A two-component fiber textile process intelligent scheduling method and system

Abstract

This invention relates to the field of intelligent textile and industrial automation technology, and discloses an intelligent scheduling method and system for bicomponent fiber textile processes. The method includes: acquiring a workshop task allocation sequence and physical operating parameters; performing dynamic load compensation calculations based on the physical operating parameters to obtain initial machine load values; analyzing energy consumption characteristics based on the initial machine load values ​​to obtain multi-dimensional energy consumption data; performing time-series analysis and fiber physical index reconstruction on the multi-dimensional energy consumption data to obtain a fiber morphology distribution matrix; analyzing batches with abnormal morphology based on the fiber morphology distribution matrix to obtain statistical results of non-conforming batches; performing incremental prediction on the statistical results of non-conforming batches to obtain a cumulative rework frequency prediction value; and using the cumulative rework frequency prediction value as a constraint condition to perform iterative optimization of the workshop task allocation sequence to generate a target process execution sequence. This method can achieve dynamic and precise scheduling of multiple machines.

Description

Technical Field

[0001] This invention relates to the field of intelligent textile and industrial automation technology, and in particular to an intelligent scheduling method and system for bicomponent fiber textile processes. Background Technology

[0002] Currently, in the field of modern textile production, the drafting machine, as the core equipment for bicomponent fiber processing, directly determines the quality stability of the final fiber product through the stability of its operating status and the accuracy of its output drafting force. Production workshops typically require multiple drafting machines to operate collaboratively according to a pre-set task allocation sequence. To ensure continuous and high-intensity production, it is necessary to monitor the load characteristics and heat parameters of multiple machines around the clock, and to combine this with multi-dimensional production line scheduling and quality feedforward control in conjunction with the process flow, thereby achieving information-based and intelligent industrial production under complex textile processes.

[0003] In a current technology, an adaptive task scheduling execution unit management system typically utilizes federated learning and hybrid neural networks to extract general execution unit features, combines deep reinforcement learning to evaluate equipment capabilities and predict status, and finally achieves task decomposition and scheduling optimization through spectral clustering and multi-armed machine algorithms. However, when dealing with complex bicomponent fiber textile processes, this current technology lacks deep modeling of the physical characteristics of textile drafting machines and fails to establish a predictive correlation between equipment depreciation, heat generation characteristics, and fiber evenness. This limits the system's ability to manage big data and fuse and analyze multi-source heterogeneous data, making it unable to predict the impact of the mechanical deterioration state of specific equipment on fiber evenness before task execution. Because it cannot accurately predict the rework frequency before physical product inspection and dynamically optimize the task sequence accordingly, some severely degraded equipment is often assigned high-intensity tasks, leading to substandard fiber evenness and yarn breakage problems during continuous operation.

[0004] In summary, existing technologies suffer from the inability to predict fiber quality defects in advance and the lack of feedforward guidance on product rework frequency in task sequence scheduling. Summary of the Invention

[0005] This invention provides an intelligent scheduling method and system for bicomponent fiber textile processes to solve the problems of being unable to predict fiber quality defects in advance and the lack of feedforward guidance on product rework frequency in task sequence scheduling.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides an intelligent scheduling method for bicomponent fiber textile processes, comprising:

[0007] Obtain the workshop task allocation sequence and the physical operating parameters of multiple drawing machines, and perform dynamic load compensation calculation based on the physical operating parameters to obtain the initial machine load value;

[0008] Based on the initial machine load values, energy consumption characteristics are analyzed to obtain multi-dimensional energy consumption data;

[0009] Time-series trend analysis is performed on the multidimensional energy consumption data to obtain current change characteristics. Based on the current change characteristics, fiber physical indicators are reconstructed to obtain a fiber morphology distribution matrix.

[0010] Based on the fiber morphology distribution matrix, analyze the batches with abnormal morphology to obtain the statistical results of unqualified batches;

[0011] Based on the statistical results of the non-conforming batches, the production line number is matched, and the pre-stored historical frequency records are retrieved based on the production line number to perform trend extrapolation prediction, thereby obtaining the cumulative rework frequency prediction value.

[0012] Using the predicted cumulative rework frequency as a constraint, the workshop task allocation sequence is iteratively optimized to generate the target process execution sequence.

[0013] Secondly, the present invention provides an intelligent scheduling system for a bicomponent fiber textile process, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described above.

[0014] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0015] Compared with the prior art, the present invention has the following beneficial effects:

[0016] (1) This invention determines the initial machine load value by analyzing the pre-established task allocation sequence to extract processing time and intensity values, and combines the real-time speed and tension fluctuation values ​​collected by high-frequency sensors. By querying a pre-calibrated speed-attenuation rate mapping table, the load attenuation rate is obtained, and then the initial machine load is corrected by multiplying the basic load data by a correction factor. This scheme, by linking real-time dynamic tension anomalies with the depth of load attenuation rate, can accurately predict and dynamically compensate for the actual power demand of the equipment when the fiber material is under unstable tension, thereby avoiding energy waste or equipment start-up obstruction caused by blindly setting the load.

[0017] (2) This invention obtains a baseline value for heat loss by synchronously collecting the surface temperature and operating frequency of core mechanical components, combined with the cumulative power-on operating time, using a load-heat mapping table, and then weighted and fused with the life consumption ratio to obtain an assessment value for the equipment depreciation degree. This scheme establishes a quantitative correlation between the physical wear and heat generation of equipment and its depreciation status, and can deeply explore the mechanical attenuation law behind the high-dimensional physical state, thereby accurately and keenly capturing the abnormal heat generation and physical aging tendency of the core components of the drawing machine before production execution.

