A digital comber equipment management method and system based on data analysis

Abstract

The present invention relates to the technical field of equipment management, and specifically to a digital comber equipment management method and system based on data analysis, including: constructing a quality-stability joint prediction model based on historical production data to evaluate the combination of fiber attributes and process parameters, and predicting the expected processing quality and expected equipment stability; combining production plan information, aiming at the switching of different fiber batches, with the management objectives of optimizing processing quality, equipment stability and equipment utilization rate, calculating the recommended process parameter combination and the recommended parameter switching path; collecting the actual processing quality and actual equipment stability, conducting management evaluation in combination with the quality-stability joint prediction model, and iteratively updating the model; analyzing the residual sequence of the actual equipment stability and the expected equipment stability, and generating predictive maintenance suggestions in combination with historical equipment maintenance data. The equipment management method of the present invention ensures that quality, efficiency and stability are taken into account during the alternate processing of different fibers.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment management, and particularly to a digital comber equipment management method and system based on data analysis. Background Art

[0002] With the rapid growth of the demand for differentiated fiber products in the textile market, the processing scenarios of combers show significant differentiation characteristics: high-value fibers, due to their high raw material costs and low error tolerance, require precise process parameter configurations to ensure the quality of finished products, but their production batches are usually small; while ordinary fibers, although the processing accuracy requirements are relatively loose, require the equipment to be in a high-yield state for a long time to meet the demand for large-scale orders. In this context, when alternating the processing of the two types of fibers on the same production line, how to ensure the process accuracy of high-value fibers, maintain the high efficiency of ordinary fibers, and avoid the decline in stability caused by frequent switching of working conditions has become the key contradiction restricting the improvement of the comprehensive efficiency of combers.

[0003] Therefore, a digital comber equipment management method and system based on data analysis are proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide a digital comber equipment management method and system based on data analysis to ensure the balance of quality, efficiency and stability during the alternating processing of differentiated fibers. It includes: constructing a quality-stability joint prediction model based on historical production data to evaluate fiber attributes and process parameter combinations, and predicting the expected processing quality and expected equipment stability; combining production plan information, aiming at the switching of different fiber batches, with the management objectives of optimizing processing quality, equipment stability and equipment utilization rate, calculating the recommended process parameter combinations and recommended parameter switching paths; collecting the actual processing quality and actual equipment stability, combining with the quality-stability joint prediction model for management evaluation, and iteratively updating the model; analyzing the residual sequence of the actual equipment stability and the expected equipment stability, and generating predictive maintenance suggestions in combination with historical equipment maintenance data.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A digital comber equipment management method based on data analysis, including.

[0007] Obtain production plan information, fiber attributes, historical production data and historical equipment maintenance data;

[0008] Based on historical production data, use machine learning algorithms to construct a quality-stability joint prediction model, receive fiber attributes and corresponding process parameter combinations, and predict the corresponding expected processing quality and expected equipment stability;

[0009] Combined with the production plan information, use the quality-stability joint prediction model to predict and evaluate the candidate process parameter combinations. For the switching of different fiber batches, with the management objectives of optimizing processing quality, equipment stability, and equipment utilization rate, calculate the recommended process parameter combinations and the recommended parameter switching paths;

[0010] Collect the actual processing quality and actual equipment stability during equipment operation, combine them with the quality-stability joint prediction model for management evaluation, and iteratively update the quality-stability joint prediction model;

[0011] Analyze the residual sequence of the actual equipment stability and the expected equipment stability, and combine the historical equipment maintenance data to generate predictive maintenance suggestions.

[0012] Preferably, LightGBM is used to construct the quality-stability joint prediction model, and a multi-objective regression strategy is adopted. A quality-stability joint prediction sub-model is trained for each index of the expected processing quality and the expected equipment stability respectively; the fiber attributes to be processed and the corresponding process parameter combinations are used to form an input feature vector, and the input feature vector is input into the trained quality-stability joint prediction sub-model to obtain the corresponding index prediction value; the index prediction values output by all quality-stability joint prediction sub-models are integrated to obtain the expected processing quality and the expected equipment stability.

[0013] Preferably, the prediction process of the quality-stability joint prediction sub-model includes: S1. Process the input feature vector through the first decision tree to obtain a preliminary index prediction value; S2. Process the input feature vector through the second decision tree to obtain a residual prediction value, and add the preliminary index prediction value and the residual prediction value to obtain the updated preliminary index prediction value; S3. Repeat step S2 to sequentially traverse the remaining decision trees, and after the traversal is completed, obtain the index prediction value.

