on-line monitoring data;

[0026] Generate dynamic weight coefficients for each factor through a fuzzy inference system, and the inference rule base is continuously optimized according to historical process data;

[0027] Non-linearly fuse the weight coefficients with the tension balance degree index, and the form of the fusion function is dynamically switched according to the plastic deformation stage of the material.

[0028] Preferably, the method for constructing the pitch dynamic prediction model includes:

[0029] When the trend component exceeds the material deformation safety boundary, generate a pitch adjustment curve based on the real-time linear velocity and material creep characteristics;

[0030] Adopt an inertial compensation algorithm to smoothly execute pitch adjustment, and the compensation intensity is dynamically optimized according to the rotational inertia of the equipment;

[0031] Synchronously correct the spindle speed during the pitch adjustment process, and the correction amount maintains a dynamic balance relationship with the pitch change rate.

[0032] Preferably, the determination process of the material deformation safety boundary includes:

[0033] Obtain the stress relaxation characteristics of the current wire through an on-line material property analysis device;

[0034] Perform boundary correction by combining the wire break accident data under similar working conditions in the historical process database;

[0035] Perform real-time compensation on the safety boundary according to the environmental temperature fluctuation.

[0036] Preferably, the implementation of the hierarchical control strategy includes:

[0037] Establish a multi-parameter joint abnormality evaluation model, which fuses the time derivative of the pressure distribution mutation rate, the spatial gradient of the strain frequency domain energy migration, and the statistical outliers of the balance degree index;

[0038] Dynamically divide the response levels according to the abnormality evaluation results, and the response level thresholds are adaptively adjusted according to the cumulative operating time of the equipment;

[0039] Introduce die reverse rotation control in the high-level response stage to eliminate residual stress, and the number of rotation cycles is positively correlated with the duration of the abnormality.

[0040] Preferably, it further includes:

[0041] After the abnormal event is processed, a self-learning optimization process is executed, including:

[0042] Collect the complete process cycle data before and after the occurrence of the abnormality, and construct a multi-dimensional event feature vector including equipment status, process parameters, and environmental factors;

[0043] Adopt a feature importance evaluation algorithm to extract key influencing factors, and the weights of the evaluation algorithm are dynamically allocated according to the event type;

[0044] Update the control model parameters through an incremental learning mechanism, and retain the effective feature association patterns;

[0045] When processing similar products, preferentially load the optimized parameter set.

[0046] Preferably, the implementation of the feature importance evaluation algorithm includes:

[0047] Perform principal component decomposition on the multi-dimensional event features, and retain the principal components whose interpretation degree exceeds the preset requirements;

[0048] Construct a deep feature extraction network to mine implicit association rules, and the network structure is dynamically adjusted according to the feature dimensions;

[0049] Dynamically allocate feature weights through an attention mechanism, and focus on the feature dimensions strongly associated with the current event type.

[0050] Advantages of the present invention:

[0051] Through real-time monitoring and precise control, the tension and strain during the stranding process are ensured to be within a reasonable range, avoiding product defects caused by uneven tension or excessive strain. The combination of the adaptive weighting mechanism and the dynamic prediction model enhances the stability and flexibility of the stranding process, and can adjust the control strategy in real time according to different working conditions, greatly improving production efficiency and product quality. At the same time, through the hierarchical control strategy and the self-learning optimization mechanism, the method can quickly respond when an abnormality occurs, and realize self-optimization in subsequent production, forming a closed-loop feedback, improving the self-adaptability and reliability of the overall system. Description of the Drawings

[0052] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only those of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0053] Figure 1 It is the step flow chart of the method of the present invention;

[0054] Figure 2Flowchart of the steps for constructing the pitch dynamic prediction model in the method of the present invention;

[0055] Figure 3 Flowchart of the steps for performing the self - learning optimization process after abnormal event handling in the method of the present invention. Detailed implementation manners

[0056] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments. For some well - known technologies, those skilled in the art can also adopt other alternative methods for implementation; moreover, the accompanying drawings are only for more specifically describing the embodiments and are not intended to specifically limit the present invention.

[0057] Please refer to Figures 1 - 3 , the embodiment of the present invention provides a cable - making machine adaptive stranding control method based on multi - dimensional tension feedback. In step 1, a phase marking device is configured on the stranding main shaft to generate a periodic reference signal, and a compound tension detection structure is set on the transmission path of each wire, including a circumferential pressure sensing unit and an axial strain sensing unit. Triggered by the reference signal, the system synchronously obtains the radial pressure distribution characteristics and axial strain evolution characteristics of each wire, and at the same time obtains their dynamic phase relationship. By this method, the tension distribution and strain state during the stranding process are monitored in real time, providing basic data for subsequent analysis.