[0018] (3) This invention obtains the rate of change feature by performing first-order difference on the active / reactive fluctuation amplitude sequence, reconstructs the fiber morphology distribution matrix by querying a pre-calibrated feature-index mapping table, realizes the estimation of physical indicators such as thickness uniformity, obtains the rework frequency prediction value by statistical analysis of non-conforming batches and extrapolation prediction based on historical trends, and finally optimizes the task sequence iteratively by using the prediction value as a constraint condition through a genetic algorithm; this scheme realizes the advance prediction of product quality and the feedforward optimization of task scheduling. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of a method for intelligent scheduling of a two-component fiber textile process provided in the first embodiment of the present invention;

[0020] Figure 2 This is a two-dimensional line graph showing the current variation characteristics provided in the first embodiment of the present invention;

[0021] Figure 3 This is a line graph showing the convergence of the fitness iteration of the genetic algorithm provided in the first embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Reference Figure 1 The first embodiment of the present invention provides an intelligent scheduling method for a two-component fiber textile process, comprising the following steps:

[0024] S11, Obtain the workshop task allocation sequence and the physical operating parameters of multiple drawing machines, and perform dynamic load compensation calculation based on the physical operating parameters to obtain the initial machine load value;

[0025] S12, perform energy consumption characteristic analysis based on the initial machine load value to obtain multi-dimensional energy consumption data;

[0026] S13, perform time-series trend analysis on the multidimensional energy consumption data to obtain current change characteristics, and reconstruct fiber physical indicators based on the current change characteristics to obtain a fiber morphology distribution matrix;

[0027] S14, perform morphological anomaly batch analysis based on the fiber morphology distribution matrix to obtain the statistical results of unqualified batches;

[0028] S15, Match the production line number according to the statistical results of the non-conforming batches, and retrieve the pre-stored historical frequency records according to the production line number to perform trend extrapolation prediction and obtain the cumulative rework frequency prediction value.

[0029] S16, using the predicted cumulative rework frequency as a constraint, perform sequence iteration optimization on the workshop task allocation sequence to generate the target process execution sequence.

[0030] In step S11, the workshop task allocation sequence and the physical operating parameters of multiple drawing machines are obtained, and dynamic load compensation calculation is performed based on the physical operating parameters to obtain the initial machine load value, including:

[0031] Analyze the workshop task allocation sequence to obtain processing time and intensity values;

[0032] Real-time rotational speed and tension fluctuation values ​​are extracted from the physical operating parameters. If the tension fluctuation value exceeds the preset tension fluctuation threshold, the load attenuation rate is obtained by dynamic mapping based on the real-time rotational speed.

[0033] The initial machine load value is obtained by predicting the load based on the processing time, the intensity value, and the load attenuation rate.

[0034] In this step, the workshop task allocation sequence refers to a digital task list issued by the production execution system, containing batches of fibers to be processed and their corresponding process parameters. The physical operating parameters refer to raw signals characterizing the current mechanical and electrical status of the equipment, collected in real time by sensors deployed on key components of the drawing machine. This embodiment calls the data interface of the production execution system to obtain the workshop task allocation sequence within the current production cycle. This sequence is encapsulated using Extensible Markup Language (XML) format, where each task node contains two attribute fields: "planned processing time" and "target drawing strength." This embodiment extracts the value of the "planned processing time" attribute field by traversing the node tree of the task allocation sequence to obtain the processing time in minutes, and simultaneously extracts the value of the "target drawing strength" attribute field to obtain the strength value in Newtons.

[0035] This embodiment reads pulse signals from an incremental rotary encoder deployed on the main shaft of the drafting machine via a fieldbus. The encoder outputs a fixed number of pulses per revolution. This embodiment measures the number of pulses per unit time and then converts the signal to a frequency-to-voltage value to obtain the real-time rotational speed in revolutions per minute. Simultaneously, differential voltage signals are read from tension sensors installed on both sides of the drafting roller assembly. These tension sensors contain strain gauges with cantilever beam structures; when fiber tension changes, the resistance of the strain gauges changes, causing a linear change in the output voltage. This embodiment converts this voltage signal into a digital value using an analog-to-digital converter and multiplies it by a pre-calibrated conversion factor to obtain the tension fluctuation value in Newtons. The conversion factor is provided by the sensor's factory calibration certificate and is measured in Newtons per volt.

[0036] The preset tension fluctuation threshold is determined through statistical analysis. After each drawing machine is installed and commissioned at the factory, and when it is in a stable operating state and the uniformity index of the produced fibers meets the superior product standard, tension fluctuation data is continuously collected for 24 hours, recording the sampling value every second to form a sample set. The arithmetic mean of this sample set is calculated, and then the square of the difference between each data point in the sample set and the mean is calculated. All squares are added together, divided by the number of data points, and the square root is taken to obtain the standard deviation of the sample set. According to the three Sigma principle in statistics, the preset tension fluctuation threshold is determined as the arithmetic mean plus three times the standard deviation, and this value is stored in the control system memory of the equipment.