[0014] Preferably, the steps of calculating the recommended process parameter combinations and the recommended parameter switching paths specifically include:

[0015] Define the decision variables as the process parameter combinations and parameter adjustment paths of the new batch of fibers, and construct an objective function with the goals of maximizing the expected processing quality, minimizing the impact of the switching process on equipment stability, and minimizing the switching time;

[0016] Set the parameter physical constraints, stability constraints, and path smoothness constraints; select the Bayesian optimization algorithm, and guide the search process by constructing a surrogate model of the objective function to iteratively generate candidate process parameter combinations and candidate parameter switching paths;

[0017] Evaluate the objective function value corresponding to the candidate process parameter combination in combination with the quality-stability joint prediction model. When the stopping condition is met, output the recommended process parameter combination that optimizes the objective function under the constraint conditions and the corresponding recommended parameter switching path.

[0018] Preferably, the step of iteratively updating the quality-stability joint prediction model specifically includes: calculating management evaluation indicators according to the expected processing quality and expected equipment stability, including prediction error, target quality achievement rate, stability compliance rate, and improvement of overall equipment efficiency; setting model update trigger conditions according to the management evaluation indicators. When the conditions are met, adopt a periodic batch retraining strategy, and use the newly obtained actual production data combined with part of the historical production data to retrain the quality-stability joint prediction sub-model that needs to be updated.

[0019] Preferably, the step of generating predictive maintenance suggestions specifically includes:

[0020] Calculate the residual sequence between the actual equipment stability and the expected equipment stability in real time; apply time series analysis technology to the residual sequence to extract trend feature vectors;

[0021] Use historical equipment maintenance data and the corresponding historical residual sequences to train a fault mode classifier for outputting the predicted fault type according to the current trend feature vectors;

[0022] Set predictive maintenance trigger rules based on the confidence level of the predicted fault type and the residual amplitude. When the predictive maintenance trigger rules are met, automatically generate predictive maintenance suggestions.

[0023] A digital comber equipment management system based on data analysis, comprising:

[0024] A data acquisition unit that acquires production plan information, fiber properties, historical production data, and historical equipment maintenance data;

[0025] A joint prediction unit that constructs a quality-stability joint prediction model based on historical production data using machine learning algorithms, receives fiber properties and corresponding process parameter combinations, and predicts the corresponding expected processing quality and expected equipment stability;

[0026] A parameter switching unit that, in combination with production plan information, uses the quality-stability joint prediction model to predict and evaluate candidate process parameter combinations, and aims at optimizing processing quality, equipment stability, and equipment utilization rate for the switching of different fiber batches, and calculates the recommended process parameter combination and the recommended parameter switching path;

[0027] An evaluation and update unit that collects the actual processing quality and actual equipment stability during equipment operation, conducts management evaluation in combination with the quality-stability joint prediction model, and iteratively updates the quality-stability joint prediction model;

[0028] A predictive maintenance unit that analyzes the residual sequence between the actual equipment stability and the expected equipment stability, and generates predictive maintenance suggestions in combination with historical equipment maintenance data.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0030] 1. By using machine learning algorithms to construct a quality-stability joint prediction model, it is possible to simultaneously predict processing quality and equipment stability based on fiber properties and process parameter combinations. The model is implemented using the LightGBM framework and adopts a multi-objective regression strategy. Sub-models are trained separately for each prediction index to form an integrated model system. The joint evaluation mechanism of the model enables the production line to consider both product quality and equipment operating status when adjusting parameters, and can pre-insight the comprehensive impact of different parameter settings on quality and stability. In addition, the model supports regular updates and iterative learning, can continuously absorb new production experience, and improve prediction accuracy. The model provides a decision-making basis for achieving the balance of quality and stability during the processing switch of different fiber batches in the comber, and solves the problem of conflicts between quality assurance and equipment stability during the alternating processing of high-value fibers and ordinary fibers.

[0031] 2. By comprehensively considering the three objectives of maximizing processing quality, minimizing the impact on equipment stability, and minimizing the switching time, a target function is constructed, and the optimal process parameter combination and parameter switching path are found under the constraint conditions. The Bayesian optimization algorithm is used to effectively handle the search problem in the high-dimensional parameter space. By iteratively generating candidate process parameter combinations and candidate parameter switching paths, and evaluating them in combination with the quality-stability joint prediction model, the optimal recommended solution is finally output. It realizes the intelligent parameter configuration and smooth transition of the comber during the fiber switching of different batches, and solves the problems of decreased equipment stability and efficiency loss caused by the alternating processing of high-value fibers and ordinary fibers.