[0058] In step 2, after obtaining the pressure and strain data, a multi - wire dynamic coupling analysis method is used to analyze the obtained pressure distribution characteristics. Through time - shift correlation calculation, the intensity of the tension interaction between wires is extracted, and combined with the frequency - domain migration law and phase relationship of the strain evolution characteristics, a tension balance degree index is generated. The function of this step is to evaluate and quantify the coupling degree of the tension between wires, providing a basis for the control strategy.

[0059] In step 3, by constructing a multi - dimensional feature space including real - time process states and historical failure modes, combined with an adaptive weighting mechanism, the tension balance degree index is fused with other process characteristics (such as equipment status, environmental parameters, material properties, etc.) to generate a stranding stability coefficient. This coefficient can reflect both instantaneous fluctuations and trend deterioration, helping to judge the stability of the stranding process. This stability coefficient is the core of subsequent decision - making and control.

[0060] In step 4, based on the trend component of the generated stability coefficient, this step generates compensation parameters by establishing a pitch dynamic prediction model and combining the deformation characteristics of the material. During the stranding process, micro-step compensation is implemented according to the fluctuation component in a specific rotation phase interval to dynamically adjust the pitch, ensuring the stability and accuracy of the stranding process. This step can adjust the stranding process parameters in real time to avoid quality problems caused by unstable processes.

[0061] In step 5, by real-time monitoring the mutation rate of pressure distribution, the migration rate of strain frequency domain, and the abnormality degree of the balance index, when the three cooperate to exceed the dynamic threshold, a hierarchical control strategy is triggered. This includes a progressive speed reduction, pre-tension balance, and a safety shutdown sequence to ensure that the system can respond in a timely manner and take measures in case of abnormalities, preventing the further expansion of faults or equipment damage.

[0062] By real-time monitoring and dynamic adjustment of the tension and strain during the stranding process, the stability of the stranding process is effectively improved, and quality problems caused by uneven tension or excessive strain are reduced. In addition, by adopting an adaptive weighting mechanism and multi-dimensional feature fusion, the stranding control becomes more flexible and can adapt to different process changes and environmental fluctuations. By constructing a pitch dynamic prediction model and implementing micro-step compensation, the stranding accuracy is further improved. Finally, the hierarchical control strategy can ensure timely response in case of abnormalities, effectively avoiding equipment failures or production interruptions, and enhancing the reliability and efficiency of the entire production process.

[0063] In a possible implementation, according to the proportional relationship between the geometric characteristics of the stranding die and the physical properties of the wire, the density of the circumferential detection units in the composite tension detection structure is dynamically determined. Specifically, the arrangement density of the circumferential detection units is calculated based on the diameter of the wire, material characteristics, and the geometric shape of the stranding die. With such a configuration, the coverage areas of adjacent detection units will form an overlapping monitoring area with a preset ratio, ensuring more accurate monitoring of the tension and strain distribution. This configuration method can ensure that the tension and strain changes in different regions can be fully sensed and monitored, avoiding imbalances in local regions.

[0064] During the stranding process, the spacing of the detection unit group needs to be dynamically adjusted according to the real-time stranding process parameters. Especially when the surface state of the wire changes (such as changes in surface friction characteristics), the spacing is adjusted based on the real-time feedback of the friction characteristics. This method ensures that when the friction changes significantly, the monitoring accuracy of the detection units will not be affected. By adjusting the spacing, the accuracy of tension and strain detection can be improved, thus avoiding monitoring blind spots or redundancies caused by too large or too small spacing.

[0065] The implantation depth of the strain sensing unit is dynamically adjusted according to the diameter of the wire and the flexural stiffness characteristics of the material. Specifically, by pre-storing the flexural stiffness data in the material property library, the depth of the strain sensing unit can be dynamically adjusted on wires of different materials and diameters. This adjustment can ensure that the sensing unit can adapt to wires of different materials and specifications, and ensure that strain information can be effectively sensed and captured under various working conditions. By adjusting the flexural stiffness characteristics, the accuracy of strain sensing can be enhanced, especially in adapting to strain changes in harder or softer wires.

[0066] This configuration method of the composite tension detection structure can significantly improve the monitoring accuracy and control efficiency of the stranding machine. First, by dynamically adjusting the density and spacing of the circumferential detection units, it can ensure that the monitoring of tension and strain is more refined under various stranding conditions, avoiding blind spots or redundant monitoring caused by unreasonable spacing and density settings. In addition, the dynamic depth adjustment method of the strain sensing unit enables appropriate strain monitoring for wires of different materials and specifications, improving the adaptability and versatility of the system. Finally, these configurations can enhance the stability and stranding quality of the entire stranding process, effectively avoiding monitoring errors caused by improper sensor configuration, and ensuring the reliable operation of the stranding machine under variable working conditions.