[0037] In this embodiment, the tension fluctuation value read in real time is compared with the preset tension fluctuation threshold read from the memory. If the tension fluctuation value is less than or equal to the preset tension fluctuation threshold, the load attenuation rate is set to zero. If the tension fluctuation value is greater than the preset tension fluctuation threshold, the load attenuation rate is obtained by querying a pre-established speed-attenuation rate mapping table. This mapping table is established individually for each device using an engineering calibration method. In the offline no-load state of the device, the motor speed is set to multiple percentage points of its rated speed by the frequency converter. At each set speed point, a simulated disturbance torque is applied to the stretching roller shaft by a magnetic powder brake, and the excitation current of the magnetic powder brake is adjusted so that the tension fluctuation value reaches 1.2 times the preset tension fluctuation threshold. The output current value automatically increased by the frequency converter to overcome the disturbance and maintain the current speed is recorded. The ratio of this increased current value to the rated current of the motor is used as the load attenuation rate corresponding to that speed point, with a value range of 0 to 0.5. Each test speed point and its corresponding load attenuation rate are made into a two-dimensional table and stored in the non-volatile memory of the control system. During real-time operation, if the current real-time speed is exactly equal to a certain test speed point, the corresponding load attenuation rate is read directly; if the current real-time speed is between two test speed points, the linear interpolation method is used to calculate the value between the two known attenuation rates according to the ratio of the speeds as the current load attenuation rate.

[0038] This embodiment obtains the base load by querying a pre-established task-load mapping table. This mapping table is established by executing multiple standardized test tasks with different combinations of processing duration and intensity values ​​on newly manufactured equipment that has not undergone long-term wear and tear. The processing duration values ​​cover the typical operating time range of the equipment, and the intensity values ​​cover the typical drawing process range. For each combination, a power analyzer deployed on the equipment's power supply line records the input active power throughout the entire process from start to finish. The arithmetic mean of the active power throughout the process is calculated and used as the base load corresponding to that task combination, in watts. The test results are then entered into a two-dimensional table with processing duration as the row index and intensity value as the column index, and stored in the control system's memory.

[0039] In this embodiment, the base load is multiplied by a correction factor to obtain the initial machine load value. The correction factor is equal to one plus the load attenuation rate, wherein the load attenuation rate ranges from 0 to 1, and is expressed as a decimal percentage of attenuation. When the tension fluctuation value does not exceed the threshold, the attenuation rate is 0; when it exceeds the threshold, the load attenuation rate is obtained by looking up a table, and ranges from 0 to 0.5.

[0040] In step S12, energy consumption characteristics are analyzed based on the initial machine load values ​​to obtain multi-dimensional energy consumption data, including:

[0041] Based on the initial machine load value, the heat loss is assessed to obtain the equipment depreciation assessment value;

[0042] Based on the equipment depreciation assessment value, energy consumption fluctuations are mapped to obtain multidimensional energy consumption data.

[0043] In this step, the initial machine load value refers to the estimated input power value output in step S11 after dynamic load compensation calculation, and its unit is watts. The multidimensional energy consumption data refers to a set of electrical parameters representing the equipment's operating status from the perspective of power quality. In this embodiment, the cumulative power-on running time of the drawing machine is read from the non-volatile memory of the control system, in hours. The cumulative power-on running time is recorded as follows: the system starts timing each time the equipment is powered on and stops timing each time the power is cut off and stops, and the duration of this operation is added to the historical storage value.

[0044] This embodiment obtains a heat loss baseline value by querying a pre-established load-heat mapping table. This mapping table is established by running the equipment at different percentages of its rated power under its new factory condition. After 30 minutes of stable operation, the highest temperature on the surface of the spindle bearing housing is measured using an infrared thermal imager. This temperature value is subtracted from the current ambient temperature measured by an ambient temperature sensor to obtain the temperature rise value. This temperature rise value is used as the heat loss baseline value corresponding to that load level, in degrees Celsius. The load percentage values ​​are 20%, 40%, 60%, 80%, and 100% of the rated power. The test results are compiled into a two-dimensional table and stored in the control system's memory. During real-time operation, the initial machine load value is divided by the equipment's rated power to obtain a load percentage. Then, by querying the mapping table and using linear interpolation, the current heat loss baseline value is obtained.

[0045] In this embodiment, the cumulative power-on running time and the heat loss benchmark value are combined to obtain a value between 0 and 100 as the equipment depreciation assessment value. Dividing the cumulative power-on operating time by the designed service life of the equipment yields a dimensionless operating life consumption ratio; dividing the heat loss benchmark value by the standard temperature rise value of the equipment under rated load yields a dimensionless heat deviation ratio; multiplying the operating life consumption ratio by 0.3 and the heat deviation ratio by 0.7, adding the two products, and then multiplying by 100. If the calculated value exceeds 100, it is taken as 100; if the calculated value is less than 0, it is taken as 0; the weights 0.3 and 0.7 are determined by the least squares regression method. During the offline testing phase, 20 drawing machines with different operating years are selected, and the operating life consumption ratio and heat deviation ratio of each machine are recorded. At the same time, the equipment maintenance personnel give a depreciation degree score from 0 to 100 based on the bearing wear measurement value and the transmission clearance measurement value. Using the two ratios as independent variables and the depreciation degree score as the dependent variable, a multiple linear regression is performed. The coefficients of the two independent variables obtained from the regression are normalized so that the sum of the two is 1, resulting in weights 0.3 and 0.7. The designed service life of the equipment and the standard temperature rise value of the equipment under rated load are provided by the equipment manufacturer and are stored in the non-volatile memory of the control system.