[0032] 3. Predictive maintenance method based on residual analysis. By analyzing the residual sequence between the actual equipment stability and the expected equipment stability, combined with time series analysis technology to extract trend feature vectors, and then predict the possible fault types. This method uses historical equipment maintenance data to train a fault mode classifier, which can predict potential faults according to the current trend feature vectors, and set trigger rules based on the confidence level of the predicted fault types and the residual amplitude. When the conditions are met, predictive maintenance suggestions are automatically generated. By detecting abnormal trends in advance, maintenance can be arranged at an appropriate time in the production plan, avoiding production losses caused by sudden shutdowns and also avoiding waste of resources caused by over-maintenance. It provides guarantee for the long-term stable operation of the comber and ensures the continuous and efficient operation of the production line when processing different fibers alternately. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a schematic flow chart of a digital comber equipment management method based on data analysis according to the present invention;

[0034] Figure 2 It is a schematic flow chart of generating predictive maintenance suggestions according to the present invention;

[0035] Figure 3 It is a schematic structural diagram of a digital comber equipment management system based on data analysis according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0036] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0037] Please refer to Figures 1 to 3 , the present invention provides a digital comber equipment management method and system based on data analysis, and the technical solutions are as follows:

[0038] Embodiment 1:

[0039] This embodiment is applied to the actual production requirements of the comber workshop of a textile enterprise. The workshop is equipped with 6 combers, which are responsible for processing high-value Egyptian long-staple cotton and ordinary Xinjiang cotton. Due to market demand fluctuations, the two types of fibers need to be processed alternately on the same production line, resulting in frequent switching of process parameters. To improve the comprehensive efficiency of comber equipment management, this embodiment applies a digital comber equipment management method based on data analysis, as Figure 1 shown, including:

[0040] Obtain production plan information, fiber properties, historical production data, and historical equipment maintenance data;

[0041] Based on historical production data, use machine learning algorithms to construct a quality-stability joint prediction model, receive fiber properties and corresponding process parameter combinations, and predict the corresponding expected processing quality and expected equipment stability;

[0042] Combined with production plan information, use the quality-stability joint prediction model to predict and evaluate candidate process parameter combinations. For the switching of different fiber batches, with the management goals of optimizing processing quality, equipment stability, and equipment utilization rate, calculate the recommended process parameter combinations and recommended parameter switching paths;

[0043] Collect the actual processing quality and actual equipment stability during equipment operation, conduct management evaluation in combination with the quality-stability joint prediction model, and iteratively update the quality-stability joint prediction model;

[0044] Analyze the residual sequence between the actual equipment stability and the expected equipment stability, and generate predictive maintenance suggestions in combination with historical equipment maintenance data.

[0045] In this embodiment, the production plan information is sourced from the enterprise resource planning system, including the fiber batch number, fiber type, planned output, planned start / end time, and target quality range for planned processing; the fiber properties are sourced from the quality management system, including detailed physical detection indicators for each fiber batch to be processed, such as average length, length dispersion, fineness, impurity content, oil content, moisture regain, and short fiber rate; the historical production data includes the fiber properties of historical processing batches, the historical process parameter combinations adopted, the historical actual processing quality of the corresponding final products, and the historical actual equipment stability. The historical equipment maintenance data is sourced from the equipment maintenance log, including failure time, failure phenomenon, failed components, maintenance content, maintenance duration, and replacement spare part information; the process parameter combinations include cylinder speed, feed roller speed, separating roller speed, key part gauge, and total draft ratio; the processing quality includes multiple indicators: evenness, number of neps, number of impurities, and short fiber content; the equipment stability includes multiple indicators: main shaft vibration amplitude, motor load rate, and energy consumption index.

[0046] Furthermore, use LightGBM to construct the quality-stability joint prediction model, adopt a multi-objective regression strategy, and train a quality-stability joint prediction sub-model for each indicator of the expected processing quality and expected equipment stability respectively; construct an input feature vector from the fiber properties to be processed and the corresponding process parameter combinations, input the input feature vector into the trained quality-stability joint prediction sub-model to obtain the corresponding indicator prediction values; integrate the indicator prediction values output by all quality-stability joint prediction sub-models to obtain the expected processing quality and expected equipment stability.

[0047] A quality-stability joint prediction model is constructed using LightGBM (Light Gradient Boosting Machine), combined with a multi-objective regression strategy, ensuring that the prediction model can efficiently process high-dimensional and non-linear data, accurately capture the complex relationships between fiber properties, process parameters, and multiple quality and stability indicators, improve the accuracy and reliability of predictions, and provide reliable decision-making support for parameter optimization of different fiber batches.