[0067] In a possible implementation, the real-time obtained spindle rotation phase signal and the feedback signal of the wire displacement detection device are used jointly to accurately calculate the actual stranding pitch. Specifically, the spindle rotation phase signal can provide the angular position information of the spindle, while the wire displacement detection device calculates the actual movement trajectory of the wire by monitoring the displacement of the wire during the stranding process. By combining these two signals, accurate actual stranding pitch data can be obtained, ensuring that the movement parameters of the wire in each stranding cycle are recorded and fed back in real time, and further providing basic data for subsequent tension control and process optimization.

[0068] Dynamically correct the friction characteristic parameters according to the analysis results of the optical characteristics of the wire surface. This process uses an optical detection device on the wire surface to scan and analyze the microscopic characteristics (such as surface roughness, glossiness, etc.) of the wire surface in real time, and then judge the change of its friction characteristics. These data can help the system adjust the friction parameters in real time, so as to compensate for the friction force fluctuation caused by the change of the wire surface state. The change of the friction force has an important impact on the tension distribution during the stranding process. Therefore, by dynamically correcting the friction characteristics, the tension error caused by inaccurate friction characteristics can be effectively avoided, and the stranding accuracy can be improved.

[0069] The reference parameter values (such as stranding speed, tension reference value, etc.) need to be updated periodically. Each time an update occurs, the system adjusts the trigger condition for the update based on the change rate of the current stranding stability coefficient. The stability coefficient is usually related to factors such as vibration and tension fluctuations during the stranding process. By monitoring the changes in these factors, it is possible to determine whether the stranding process is in a stable state. If the system detects that the change in the stability coefficient during the stranding process exceeds a preset threshold, the update of the reference parameters will be triggered to ensure that the stranding process is always in an optimized state. In this way, the stability of the stranding process can be continuously improved, and operation deviations caused by outdated process parameters can be avoided.

[0070] By combining the spindle rotation phase signal with the feedback from the wire displacement detection device, the stranding pitch can be accurately obtained, thereby providing accurate real-time data for the tension control system and ensuring the pitch accuracy during the stranding process. Dynamically correcting the friction characteristic parameters can cope with the friction force fluctuations caused by changes in the surface state of the wire, avoiding inaccurate tension control caused by unstable friction characteristics. Periodically updating the reference parameters and adjusting the trigger condition according to the change in the stability coefficient can optimize the stranding process in real time, avoiding production instability problems caused by lagging process parameters. The implementation of these steps greatly improves the adaptive ability of the cabling machine under dynamic working conditions, enhances the stranding accuracy and stability, and ensures an efficient and stable production process.

[0071] In a possible implementation, before analyzing the pressure distribution data of multiple wires, it is first normalized to eliminate interference caused by magnitude differences. Different wires may generate pressure data of different magnitudes during the stranding process due to factors such as material, shape, and surface smoothness. To ensure that the pressure data of each wire has the same reference value during the analysis process, the normalization process converts these data to the same magnitude, enabling them to be effectively compared and comprehensively analyzed. This process can reduce the impact of the pressure fluctuations of a single wire on the overall analysis result, ensuring that the pressure data of each wire in the entire stranding system can be reasonably analyzed under a unified standard.

[0072] As the spindle speed changes, the production process rhythm of the cabling machine also changes. Therefore, the width of the dynamic coupling analysis window needs to match the current production process rhythm. By real-time monitoring the spindle speed, the width of the time-shift analysis window is dynamically adjusted so that the data segment analyzed each time can accurately capture the stranding state and pressure fluctuations at the spindle speed. This dynamic adjustment of the time-shift window can ensure that the analyzed data range always meets the actual process requirements at different production speeds, avoiding data loss or incomplete analysis caused by an overly large or small analysis window.

[0073] In the time-shift analysis window, by calculating the correlation strength of different wire pressure fluctuation sequences at different time-shift amounts, the coupling relationship of pressure fluctuations between different wires can be revealed. Specifically, the correlation calculation of the pressure fluctuation sequence can use the correlation coefficient or cross-correlation method to evaluate the degree of cooperative movement or mutual interference between wires. This step can help analyze the cooperation situation of each wire during the stranding process, reveal which wires have a strong coupling effect on the tension fluctuation, and thus provide a basis for subsequent tension optimization.

[0074] For the key indicators characterizing the tension coupling degree extracted from the pressure fluctuation sequence, dynamic smoothing processing is required to remove abnormal data caused by short-term fluctuations or noise interference. The intensity of the smoothing processing is negatively correlated with the wire feeding speed. That is, when the feeding speed is fast, the intensity of the smoothing processing is small to avoid losing important signals due to over-smoothing; while when the wire feeding speed is slow, the smoothing intensity is large to reduce the wrong analysis caused by accidental interference or sudden fluctuations. This dynamic smoothing strategy can effectively balance the fineness of the data and the stability of the signal, ensuring reliable coupling analysis results in actual production.