[0046] This embodiment obtains multidimensional energy consumption data by querying a pre-established depreciation-energy consumption fluctuation mapping table. This mapping table is established using an engineering calibration method. With the equipment offline, multiple drawing machines of the same model at different depreciation levels are selected as test samples, and the depreciation assessment value of each test sample is used as the horizontal axis. An initial machine load value determined in step S11 is applied to each test sample as the load. After stable operation, three-phase voltage and three-phase current waveform data are continuously collected over a period of time using a power quality analyzer. Based on the collected waveform data, the average active power within each power frequency cycle is calculated, resulting in an active power time series arranged periodically. The difference between the maximum and minimum values ​​of this time series is calculated and divided by the average value of the time series to obtain the active power fluctuation amplitude. Similarly, the average reactive power within each power frequency cycle is calculated, resulting in a reactive power time series arranged periodically. The difference between the maximum and minimum values ​​of this time series is calculated and divided by the average value of the time series to obtain the reactive power fluctuation amplitude. Two two-dimensional mapping tables are created for each test sample, along with its corresponding active power fluctuation amplitude and reactive power fluctuation amplitude, and then stored in the control system memory.

[0047] In real-time operation, the calculated equipment depreciation assessment value is used as the retrieval key to query the depreciation-active power fluctuation mapping table and the depreciation-reactive power fluctuation mapping table respectively. If the depreciation assessment value lies between two test points, linear interpolation is used to calculate a value between two known fluctuation amplitudes according to the proportional relationship of the depreciation assessment values. The retrieved active power fluctuation amplitude and reactive power fluctuation amplitude are encapsulated into a data structure to form multi-dimensional energy consumption data; the active power fluctuation amplitude and reactive power fluctuation amplitude calculated at the current moment are written together with the timestamp into the historical circular cache to form a fluctuation amplitude sequence arranged by time for subsequent steps.

[0048] In step S13, time-series trend analysis is performed on the multidimensional energy consumption data to obtain current variation characteristics. Based on the current variation characteristics, fiber physical indicators are reconstructed to obtain a fiber morphology distribution matrix, including:

[0049] The fluctuation amplitude sequences of active power and reactive power are extracted from the multidimensional energy consumption data and first-order difference calculation is performed to obtain the active power change rate sequence and the reactive power change rate sequence.

[0050] The active power rate sequence and the reactive power rate sequence at the same sampling time are combined into a two-dimensional current change feature vector to obtain the current change characteristics of the drawing machine under continuous operation.

[0051] Based on the current change characteristics, a pre-established feature-index mapping table is queried to obtain multiple physical index values;

[0052] The values ​​of multiple physical indicators are arranged in a predetermined order to obtain a fiber morphology distribution matrix that includes the values ​​of thickness uniformity.

[0053] In this step, the multidimensional energy consumption data refers to the set of electrical parameters output in step S12, which includes two dimensions: active power fluctuation amplitude and reactive power fluctuation amplitude. The fiber morphology distribution matrix refers to a set of values ​​organized in matrix form that characterizes multiple physical indicators of the fiber.

[0054] This embodiment performs time-series trend analysis on the multidimensional energy consumption data to obtain the current variation characteristics of the drawing machine under continuous operation. In one implementation, this embodiment first reads the multidimensional energy consumption data sequence stored in the historical database of the control system over a past period. This sequence records the active power fluctuation amplitude and reactive power fluctuation amplitude at each sampling time according to a fixed sampling interval. The sampling interval is consistent with the duration of waveform data acquisition in step S12, and is set to sixty seconds in this embodiment. This embodiment performs first-order difference calculations on the read active power fluctuation amplitude sequence and reactive power fluctuation amplitude sequence respectively. The first-order difference calculation method is to subtract the value of the previous sampling point from the value of the current sampling point to obtain the change at that sampling point, and then divide by the sampling interval to obtain the rate of change. Through first-order difference calculation, the active power change rate sequence and the reactive power change rate sequence are obtained.

[0055] This embodiment combines the active power change rate sequence and the reactive power change rate sequence to form a two-dimensional current change feature vector sequence. Specifically, the active power change rate and reactive power change rate at the same sampling moment are used as the two components of the feature vector at that moment, arranged in chronological order to form a vector sequence that evolves over time; this sequence represents the current change characteristics. It should be noted that the fluctuation of active power mainly reflects the change of the fundamental component of the motor stator current used for work, while the fluctuation of reactive power mainly reflects the change of the excitation current component required to establish the internal magnetic field of the motor. The combination of these two rates of change can characterize the dynamic fluctuation law of the total stator current of the motor under continuous operation.

[0056] This embodiment reconstructs fiber physical indicators based on the current change characteristics, obtaining a fiber morphology distribution matrix including thickness uniformity values. This embodiment reconstructs the fiber physical indicators by querying a pre-established feature-indicator mapping table. This mapping table is stored in a two-dimensional table structure, with row indices representing the rate of change of active power and column indices representing the rate of change of reactive power. Each table cell stores a set of physical indicator values, including thickness uniformity, breaking strength, and evenness. The ranges of both the active and reactive power rates are divided into twenty equal intervals, with the step size of each interval determined by dividing the range of data collected during offline testing by twenty. This mapping table is established using an engineering calibration method; that is, during the offline testing phase, multiple batches of fiber samples from different groups are selected and processed on a drawing machine according to standardized processes. During processing, multidimensional energy consumption data of the drawing machine are collected simultaneously. Then, following the time-series trend analysis method described in step S13, first-order difference calculations are performed on the active power fluctuation amplitude sequence and the reactive power fluctuation amplitude sequence to obtain the active power change rate sequence and the reactive power change rate sequence, which in turn constitute the current change characteristic vector sequence. After processing is completed, physical index tests are performed on each batch of fiber samples.