[0048] Furthermore, the prediction process of the quality-stability joint prediction sub-model includes: S1. Processing the input feature vector through a first decision tree to obtain a preliminary index prediction value; S2. Processing the input feature vector through a second decision tree to obtain a residual prediction value, and adding the preliminary index prediction value and the residual prediction value to obtain the updated preliminary index prediction value; S3. Repeating step S2 to sequentially traverse the remaining decision trees, and obtaining the index prediction value after the traversal ends.

[0049] Through the residual learning mechanism of the gradient boosting decision tree, the prediction model can gradually learn and correct prediction errors, continuously improve the prediction accuracy, effectively capture the complex interaction relationships between fiber properties and process parameters, and improve the model's ability to identify marginal conditions.

[0050] Table 1 shows the performance evaluation results of the quality-stability joint prediction sub-model corresponding to each index on the training set and the test set. Based on the collected historical production data, more than 5,000 valid samples were extracted and divided into a training set and a test set. Each sample includes fiber properties, process parameter combinations, actual processing quality, and actual equipment stability. As can be seen from Table 1, the sub-models of all indicators show good prediction capabilities, indicating that the model can accurately capture the influence relationships of fiber properties and process parameters on processing quality and equipment stability.

[0051] Table 1 Quality-Stability Joint Prediction Sub-Model Evaluation Results

[0052]

[0053] Furthermore, the steps for calculating the recommended process parameter combination and the recommended parameter switching path specifically include:

[0054] Define the decision variables as the process parameter combination and the parameter adjustment path of the new batch of fibers, and construct an objective function with the goal of maximizing the expected processing quality, minimizing the impact of the switching process on equipment stability, and minimizing the switching time;

[0055] Set physical constraints, stability constraints, and path smoothness constraints for the parameters; select the Bayesian optimization algorithm, and guide the search process by constructing a surrogate model of the objective function, iteratively generating candidate process parameter combinations and candidate parameter switching paths;

[0056] Combine the quality-stability joint prediction model to evaluate the objective function value corresponding to the candidate process parameter combination. When the stopping condition is met, output the recommended process parameter combination that optimizes the objective function under the constraint conditions and the corresponding recommended parameter switching path.

[0057] The process parameter recommendation mechanism based on Bayesian optimization comprehensively considers processing quality, equipment stability, and switching efficiency. Guided by the surrogate model in the search process, it greatly improves the efficiency of parameter optimization. The setting of multiple constraint conditions ensures the practicability and safety of the recommended parameter scheme, while the smooth path planning reduces the equipment impact during the parameter switching process, effectively solves the problem of process parameter switching when processing high-value fibers and ordinary fibers alternately, and improves the adaptability and flexibility of the production line.

[0058] Table 2 gives an example of the batch switching process parameter combination and the optimization result of the parameter switching path from Egyptian long-staple cotton to Xinjiang cotton:

[0059] Table 2 Example of process parameter combination and optimization result of parameter switching path

[0060]

[0061] The parameter switching path divides the switching process into a starting state, a first stage, a second stage, and a final state, and gradually reaches each parameter value in the recommended process parameter combination through different time steps. The time steps of the recommended parameter switching path in the example in Table 2 are [0, 10min, 20min, 30min, 35min]. The parameter changes in each stage are within the acceptable range of the equipment, avoiding equipment impact caused by sudden changes and achieving a smooth transition.

[0062] Furthermore, maximizing the expected processing quality means predicting the values of various indicators of the processing quality when processing a new batch of fibers with the candidate process parameter combination by calling the quality-stability joint prediction model. The goal is to make these indicators as close to or better than the target quality range in the production plan information. Deviating from the target quality range will result in a negative evaluation;

[0063] Minimizing the impact of the switching process on equipment stability means using a combined quality-stability prediction model to evaluate the degree of fluctuation of various index values of equipment stability during the process of adjusting from the current process parameter combination to the candidate process parameter combination according to the candidate parameter switching path. The goal is to make these degrees of fluctuation as small as possible and avoid exceeding the safe fluctuation threshold. Exceeding the safe fluctuation threshold will result in a significant negative evaluation.

[0064] Minimizing the switching time means that the total time consumed by the parameter switching process should be as short as possible to improve equipment utilization. The longer the time, the lower the evaluation.