[0075] By performing normalization preprocessing on the multi-wire pressure distribution data, the influence of the pressure magnitude difference of different wires on the analysis results can be eliminated, and the comparability of the data can be improved. Dynamically adjusting the width of the time-shift analysis window enables each analysis to match the current production rhythm, thereby improving the accuracy and timeliness of the analysis. By calculating the correlation of the wire pressure fluctuation sequence, the degree of tension coupling between wires can be deeply understood, providing a theoretical basis for tension control. The dynamic smoothing processing of key indicators can effectively suppress short-term fluctuations and noise interference, avoiding the influence of false signals on the control system, and thus improving the stability and reliability of the stranding process. The implementation of these steps can significantly enhance the adaptive control ability of the cabling machine under multi-dimensional tension feedback, optimize the stranding process, and ensure the stability of the quality of the cabled product.

[0076] In a possible implementation manner, in the adaptive weighting mechanism, first, a set containing multiple factors needs to be constructed. These factors cover the operating state of the equipment, environmental conditions, and material properties. These factors can affect the stranding process of the cabling machine from different perspectives and thus need to be considered comprehensively. For example, the operating state of the equipment includes the spindle speed, load conditions, etc.; environmental conditions can involve factors such as temperature and humidity; material properties include physical properties such as the strength and ductility of the wire. These factors are the key variables affecting the tension control during the stranding process and the quality of the final product.

[0077] Each influencing factor may have different magnitudes and units, so they need to be normalized so that they can be compared and fused on the same scale. The normalization interval should be automatically adjusted according to the distribution characteristics of the on-line monitoring data, which means that the system will monitor the change range of each influencing factor in real time and dynamically adjust the normalization method according to the current production conditions. This flexible normalization process can ensure that under different production conditions, the changes of each factor can be properly measured and adjusted, thus improving the accuracy of analysis.

[0078] Based on the multi-dimensional set of influencing factors, a fuzzy inference system is used to generate the dynamic weight coefficients of each factor. The fuzzy inference system determines the relative importance of each factor by analyzing historical process data and combining the current production conditions. The inference rule base can be continuously optimized based on the analysis of a large amount of historical process data to adapt to different production requirements and condition changes. The advantage of the fuzzy inference system is that it can handle uncertainty and ambiguity, especially when the process parameters are relatively complex or difficult to accurately quantify, it can reasonably estimate the weights of each factor.

[0079] The generated dynamic weight coefficients are combined with the tension balance degree index and processed through a non-linear fusion function. The core of this process is to dynamically switch the form of the fusion function according to the plastic deformation stage of different materials. During the plastic deformation stage of the material, different material properties will have different effects on the stranding tension, so specific non-linear functions are needed to process the tension balance degree at different stages. For example, in the initial stage of material deformation, the focus of tension control may be to avoid excessive stretching, while when the material reaches a higher deformation stage, the focus may shift to optimizing the distribution and balance of tension. Dynamically switching the form of the fusion function can flexibly adapt to different production states, thus optimizing tension control.

[0080] By constructing a multi-dimensional set of influencing factors including equipment operating status, environmental conditions, and material properties, the impacts of multiple factors on the tension control of the cabling machine can be comprehensively considered, making the control more precise and accurate. The dynamic normalization process ensures that the impacts of each factor can be compared on the same scale, avoiding the interference of factors with different magnitudes on the analysis results. The application of the fuzzy inference system can intelligently adjust the weight coefficients of each factor in a complex and uncertain production environment, thereby optimizing the tension control effect. Non-linearly fusing the weight coefficients of the tension balance degree index can intelligently adjust the control strategy according to different deformation stages of the material, ensuring the best tension distribution in different process stages. The comprehensive application of these steps not only improves the accuracy of the adaptive stranding control method but also enables the cabling machine to maintain stable working performance in a changing production environment, further improving the cabling quality and production efficiency.

[0081] In a possible implementation, when performing dynamic pitch prediction, it is first necessary to monitor the safety margin of material deformation. When it is detected that the trend component (such as the change trend of tension or speed) exceeds the safety margin of material deformation, the system will issue a warning signal and activate the pitch adjustment mechanism. The safety margin of material deformation is determined based on factors such as the creep characteristics, ductility, and mechanical properties of the material. When these trend components exceed the set range, it means that there may be a risk of overstretching or rupture. Therefore, measures must be taken to adjust the pitch in real time to avoid damaging the material.