[0057] The physical index testing includes at least three indicators: fiber uniformity, breaking strength, and yarn evenness. The fiber uniformity test method uses an online yarn evenness tester to collect fiber diameter data at a density of 100 sampling points per meter. The ratio of the standard deviation of the diameters at all sampling points to the average diameter is calculated, and then multiplied by 100% to obtain the fiber uniformity value, expressed as a percentage. The breaking strength test method uses a single-fiber tensile tester to apply tension to the fiber at a standard tensile rate until it breaks. The maximum tensile force at break is recorded and divided by the fiber's linear density to obtain the breaking strength value, expressed as centineutcisedes per tex (cN / tex). The yarn evenness test method uses a capacitive yarn evenness meter to measure the coefficient of variation of the fiber's mass per unit length, expressed as a percentage.

[0058] The current change feature vector sequence of each test sample at each sampling time is paired with the corresponding detection results of the three physical indicators. For each sampling time in each test sample, the active power change rate value is used as the row index and the reactive power change rate value is used as the column index. The corresponding three physical indicator values ​​are stored in the cell corresponding to the index in a two-dimensional table. If multiple test samples fall into the same table cell, the arithmetic mean of the physical indicator values ​​in that cell is taken and stored. After all test samples are filled, for the still empty table cells, bilinear interpolation is used to fill them based on the values ​​of adjacent non-empty cells. The two-dimensional mapping table constructed above is stored in the control system memory. This mapping table is established separately for fiber varieties. During implementation, the corresponding mapping table is automatically switched according to the variety code being processed. The variety code is issued by the production execution system along with the task allocation sequence. For the current change feature vector collected in real time, first determine the row index interval where the active change rate of the vector is located and the column index interval where the reactive change rate is located, locate the four adjacent cells in the mapping table, and then use bilinear interpolation to calculate the physical index value corresponding to the current feature vector, where the coarseness uniformity value is the first component of the interpolation result.

[0059] In this embodiment, the obtained uniformity of thickness, breaking strength, and unevenness values ​​are arranged in a predetermined order to form a matrix with one row and three columns. This matrix is ​​the fiber morphology distribution matrix. The first element is the uniformity of thickness, the second element is the breaking strength, and the third element is the unevenness.

[0060] like Figure 2As shown, this figure is a line chart of the current change characteristics collected by the drafting machine under typical operating conditions; the red line in the figure represents the sequence of active power change rate, and the blue line represents the sequence of reactive power change rate; by comparing the fluctuation amplitudes and trends of the two curves at different times, the transition of the drafting machine load state can be clearly identified. For example, the significant peak near the 6th sampling period corresponds to the state of sudden change in the drafting machine load, which provides the core basis for subsequent reconstruction of the fiber morphology distribution matrix.

[0061] In step S14, according to the fiber morphology distribution matrix, analyze the batches with abnormal morphology to obtain the statistical results of unqualified batches, including:

[0062] Extract the fineness uniformity value from the fiber morphology distribution matrix. If the fineness uniformity value is lower than the preset uniformity threshold, obtain the corresponding trace code for the abnormal batch.

[0063] Perform cumulative frequency increment statistics according to the trace code of the abnormal batch to obtain the statistical results of unqualified batches.

[0064] In this step, the fiber morphology distribution matrix refers to a one-row and three-column matrix output by step S13, which contains the fineness uniformity value, the breaking strength value, and the evenness variation coefficient value. The statistical results of unqualified batches refer to the total number of fiber batches determined to be unqualified within a statistical period. In this embodiment, the fineness uniformity value is extracted from the fiber morphology distribution matrix. This value is located at the first element position of the matrix, and directly read the value of this element to obtain a value in percentage.

[0065] In this embodiment, the extracted fineness uniformity value is compared with the preset uniformity threshold. The preset uniformity threshold is determined by statistical analysis. Specifically, from the historical production records, collect the fineness uniformity test data corresponding to all fiber batches determined to be qualified by the final quality inspection department to form a sample set. Calculate the arithmetic mean of this sample set, then calculate the square of the difference between each data point in this sample set and this mean, add all the squared values and divide by the number of data points, and then take the square root to obtain the standard deviation of this sample set. Determine the preset uniformity threshold as the arithmetic mean minus three times the standard deviation. This value represents the lower limit boundary of the qualified products in terms of the fineness uniformity index, and values lower than this boundary are statistically outlier points.

[0066] If the uniformity value is greater than or equal to the preset uniformity threshold, the batch of fibers is determined to be qualified, and no statistical operation is performed on the batch corresponding to this sampling. If the uniformity value is lower than the preset uniformity threshold, the batch of fibers is determined to be unqualified. In this case, this embodiment obtains the corresponding abnormal batch traceability code. This embodiment obtains the corresponding abnormal batch traceability code by calling the data interface of the production execution system and querying the corresponding abnormal batch traceability code based on the fiber batch information currently being processed. It should be noted that the abnormal batch traceability code is a unique identifier assigned by the production execution system to each processing batch, which is usually composed of the production date, production line number, batch number of the day, and defect type code.