[0065] Set different weights according to the focus of production management, and weighted sum the evaluation results of these three aspects to form the objective function value of the process parameter combination and the corresponding parameter adjustment path. The goal of the optimization algorithm is to find the optimal solution that makes the objective function value optimal.

[0066] Parameter physical constraints mean that each parameter value in the recommended process parameter combination must be within the physical adjustment range allowed by the equipment.

[0067] Stability constraints mean that at any moment during the entire parameter switching path, the expected equipment stability shall not exceed the preset safe operation threshold.

[0068] Path smoothness constraints mean that the speed of parameter adjustment cannot exceed the maximum change rate allowed by the equipment execution to ensure the smoothness of the adjustment process.

[0069] The Bayesian optimization iteration process includes: The Bayesian optimization algorithm first calls the prediction model according to a few candidate solutions (candidate process parameter combinations and candidate parameter switching paths) to obtain the true objective function score, and then uses the Gaussian process model to construct a surrogate model of the objective function. In each iteration, the surrogate model is used to quickly predict the potential objective function values of the unevaluated candidate solutions, and a acquisition function is used to select the currently most worthy candidate solution for real evaluation from the unevaluated candidate solutions. Call the combined quality-stability prediction model to calculate the true objective function value of the selected candidate solution, add the new true objective function value to the known data, and update the surrogate model to make it approximate the true objective function more accurately; repeat this process until the preset stop condition is reached. The final output is the recommended process parameter combination and the corresponding recommended parameter switching path that make the objective function score optimal under all constraint conditions. The stop condition can be set as the maximum number of iterations, the calculation time limit, or the objective function value no longer improves significantly.

[0070] Further, the steps of iteratively updating the quality-stability joint prediction model specifically include: calculating management evaluation indicators based on the expected processing quality and expected equipment stability, including prediction error, target quality achievement rate, stability compliance rate, and improvement of overall equipment effectiveness (OEE); setting model update trigger conditions according to the management evaluation indicators, and when the conditions are met, adopting a periodic batch retraining strategy, merging the newly obtained actual production data with part of the historical production data to retrain the quality-stability joint prediction sub-model that needs to be updated.

[0071] By setting model update trigger conditions and adopting a periodic batch retraining strategy, continuous optimization of the joint prediction model is achieved. This iterative update mechanism can adapt to long-term influencing factors such as equipment aging and seasonal changes, maintaining the timeliness and accuracy of the model. The design of the management evaluation indicators directly reflects the application effect of the method, provides a quantitative basis for continuous improvement, ensures that the comber management system can continuously self-improve with production practice, and improves the overall management level.

[0072] Further, the calculation method of the prediction error is: by comparing the index values of the expected processing quality and expected equipment stability predicted by the model in real time during actual production with the corresponding index values actually collected, calculating the mean absolute error between the two, and the smaller the error, the more accurate the prediction;

[0073] The calculation method of the target quality achievement rate is: counting the proportion of the quantity within the range from the actual processing quality to the target quality after producing with the recommended process parameters, and the higher the proportion, the better the effectiveness of the method in ensuring the target quality;

[0074] The calculation method of the stability compliance rate is: counting the proportion of the time when the actual equipment stability does not exceed the preset safe operation threshold during actual production operation, and the higher the proportion, the better the effect of the method in maintaining equipment stability;

[0075] The improvement of overall equipment effectiveness (OEE) is obtained by comparing the change in the overall equipment effectiveness of the comber before and after applying this method; the overall equipment effectiveness takes into account the time utilization rate, performance utilization rate, and quality qualification rate of the equipment.

[0076] To ensure the timeliness and accuracy of the model, reasonable trigger conditions need to be set to initiate the iterative update of the model. The model update trigger conditions can be set as a combination of one or more of the following:

[0077] Based on time period: Set a fixed update period, for example, perform model update once a week, every two weeks, or every month;

[0078] Based on data volume accumulation: Trigger an update when the amount of newly collected valid production data samples reaches a certain threshold;

[0079] Based on model performance degradation: Continuously monitor the prediction error. When the prediction error corresponding to the processing quality index exceeds the preset threshold for a continuous period of time, or the accuracy is significantly lower than the historical level, it indicates that the model performance may have degraded and an update needs to be triggered.

[0080] Based on significant changes in working conditions or environment: When it is detected that there are major changes in the production environment, such as equipment renovation, long-term deviation of environmental temperature and humidity from the normal range, etc., which may cause the original model to fail, actively trigger model update.