[0082] Once the trend component exceeds the safety margin, the system will generate a pitch adjustment curve based on the real-time linear velocity and the creep characteristics of the material. The real-time linear velocity reflects the actual processing speed of the material, while the creep characteristics of the material determine its deformation behavior at different speeds. By combining these two factors, the system can generate an optimal pitch adjustment curve to ensure that the material does not exceed the safe deformation range during the adjustment process.

[0083] During the pitch adjustment process, in order to reduce the vibration and instability of the equipment, the system uses an inertia compensation algorithm to smooth the pitch adjustment process. The inertia compensation algorithm dynamically optimizes the compensation intensity and makes real-time adjustments according to the change of the equipment's moment of inertia. This process ensures that during the pitch adjustment, an unstable operating state will not be caused due to the lag reaction of the equipment's inertia, thus ensuring the control accuracy and the smooth operation of the equipment.

[0084] During the pitch adjustment process, the system will synchronously correct the spindle speed. The correction amount maintains a dynamic balance relationship with the pitch change rate. That is, the adjustment of the spindle speed is associated with the change amount of the pitch to ensure the mutual coordination between the two. This correction process can keep the wire tension stable during the stranding process, thus avoiding tension fluctuations caused by too fast or too slow speeds and ensuring the smooth progress of the stranding process.

[0085] The construction method of the lay length dynamic prediction model can significantly improve the operation stability and production efficiency of the stranding machine. First, by real-time monitoring the safety boundary of material deformation, it can effectively avoid damage caused by excessive stretching or deformation of materials. Second, by combining the real-time line speed with the creep characteristics of the material to generate a lay length adjustment curve, the system can adjust the lay length according to the actual situation to ensure that the material is always within the safe deformation range during processing. The application of the inertia compensation algorithm further smooths the lay length adjustment process and reduces the operation instability of the equipment caused by inertia lag. Finally, by synchronously correcting the spindle speed and the lay length change rate, the system ensures the accuracy and consistency of the stranding process, thereby improving the cable laying quality and production efficiency and reducing the failure rate of the equipment. The effective combination of these steps not only improves the intelligent level of the production process but also ensures the safety and stability of the production process, reducing downtime and material waste caused by equipment instability.

[0086] In a possible implementation, the system first performs real-time detection on the wire being processed through an on-line material property analysis device installed on the stranding equipment. This device mainly detects the stress relaxation characteristics of the material under different tensions and lay lengths (i.e., the stress decay process that occurs over time under a constant tensile state) to obtain the immediate mechanical properties of this batch of materials. These characteristic data are used to evaluate the ultimate deformation ability of the material under the current working conditions and serve as an important reference for setting the initial safety boundary.

[0087] After obtaining the real-time material properties, the system compares the collected data with the historical process database. The database records the wire breakage accidents and operation parameters under different material types, tension levels, line speeds, and environmental conditions. The system analyzes the critical conditions for wire breakage at that time by matching the current processing conditions with the "similar working conditions" cases in the historical data, thereby making an empirical correction to the initial safety boundary to improve the practical adaptability and reliability of the safety boundary.

[0088] The mechanical behavior of materials is significantly affected by the environmental temperature. The system is equipped with temperature sensors to real-time monitor the temperature changes in the processing environment and incorporates the temperature factor into the safety boundary model. When it detects that the temperature rises or falls beyond the set range, the system will dynamically adjust the safety boundary according to thermodynamic data such as the material's thermal expansion coefficient and creep characteristics to adapt to the material property fluctuations brought about by temperature changes. For example, in a high-temperature environment, the material is more prone to creep, and the system will automatically lower the safety boundary to enhance the protection measures.

[0089] The above three steps complement each other and constitute a dynamic closed-loop system. Online detection provides real-time material data as the basis for preliminary judgment; the historical database provides empirical feedback for model correction; temperature compensation ensures the real-time adaptability of the model. This significantly improves the safety of the stranding process, reduces the risk of wire breakage, and at the same time enhances the adaptability to new materials and complex working conditions, and improves the intelligent and automated control level of the system.

[0090] In a possible implementation, first, a multi-parameter joint abnormality evaluation model is established. The core of this model is to combine multiple parameters reflecting the state of materials and equipment. These parameters include:

[0091] The time derivative of the pressure distribution mutation rate: This parameter reflects the rate of pressure change during the stranding process. An increase in the mutation rate usually indicates the existence of equipment failures or abnormal material properties.

[0092] The spatial gradient of the energy migration in the response frequency domain: Through frequency domain energy analysis, the changing trend of strain is captured. The spatial gradient reflects the strain non-uniformity in different parts of the material and may be a sign of improper equipment adjustment or material property problems.

[0093] The statistical outlier of the balance index: The balance refers to the tension balance of each wire during the stranding process. The appearance of outliers indicates the existence of uneven stress distribution, which may lead to wire breakage or other operation abnormalities.