[0067] This embodiment performs cumulative frequency increment statistics based on the abnormal batch traceability code to obtain the statistical results of non-conforming batches. In one implementation, this embodiment maintains a non-conforming batch statistics table in memory. This statistics table uses natural days as the statistical period and the cumulative number of non-conforming batches as the value. The rule for dividing the statistical period is: 00:00:00 of each day is the start time of the statistical period, and 00:00:00 of the next day is the end time of the statistical period. At the same time, the system automatically resets the cumulative number of non-conforming batches to zero. When the abnormal batch traceability code is obtained, this embodiment parses the date information contained in the traceability code, extracts the year, month, and day fields, and takes the natural day that completely matches the three fields as the statistical period to which the batch belongs. Then, it searches for the cumulative number of non-conforming batches corresponding to the statistical period in the non-conforming batch statistics table. If found, the cumulative value is incremented by one and written back; if not found, the statistical period is inserted into the table as a new index, and the cumulative number of non-conforming batches is set to the initial value of one. This embodiment reads the cumulative number of non-conforming batches corresponding to the current statistical period in the non-conforming batch statistics table and outputs this value as the non-conforming batch statistical result. This value represents the total number of fiber batches that are deemed unqualified due to failure to meet the uniformity index within the current statistical period.

[0068] In step S15, the production line number is matched according to the statistical results of the non-conforming batches, and the pre-stored historical frequency records are retrieved according to the production line number to perform trend extrapolation prediction, thereby obtaining the cumulative rework frequency prediction value, including:

[0069] If the statistical result of the non-conforming batches is equal to zero, then the predicted value of the cumulative rework frequency is set to zero;

[0070] If the statistical result of the non-conforming batch is greater than zero, the defect type identifier is extracted based on the abnormal batch traceability code corresponding to the statistical result of the non-conforming batch and matched with the corresponding production line number.

[0071] Based on the production line number and the defect type identifier, read the historical rework frequency sequence from the pre-stored historical frequency record;

[0072] The historical rework frequency sequence is extrapolated to predict the predicted value for the next statistical period. The predicted value is then rounded to obtain the cumulative rework frequency prediction value.

[0073] In this step, the non-conforming batch statistics result refers to the value output in step S14 that represents the total number of non-conforming batches within the current statistical period. The cumulative rework frequency prediction value refers to the estimated number of fiber batches that need to be reworked due to the same defect type within a future complete statistical period. In this embodiment, defect type identifiers are extracted based on the non-conforming batch statistics result, and production line numbers are obtained by matching the defect type identifiers.

[0074] In one implementation, this embodiment first determines whether the value of the statistical result of the non-conforming batch is greater than zero. If the value is equal to zero, it is determined that no non-conforming batch was generated in the current statistical period, the cumulative rework frequency prediction value is set to zero, and all subsequent processing steps are skipped. If the value of the statistical result of the non-conforming batch is greater than zero, it indicates that there is a non-conforming batch in the current statistical period. This embodiment obtains the defect type identifier by parsing the abnormal batch traceability code obtained in step S14. Specifically, the defect type code is extracted from the abnormal batch traceability code. The traceability code format is YYYYMMDD-production line number-batch number-defect code, with each field separated by "-". The defect code is the fourth field. In this embodiment, the defect code corresponding to the non-compliance of thickness uniformity is fixed as 01. Other defect codes, such as non-compliance of fracture strength and non-compliance of unevenness, can be configured as other codes by the production execution system during implementation. This embodiment does not limit this. The defect type code is written into a specific field of the traceability code by the production execution system when determining that a batch is non-conforming. Different codes correspond to different quality defect categories.

[0075] This embodiment obtains the production line number by matching the defect type identifier. Specifically, this embodiment calls the equipment ledger interface of the production execution system, using the currently running drawing machine equipment number as the query key value to retrieve the production line number of that equipment. The mapping relationship between equipment number and production line number is uniformly maintained by the equipment ledger module of the production execution system, and is entered and updated by the system administrator when new equipment is added or production lines are adjusted. The production line number consists of letters and numbers and is used to uniquely identify a physical production line. This embodiment retrieves pre-stored historical frequency records based on the production line number to perform trend extrapolation prediction and obtain the cumulative rework frequency prediction value. The pre-stored historical frequency records refer to a time series data table stored in the control system database. This data table uses the production line number as the first index and the defect type code as the second index, recording the actual rework frequency of the production line for a specific defect type in multiple complete statistical periods in the past. The actual rework frequency refers to the number of non-conforming batches that the quality inspection department confirms need to be reworked after completing physical sampling inspection. The records in this data table are automatically updated by the production execution system after each physical sampling inspection is completed.

[0076] This embodiment uses the current production line number and defect type identifier as joint query conditions to read historical frequency data for several complete statistical periods from the historical frequency records. The number of periods read is determined by a preset sliding window length. The preset sliding window length is 4 statistical periods, determined through offline testing on a production scale of 200 batches per day. In practice, it can be configured to other integer values ​​according to the actual production scale. Specifically, historical data from the most recent 90 statistical periods are taken, and linear regression prediction is performed with window lengths of 2 to 10 periods respectively. The average absolute error under each window length is calculated, and the window length with the smallest error is selected as the configured value.

[0077] This embodiment performs trend extrapolation prediction on the read historical frequency data sequence. In one implementation, linear regression is used for prediction. Specifically, the statistical period number is used as the independent variable, and the historical frequency value of the corresponding period is used as the dependent variable. The arithmetic mean of the independent variable sequence and the arithmetic mean of the dependent variable sequence are calculated. The difference between the independent variable and its mean for each statistical period is multiplied by the difference between the dependent variable and its mean, and all products are added together to obtain the numerator. The square of the difference between the independent variable and its mean for each statistical period is calculated, and all squares are added together to obtain the denominator. The numerator is divided by the denominator to obtain the regression slope. The mean of the dependent variable is subtracted from the regression slope multiplied by the mean of the independent variables to obtain the regression intercept. The independent variable number corresponding to the next statistical period is substituted into the linear equation, that is, the regression slope is multiplied by the number and the regression intercept is added. The calculated value is the cumulative rework frequency prediction value. It should be noted that if the calculated predicted value is less than the statistical result of non-conforming batches in the current statistical period, the statistical result of non-conforming batches in the current period will be used as the predicted value; if the calculated cumulative rework frequency predicted value is a decimal, the rounding method will be used, that is, regardless of the size of the decimal part, the smallest integer greater than the value will be taken as the final predicted value output.