[0081] Furthermore, to calculate the residual sequence in the management evaluation index and predictive maintenance, during the normal production operation of the comber equipment, continuously use the quality-stability joint prediction model for real-time prediction, specifically including: obtaining the combination of process parameters of the equipment actually running at the current moment and the attribute data of the fiber being processed currently, to obtain real-time fiber attributes and real-time process parameter combinations; using the real-time fiber attributes and real-time process parameter combinations as inputs, calling the quality-stability joint prediction model, and outputting the expected processing quality and expected equipment stability at the current moment; subtracting the index value predicted by the model from the actual index value at the same moment to obtain the residual value at the corresponding moment.

[0082] Furthermore, the flow chart for generating predictive maintenance suggestions is shown in Figure 2 , and the steps specifically include:

[0083] Real-time calculate the residual sequence between the actual equipment stability and the expected equipment stability; apply time series analysis techniques to the residual sequence to extract trend feature vectors; time series analysis techniques can be moving average, exponential smoothing or Fourier transform, etc.;

[0084] Use historical equipment maintenance data and the corresponding historical residual sequences to train a fault mode classifier for outputting the predicted fault type according to the current trend feature vector; optional models of the fault mode classifier include support vector machine, random forest or recurrent neural network;

[0085] Set predictive maintenance trigger rules based on the confidence level of the predicted fault type and the residual amplitude. When the predictive maintenance trigger rules are met, automatically generate predictive maintenance suggestions.

[0086] The predictive maintenance method based on residual analysis realizes the early detection of potential problems of the equipment through time series analysis and fault mode classification. Reduces the risk of sudden failures and unplanned downtime. The automatically generated maintenance suggestions are based on data analysis rather than empirical judgment, improving the scientificity and pertinence of maintenance decisions, and providing a strong guarantee for the stable operation of the comber under frequent working condition switching.

[0087] The rules for triggering predictive maintenance need to comprehensively consider the likelihood and severity of failures. The specific setting methods for triggering rules of predictive maintenance based on confidence level and residual amplitude include:

[0088] Based on confidence level: When the predicted confidence level of the fault type output by the fault mode classifier exceeds the confidence level threshold set for this fault type, predictive maintenance is directly triggered; different confidence level thresholds can be set for different fault types;

[0089] Based on residual amplitude: When the residual value corresponding to the index in the equipment stability exceeds the set warning threshold for a period of time, predictive maintenance is triggered; the warning threshold can be set according to historical production data analysis and expert experience;

[0090] When any of the above predictive maintenance triggering rules is satisfied, predictive maintenance suggestions are automatically generated through the following process:

[0091] Obtain the predictive maintenance trigger time as the early warning timestamp; determine the most likely faulty component according to the predicted fault type with the highest output confidence of the fault mode classifier; retrieve the matching, standardized inspection steps, diagnostic methods, and recommended repair operation lists from the predefined maintenance knowledge base according to the predicted fault type; determine the urgency of the fault based on the predicted fault type, confidence level, and residual amplitude, and combine the production plan information to determine the available downtime information, and recommend specific maintenance execution time suggestions.

[0092] Table 3 shows the effect comparison before and after implementing the method of this embodiment. It can be seen that all key indicators have been significantly improved. In particular, the first-class product rate of high-value fibers has increased by 17.7%, the batch switching time has been reduced by 57.8%, and the unplanned downtime has been reduced by 72.2%.

[0093] Table 3 Effect Comparison of Implementation

[0094]

[0095] The present invention forms a closed-loop digital management solution by constructing a quality-stability joint prediction model, realizing intelligent recommendation of process parameters and optimization of parameter switching paths, and predictive maintenance based on residual analysis. Taking fiber processing quality and equipment operation stability as the joint optimization objectives, when alternately processing different fiber types on the same production line, it can automatically calculate the optimal combination of process parameters and its smooth switching path according to the differences in fiber physical properties, ensuring both the fine processing quality of high-value fibers and the high-efficiency output of ordinary fibers. At the same time, by real-time monitoring the deviation between the equipment state and the theoretical predicted value, potential faults can be identified early, effectively avoiding equipment abnormalities caused by frequent switching of working conditions. It solves the key contradictions among quality, efficiency, and stability in the scenario of alternate processing of different fibers by combers.