[0094] The joint evaluation of these parameters provides a comprehensive evaluation of the equipment and process status, which helps to accurately identify potential abnormalities in the system operation.

[0095] After the abnormality evaluation is completed, the system dynamically divides different response levels according to the evaluation results. Specifically, when the abnormality value is high, the system will judge a higher abnormality risk, automatically adjust the operation strategy, and execute more stringent control measures. The threshold of the response level will be adaptively adjusted according to the cumulative operation time of the equipment, which means that as the equipment usage time increases, the system will adjust the response strategy according to the aging and wear degree of the equipment. For example, the aging of the equipment may increase the probability of failures. At this time, the system will give a higher response level to a lower abnormality degree to prevent accidents.

[0096] In the case of a relatively high abnormal response level, the system will enter the high-level response stage, and at this time, the die reverse rotation control is introduced. The die reverse rotation control eliminates the residual stress by changing the rotation direction of the die. This process helps to improve the material quality and the physical properties of the wire during the stranding process. The number of rotation cycles of this control process is positively correlated with the duration of the abnormality. That is to say, the longer the duration of the abnormality, the more cycles of die reverse rotation, so as to ensure that the residual stress is effectively removed through a longer reverse rotation and avoid wire breakage or other quality problems caused by stress concentration.

[0097] These three steps work highly collaboratively to form a dynamic and intelligent response mechanism. First, by establishing a multi-parameter joint abnormality evaluation model, the system can monitor in real time and accurately identify potential problems in the process. Second, based on these evaluation results, the system adaptively adjusts the response level to ensure that the equipment is always in the best control state. Finally, when a high-level abnormality occurs, the step of introducing die reverse rotation control is added, which further improves the emergency response ability of the system, eliminates the residual stress, and guarantees the quality of the cable laying. Generally speaking, this hierarchical control strategy not only improves the stability and reliability of the system, but also enables the equipment to make flexible adjustments under different working conditions, thereby improving the production efficiency and reducing the occurrence of equipment failures and quality problems.

[0098] In a possible implementation manner, the first step of the self-learning optimization process is to collect complete process cycle data before and after the occurrence of an abnormal event. These data cover multiple dimensions such as equipment status, process parameters, and environmental factors. For example, process parameters such as the working temperature, humidity, stranding tension, and speed of the equipment will be recorded and compared with the parameters at the time of the abnormality. Through this step, the subtle changes in the process and equipment status at the time of the abnormality can be comprehensively captured, forming a multi-dimensional event feature vector.

[0099] After the data is collected, the system will construct a multi-dimensional event feature vector that includes equipment status, process parameters, environmental factors, etc. These vectors reflect various factors at the time of the abnormality and can provide rich information for subsequent analysis. By integrating these multi-dimensional data to form a comprehensive event feature, it is possible to more accurately identify which factors are closely related to the occurrence of the abnormal event.

[0100] Subsequently, the system will use a feature importance evaluation algorithm to evaluate the influence degree of each factor on the abnormal event. This algorithm will analyze all feature vectors, evaluate the role of each parameter in the occurrence of the abnormality, and dynamically adjust the weight of the algorithm according to the event type. In other words, different types of abnormal events may have different key influencing factors. For example, some abnormalities may be related to temperature changes, while other abnormalities are closely related to tension fluctuations or speed changes.

[0101] After extracting the key influencing factors, the system updates the control model parameters using an incremental learning mechanism. During this process, the system compares the new dataset and features with historical data, retains the effective feature association patterns, and gradually optimizes the accuracy of the control model. The advantage of incremental learning is that it can update the model parameters in real time, avoiding the high cost of retraining the entire model in traditional machine learning methods, while ensuring the continuous optimization of the model.

[0102] Finally, during the processing of similar products, the system preferentially loads the optimized control parameter set. Through continuous accumulation of past processing experience, the system can quickly adjust the control parameters according to the historical optimization results during new processing, thereby improving efficiency and quality. The optimized parameter set will help the cabling machine quickly reach the optimal working state under similar working conditions, thus enhancing product consistency and production efficiency.

[0103] The five steps of the self-learning optimization process are interrelated. Through data collection, feature extraction, evaluation, and incremental learning, the system can continuously update and optimize the control strategy. The greatest advantage of this optimization process lies in its dynamic adaptability. It can automatically adjust the control model as the process and equipment change, and repeatedly verify and adjust the parameters during the processing, thereby improving the adaptive ability of the production process and reducing the probability of anomalies. By preferentially loading the optimized parameter set, the system can quickly return to the optimal working state in subsequent production, avoiding repeated debugging and experiments, and greatly improving production efficiency and product quality consistency.