[0078] In step S16, the predicted cumulative rework frequency is used as a constraint to perform sequence iterative optimization on the workshop task allocation sequence to generate the target process execution sequence, including:

[0079] Using the predicted cumulative rework frequency as a constraint, the workshop task allocation sequence is subjected to population initialization processing to obtain the initial allocation population.

[0080] The fitness of the initially assigned population is evaluated to obtain fitness values;

[0081] The initial population is iteratively evolved based on the fitness value to generate the target process execution sequence.

[0082] In this step, the cumulative rework frequency prediction value refers to the estimated number of rework batches within a future statistical period output in step S15. The workshop task allocation sequence refers to the original task list obtained in step S11, which includes the batches of fibers to be processed and their corresponding process parameters. The target process execution sequence refers to the task execution order that, after optimization and adjustment, can reduce the predicted rework frequency. In this embodiment, the cumulative rework frequency prediction value is used as a constraint condition to perform population initialization processing on the workshop task allocation sequence to obtain an initial allocation population.

[0083] In one implementation, this embodiment first determines whether the predicted cumulative rework frequency is greater than zero. If the predicted value is zero, it is determined that the current scheduling scheme meets the quality requirements, and the original workshop task allocation sequence is directly output as the target process execution sequence, skipping all subsequent processing steps. If the predicted cumulative rework frequency is greater than zero, the sequence optimization process is triggered. This embodiment performs population initialization processing on the workshop task allocation sequence. Each individual in the population represents a task allocation scheme, and the encoding method adopts permutation encoding, that is, arranging all fiber batch numbers to be processed into a sequence according to the execution order. The population size is preset to 50, the maximum evolution generation is preset to 200, and the mutation probability is preset to 5%, that is, fifty different task execution order schemes are generated simultaneously. The above parameters were determined through offline experiments for a production scenario with a daily output of 200 batches and 10 machines. During implementation, when the number of tasks exceeds 100 batches or the number of machines exceeds 20, recalibration is required. Orthogonal experiments are conducted with different population sizes between 20 and 200 and different generations between 50 and 500, selecting the parameter combination with the highest fitness while ensuring computation time does not exceed 60 seconds and fitness values ​​converge (change rate less than 1% over 10 consecutive generations). The initial population is generated by using the original workshop task allocation sequence as an initial individual; the remaining forty-nine individuals are generated by performing a random swap operation on this original sequence, i.e., randomly selecting two different positions in the sequence, swapping the batch numbers at those two positions, and repeating this operation several times to obtain a mutated individual.

[0084] This embodiment evaluates the fitness of the initial allocation population to obtain a fitness value. The core objective of fitness evaluation is to measure the merits of a task allocation scheme under the constraint of the cumulative rework frequency prediction value. This embodiment constructs a fitness function, whose input is an individual in the population, and whose output is the fitness value of that individual. The higher the value, the better the scheme. First, the input task allocation sequence is simulated. For each fiber batch in the sequence, based on the drawing machine to which the batch will be allocated, and using the historical frequency records and the same trend extrapolation prediction method used in step S15, the predicted number of non-conforming batches for that drawing machine in the next statistical period is calculated. The calculated predicted number of non-conforming batches is compared with the cumulative rework frequency prediction value output in step S15. If the calculated predicted number of non-conforming batches exceeds the cumulative rework frequency prediction value output in step S15, a penalty score is applied to that batch. The penalty score is calculated by multiplying the number exceeding the prediction value by a penalty coefficient. If the number does not exceed the prediction value, the penalty score is zero. The penalty coefficient is preset to -10. In an offline simulation environment, the population convergence performance was tested with penalty coefficients of -5, -10, -15, -20, and -25, respectively. The objective function value of the final solution (the reciprocal of the total processing time) was used as the evaluation metric. Test results show that when the absolute value of the penalty coefficient is less than 10, the penalties for solutions exceeding the constraints are insufficient, and the algorithm cannot effectively eliminate unacceptable solutions; when the absolute value of the penalty coefficient is greater than 10, the algorithm is overly conservative, leading to premature shrinkage of the search space. Therefore, -10 was selected as the penalty coefficient.

[0085] Then, the total processing time for all batches in the sequence is calculated. The total processing time is calculated by summing the standard processing times of each batch on its assigned equipment. Finally, the reciprocal of the total processing time is multiplied by 100, and then the penalty score is subtracted to obtain the fitness value of the individual. All parameters are used in the calculation, with the total processing time in minutes, and a conversion factor of 100 used to scale the time value to the same order of magnitude as the penalty score. A higher fitness value indicates a shorter total processing time while meeting the rework frequency constraint. In this embodiment, the initial assigned population is iteratively evolved based on the fitness value to generate the target process execution sequence. The iterative evolution process is implemented using a genetic algorithm, specifically including three steps: selection, crossover, and mutation. The selection operation is used to select individuals with higher fitness values ​​from the current population as parents. This embodiment uses a roulette wheel selection method, where the probability of each individual being selected is proportional to the proportion of its fitness value to the total fitness value of the population.