[0096] Example 2:

[0097] This embodiment provides a digital comber equipment management system based on data analysis, as Figure 3 shown, including:

[0098] A data acquisition unit that acquires production plan information, fiber properties, historical production data, and historical equipment maintenance data;

[0099] A joint prediction unit that constructs a quality-stability joint prediction model based on historical production data using machine learning algorithms, receives fiber properties and corresponding process parameter combinations, and predicts the corresponding expected processing quality and expected equipment stability;

[0100] A parameter switching unit that combines production plan information, uses the quality-stability joint prediction model to predict and evaluate candidate process parameter combinations, and aims at optimizing processing quality, equipment stability, and equipment utilization rate for different fiber batch switches, calculates recommended process parameter combinations and recommended parameter switching paths;

[0101] An evaluation and update unit that collects the actual processing quality and actual equipment stability during equipment operation, conducts management evaluation in combination with the quality-stability joint prediction model, and iteratively updates the quality-stability joint prediction model;

[0102] A predictive maintenance unit that analyzes the residual sequence of actual equipment stability and expected equipment stability, and generates predictive maintenance suggestions in combination with historical equipment maintenance data.

[0103] Furthermore, LightGBM is used to construct the quality-stability joint prediction model, and a multi-objective regression strategy is adopted. A quality-stability joint prediction sub-model is trained for each index of expected processing quality and expected equipment stability respectively; the fiber properties to be processed and the corresponding process parameter combinations are used to form an input feature vector, and the input feature vector is input into the trained quality-stability joint prediction sub-model to obtain the corresponding index prediction value; the index prediction values output by all quality-stability joint prediction sub-models are integrated to obtain the expected processing quality and expected equipment stability.

[0104] Furthermore, the prediction process of the quality-stability joint prediction sub-model includes: S1. Processing the input feature vector through a first decision tree to obtain a preliminary index prediction value; S2. Processing the input feature vector through a second decision tree to obtain a residual prediction value, and adding the preliminary index prediction value and the residual prediction value to obtain the updated preliminary index prediction value; S3. Repeating step S2 to traverse the remaining decision trees in turn, and obtaining the index prediction value after the traversal ends.

[0105] Further, the steps of calculating the recommended process parameter combinations and the recommended parameter switching paths specifically include:

[0106] Define the decision variables as the process parameter combinations and parameter adjustment paths of the new batch of fibers, and construct an objective function with the goal of maximizing the expected processing quality, minimizing the impact of the switching process on equipment stability, and minimizing the switching time;

[0107] Set physical constraints, stability constraints, and path smoothness constraints for the parameters; select the Bayesian optimization algorithm, and guide the search process by constructing a surrogate model of the objective function to iteratively generate candidate process parameter combinations and candidate parameter switching paths;

[0108] Combine the quality-stability joint prediction model to evaluate the objective function values corresponding to the candidate process parameter combinations. When the stopping condition is met, output the recommended process parameter combinations that optimize the objective function under the constraint conditions and the corresponding recommended parameter switching paths.

[0109] Further, the steps of iteratively updating the quality-stability joint prediction model specifically include: calculating management evaluation indicators based on the expected processing quality and expected equipment stability, including prediction error, target quality achievement rate, stability compliance rate, and improvement of overall equipment efficiency; setting model update trigger conditions according to the management evaluation indicators. When the conditions are met, adopt a periodic batch retraining strategy, and use the newly obtained actual production data combined with some historical production data to retrain the quality-stability joint prediction submodels that need to be updated.

[0110] Further, the steps of generating predictive maintenance suggestions specifically include:

[0111] Real-time calculate the residual sequence between the actual equipment stability and the expected equipment stability; apply time series analysis techniques to the residual sequence to extract trend feature vectors;

[0112] Use historical equipment maintenance data and the corresponding historical residual sequences to train a fault mode classifier for outputting predicted fault types according to the current trend feature vectors;

[0113] Set predictive maintenance trigger rules based on the confidence level of the predicted fault type and the residual amplitude. When the predictive maintenance trigger rules are met, automatically generate predictive maintenance suggestions.

[0114] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A digital comber equipment management method based on data analysis, characterized in that, Including: Obtain production plan information, fiber properties, historical production data, and historical equipment maintenance data; Based on historical production data, use the LightGBM algorithm to construct a quality-stability joint prediction model, adopt a multi-objective regression strategy, and train a quality-stability joint prediction sub-model for each of the expected processing quality including evenness, nep count, impurity count, and short fiber content, and the expected equipment stability including main shaft vibration amplitude, motor load rate, and energy consumption index; Receive the fiber properties and the corresponding process parameter combinations to form an input feature vector, input the input feature vector into the trained quality-stability joint prediction sub-model to obtain the corresponding index prediction value; Integrate the index prediction values output by all quality-stability joint prediction sub-models to obtain the expected processing quality and expected equipment stability; Combined with the production plan information, use the quality-stability joint prediction model to predict and evaluate the candidate process parameter combinations, and for the switching of different fiber batches, with the management objectives of optimizing processing quality, equipment stability, and equipment utilization rate, calculate the recommended process parameter combinations and recommended parameter switching paths; Collect the actual processing quality and actual equipment stability during equipment operation, conduct management evaluation in combination with the quality-stability joint prediction model, and iteratively update the quality-stability joint prediction model; Analyze the residual sequence of the actual equipment stability and the expected equipment stability, and generate predictive maintenance suggestions in combination with the historical equipment maintenance data.