[0104] In a possible implementation, first, the principal component analysis (PCA) is performed on the collected multi-dimensional event features. Principal component analysis is a dimensionality reduction technique that maps the original feature set to a new feature space through a linear transformation, where the new features (principal components) have the maximum variance and the strongest representativeness. The system determines which principal components can retain the core information of the data to the greatest extent by analyzing the explanatory power of each principal component, that is, the ability of the principal component to explain the variation of the original data. Only the principal components with an explanatory power exceeding the preset requirements will be retained, while the principal components with low explanatory power will be discarded, which can effectively reduce redundant data and improve the calculation efficiency.

[0105] After principal component decomposition, the system will construct a deep feature extraction network to further explore potential feature correlation patterns. Deep neural networks, such as convolutional neural networks or fully connected networks, can automatically extract and learn complex patterns in input features through multi-level non-linear transformations. Here, the structure of the network is dynamically adjusted according to the feature dimensions. For example, for high-dimensional features, the network may be designed with deeper layers to capture more complex relationships; while for low-dimensional features, the network may adopt a shallower structure to avoid overfitting. The flexibility of the deep network enables it to adapt to different feature sets and effectively discover complex patterns hidden in the data.

[0106] Next, the system will introduce an attention mechanism to dynamically assign feature weights. The attention mechanism can automatically focus on those feature dimensions that are highly relevant to the event type according to the type of the current event and the actual situation being processed. For example, in some abnormal events, temperature changes may be more critical than tension changes, while in other cases, fluctuations in tension may be the main influencing factor. Through the attention mechanism, the system can adjust the weights of each feature in real time to ensure that the features having the greatest impact on the current event are focused on during the analysis process.

[0107] These steps form a close synergy in the feature importance evaluation algorithm. Principal component decomposition retains the most explanatory features through dimensionality reduction, reducing the computational burden; the deep feature extraction network further explores the complex correlations between features, enhancing the system's ability to identify potential patterns; while the attention mechanism ensures that the system can dynamically adjust the feature weights when each event occurs, thereby improving the accuracy and reliability of the evaluation results. Through the cooperation of these three steps, the system can more accurately identify the key factors affecting abnormal events, providing more refined support for subsequent adaptive control and optimization.

[0108] The implementation of these steps helps to improve the system's processing ability for multi-dimensional data. Especially when facing complex working conditions and changing environments, it can adaptively adjust and optimize control strategies, thereby reducing the probability of anomalies and improving the stability and reliability of production.

[0109] This invention covers any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To enable the public to have a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments of this invention. However, those skilled in the art can fully understand this invention even without the description of these details. Additionally, well-known methods, processes, procedures, components, and circuits, etc. are not described in detail to avoid unnecessary confusion to the essence of this invention.

[0110] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. An adaptive stranding control method for a cabling machine based on multi-dimensional tension feedback, characterized in that It includes the following steps: Step 1: Configure a phase marking device on the stranding spindle to generate a periodic reference signal, and set a composite tension detection structure on each wire transmission path. This structure includes circumferentially distributed pressure sensing units and axial strain sensing units. Under the trigger of the reference signal, synchronously obtain the radial pressure distribution characteristics, axial strain evolution characteristics of each wire, and the dynamic phase relationship between the two; Step 2: Conduct multi-wire dynamic coupling analysis on the pressure distribution characteristics, extract the intensity of tension interaction between wires through time-shift correlation calculation, and generate a tension balance degree index by combining the frequency-domain migration law and phase relationship of the strain evolution characteristics; Step 3: Construct a multi-dimensional feature space including real-time process states and historical fault modes, and use an adaptive weighting mechanism to fuse the tension balance degree index with the multi-dimensional features to generate a stranding stability coefficient that reflects both instantaneous fluctuations and trend deterioration; Step 4: Establish a dynamic lay length prediction model based on the trend component of the stability coefficient, generate compensation parameters by combining the material deformation characteristics, and at the same time implement micro-step compensation based on the fluctuation component in a specific rotation phase interval; Step 5: Real-time monitor the pressure distribution mutation rate, strain frequency-domain migration rate, and the abnormality degree of the balance degree index. When the three cooperate to exceed the dynamic threshold, execute a hierarchical control strategy, including progressive speed reduction, pre-tension balance, and safety shutdown sequence.

2. The adaptive stranding control method for a cabling machine based on multi-dimensional tension feedback according to claim 1, wherein: The configuration method of the composite tension detection structure includes: Dynamically determine the density of circumferential detection units according to the proportional relationship between the geometric characteristics of the stranding die and the physical properties of the wires, so that the overlapping monitoring areas formed by the coverage areas of adjacent detection units form a preset proportion; Dynamically adjust the spacing between detection unit groups based on real-time stranding process parameters, where the spacing adjustment amount depends on the friction characteristics feedback by the real-time surface state of the wires; The implantation depth of the strain sensing unit is dynamically adjusted according to the wire diameter and the material bending stiffness characteristics, and the bending stiffness characteristics are extracted from the pre-stored material characteristics library.