2. The method for managing a digital comber device based on data analysis according to claim 1, wherein The prediction process of the quality-stability joint prediction sub-model includes: S1. Process the input feature vector through the first decision tree to obtain a preliminary index prediction value; S2. Process the input feature vector through the second decision tree to obtain a residual prediction value, and add the preliminary index prediction value and the residual prediction value to obtain the updated preliminary index prediction value; S3. Repeat step S2 to sequentially traverse the remaining decision trees, and after the traversal is completed, obtain the index prediction value.

3. A digital comber equipment management method based on data analysis according to claim 1, characterized in that, The steps of calculating the recommended process parameter combinations and recommended parameter switching paths specifically include: Define the decision variables as the process parameter combinations and parameter adjustment paths of the new batch of fibers, and construct an objective function with the goals of maximizing the expected processing quality, minimizing the impact of the switching process on equipment stability, and minimizing the switching time; Set parameter physical constraints, stability constraints, and path smoothness constraints; Select the Bayesian optimization algorithm, and guide the search process by constructing a surrogate model of the objective function to iteratively generate candidate process parameter combinations and candidate parameter switching paths; Evaluate the objective function values corresponding to the candidate process parameter combinations in combination with the quality-stability joint prediction model, and until the stop condition is met, output the recommended process parameter combinations that optimize the objective function under the constraint conditions and the corresponding recommended parameter switching paths.

4. A digital comber equipment management method based on data analysis according to claim 1, characterized in that, The steps for iteratively updating the quality-stability joint prediction model specifically include: calculating management evaluation indicators according to the expected processing quality and expected equipment stability, including prediction error, target quality achievement rate, stability compliance rate, and improvement of overall equipment efficiency; setting model update trigger conditions based on the management evaluation indicators, and when the conditions are met, adopting a periodic batch retraining strategy, merging newly acquired actual production data with some historical production data, and retraining the quality-stability joint prediction submodel that needs to be updated.

5. A digital comber equipment management method based on data analysis according to claim 1, characterized in that, The steps for generating predictive maintenance suggestions specifically include: Calculating the residual sequence between the actual equipment stability and the expected equipment stability in real time; applying time series analysis techniques to the residual sequence to extract trend feature vectors; Training a fault mode classifier using historical equipment maintenance data and the corresponding historical residual sequences to output the predicted fault type according to the current trend feature vector; Setting predictive maintenance trigger rules based on the confidence level of the predicted fault type and the residual amplitude, and automatically generating predictive maintenance suggestions when the predictive maintenance trigger rules are met.

6. A digital comber equipment management system based on data analysis, characterized in that, Including: A data acquisition unit that acquires production plan information, fiber properties, historical production data, and historical equipment maintenance data; A joint prediction unit that constructs a quality-stability joint prediction model based on historical production data using the LightGBM algorithm, adopts a multi-objective regression strategy, and trains a quality-stability joint prediction submodel for each of the evenness, neps count, impurity count, and short fiber content included in the expected processing quality and the spindle vibration amplitude, motor load rate, and energy consumption index included in the expected equipment stability; receives the fiber properties and the corresponding process parameter combinations to form an input feature vector, inputs the input feature vector into the trained quality-stability joint prediction submodel to obtain the corresponding index prediction values; integrates the index prediction values output by all quality-stability joint prediction submodels to obtain the expected processing quality and expected equipment stability; A parameter switching unit that combines the production plan information and uses the quality-stability joint prediction model to predict and evaluate candidate process parameter combinations, and calculates the recommended process parameter combinations and recommended parameter switching paths for the switching of different fiber batches with the management objectives of optimizing processing quality, equipment stability, and equipment utilization rate; An evaluation and update unit that collects the actual processing quality and actual equipment stability during equipment operation, conducts management evaluation in combination with the quality-stability joint prediction model, and iteratively updates the quality-stability joint prediction model; A predictive maintenance unit that analyzes the residual sequence of the actual equipment stability and the expected equipment stability, and generates predictive maintenance suggestions in combination with historical equipment maintenance data.

Images