3. The adaptive stranding control method for a cabling machine based on multi-dimensional tension feedback according to claim 2, wherein: The acquisition process of the real-time stranding process parameters includes: Jointly calculate the actual stranding lay length through the spindle rotation phase signal and the wire displacement detection device; Dynamically correct the friction characteristic parameters according to the analysis results of the surface optical characteristics of the wires; Periodically update the reference parameter values, and the update trigger condition is associated with the change rate of the stranding stability coefficient.

4. The adaptive stranding control method for a cabling machine based on multi-dimensional tension feedback according to claim 1, characterized in that: The specific implementation of the dynamic coupling analysis includes: Perform normalization preprocessing on the multi-wire pressure distribution data to eliminate magnitude differences; Establish a time-shift analysis window matching the current production process rhythm, and the window width is dynamically adjusted according to the spindle speed; Calculate the correlation intensity of the pressure fluctuation sequences of each wire at different time-shift amounts, and extract the key indicators characterizing the tension coupling degree; Perform dynamic smoothing processing on the key indicators, and the smoothing intensity is negatively correlated with the wire feeding speed.

5. The adaptive stranding control method for a cabling machine based on multi-dimensional tension feedback according to claim 1, wherein: The implementation process of the adaptive weighting mechanism includes: Construct a multi-dimensional influence factor set including equipment operating status, environmental conditions, and material characteristics; Perform dynamic normalization processing on each influence factor, and the normalization interval is automatically adjusted according to the distribution characteristics of on-line monitoring data; Generate dynamic weight coefficients for each factor through a fuzzy inference system, and the inference rule base is continuously optimized according to historical process data; Non-linearly fuse the weight coefficient with the tension balance degree index, and the form of the fusion function is dynamically switched according to the plastic deformation stage of the material.

6. The adaptive stranding control method for a cabling machine based on multi-dimensional tension feedback according to claim 1, wherein: The method for constructing the pitch dynamic prediction model includes: When the trend component exceeds the material deformation safety boundary, generate a pitch adjustment curve based on the real-time linear velocity and the material creep characteristics; Adopt an inertial compensation algorithm to smoothly execute the pitch adjustment, and the compensation intensity is dynamically optimized according to the rotational inertia of the equipment; During the pitch adjustment process, synchronously correct the spindle speed, and the correction amount maintains a dynamic balance relationship with the pitch change rate.

7. The adaptive stranding control method for a cabling machine based on multi-dimensional tension feedback according to claim 6, characterized in that: The determination process of the material deformation safety boundary includes: Obtain the stress relaxation characteristics of the current wire through an on-line material property analysis device; Combine the wire break accident data under similar working conditions in the historical process database for boundary correction; Perform real-time compensation on the safety boundary according to the environmental temperature fluctuation.

8. The adaptive stranding control method for a cabling machine based on multi-dimensional tension feedback according to claim 1, wherein: The implementation of the hierarchical control strategy includes: Establish a multi-parameter joint abnormality evaluation model, which fuses the time derivative of the pressure distribution mutation rate, the spatial gradient of the strain frequency domain energy migration, and the statistical outliers of the balance degree index; Dynamically divide the response levels according to the abnormality evaluation results, and the response level thresholds are adaptively adjusted according to the cumulative operation time of the equipment; Introduce die reverse rotation control in the high-level response stage to eliminate the residual stress, and the number of rotation cycles is positively correlated with the duration of the abnormality.

9. The adaptive stranding control method for a cabling machine based on multi-dimensional tension feedback according to claim 1, characterized in that: It also includes: Execute a self-learning optimization process after the abnormal event is processed, including: Collect the complete process cycle data before and after the abnormality occurs, and construct a multi-dimensional event feature vector including equipment status, process parameters, and environmental factors; Adopt a feature importance evaluation algorithm to extract key influencing factors, and the weights of the evaluation algorithm are dynamically allocated according to the event type; Update the control model parameters through an incremental learning mechanism, and retain the effective feature association patterns; Preferentially load the optimized parameter set when processing similar products.

10. The adaptive stranding control method for a cabling machine based on multi-dimensional tension feedback according to claim 9, wherein: The implementation of the feature importance evaluation algorithm includes: Perform principal component decomposition on the multi-dimensional event features, and retain the principal components with an interpretation degree exceeding the preset requirements; Construct a deep feature extraction network to mine the implicit association rules, and the network structure is dynamically adjusted according to the feature dimensions; Dynamically allocate feature weights through an attention mechanism, and focus on the feature dimensions strongly associated with the current event type.

Images