[0086] Crossover is used to exchange partial sequences of two parent individuals to produce offspring. This embodiment uses partial mapping crossover, where two crossover points are randomly selected. The fragment from the first parent between these two crossover points is copied to the same position in the offspring. Then, the genes from the second parent that were not copied to the offspring are sequentially filled into the remaining positions in the offspring. It should be noted that the crossover probability is preset to 80%. This value was tested offline at four levels (40%, 60%, 80%, and 100%), and the 80% with the fastest convergence speed was selected as the preset value. In practice, it can be adjusted between 60% and 90%. Mutation is used to randomly perturb the sequences of offspring individuals to maintain population diversity. This embodiment uses exchange mutation, where two different positions in the offspring are randomly selected with a preset mutation probability, and the batch numbers at these two positions are exchanged. The mutation probability is preset to 5%. This value was tested offline at four levels (1%, 3%, 5%, and 10%), and the 5% with the fastest convergence speed and best solution quality was selected as the preset value.

[0087] The selection, crossover, and mutation operations described above constitute an evolutionary generation. This embodiment repeats this evolutionary generation until a termination condition is met. The termination condition is either one of the following two conditions being met first: the highest fitness value in the current population no longer increases for ten consecutive generations; or the total number of evolutionary generations executed reaches a preset maximum number of evolutionary generations, which is preset to be two hundred; the maximum number of generations can be increased based on actual production scenarios. When the termination condition is met, this embodiment selects the individual with the highest fitness value from the final population and outputs the task execution sequence corresponding to that individual as the target process execution sequence.

[0088] It is worth noting that after executing the target process sequence and completing the processing, the actual rework frequency data is obtained. The actual rework frequency that occurs within the current statistical period is written into the historical frequency record according to the production line number and defect type code, which is used to update the data benchmark for the next round of prediction.

[0089] like Figure 3 As shown in the figure, this is the fitness evolution curve of the sequence iterative optimization step in the embodiment; the horizontal axis is the number of algorithm iterations, and the vertical axis is the fitness value. The larger the value, the lower the overall cost of the current process allocation sequence; the curve shows that the algorithm experienced a rapid fitness increase in the early stage, and then oscillated for optimization in the range of 70 to 85. Finally, around the 15th generation, it found the optimal process allocation scheme with a fitness of 85.0 (shown by the red dot in the figure), and successfully achieved task scheduling optimization based on rework frequency prediction constraints.

[0090] In summary, this invention achieves forward prediction of quality defects by dynamically compensating the initial load of the drawing machine through multi-source heterogeneous sensors, and then reconstructing the fiber morphology distribution matrix through active and reactive power fluctuation analysis and first-order difference lookup table. Based on this, the genetic algorithm task sequence is optimized with the rework frequency prediction value as a constraint, thus achieving accurate prediction of quality defects and rework frequency before product sampling inspection.

[0091] The second embodiment of the present invention provides an intelligent scheduling system for a two-component fiber textile process, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the method described above.

[0092] It should be noted that the intelligent scheduling system for bicomponent fiber textile processes provided in this embodiment of the invention is used to execute all the process steps of the intelligent scheduling method for bicomponent fiber textile processes described in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0093] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0094] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for intelligent scheduling of a two-component fiber textile process, characterized in that, include: Obtain the workshop task allocation sequence and the physical operating parameters of multiple drawing machines, and perform dynamic load compensation calculation based on the physical operating parameters to obtain the initial machine load value; Based on the initial machine load values, energy consumption characteristics are analyzed to obtain multi-dimensional energy consumption data; Time-series trend analysis is performed on the multidimensional energy consumption data to obtain current change characteristics. Based on the current change characteristics, fiber physical indicators are reconstructed to obtain a fiber morphology distribution matrix. Based on the fiber morphology distribution matrix, analyze the batches with abnormal morphology to obtain the statistical results of unqualified batches; Based on the statistical results of the non-conforming batches, the production line number is matched, and the pre-stored historical frequency records are retrieved based on the production line number to perform trend extrapolation prediction, thereby obtaining the cumulative rework frequency prediction value. Using the predicted cumulative rework frequency as a constraint, the workshop task allocation sequence is iteratively optimized to generate the target process execution sequence.

2. The intelligent scheduling method for bicomponent fiber textile processes according to claim 1, characterized in that, The process of acquiring the workshop task allocation sequence and the physical operating parameters of multiple drawing machines, and performing dynamic load compensation calculations based on the physical operating parameters to obtain the initial machine load values, includes: Analyze the workshop task allocation sequence to obtain processing time and intensity values; Real-time rotational speed and tension fluctuation values ​​are extracted from the physical operating parameters. If the tension fluctuation value exceeds the preset tension fluctuation threshold, the load attenuation rate is obtained by dynamic mapping based on the real-time rotational speed. The initial machine load value is obtained by predicting the load based on the processing time, the intensity value, and the load attenuation rate.

3. The intelligent scheduling method for bicomponent fiber textile processes according to claim 1, characterized in that, The step of analyzing energy consumption characteristics based on the initial machine load value to obtain multi-dimensional energy consumption data includes: Based on the initial machine load value, the heat loss is assessed to obtain the equipment depreciation assessment value; Based on the equipment depreciation assessment value, energy consumption fluctuations are mapped to obtain multidimensional energy consumption data.

4. The intelligent scheduling method for bicomponent fiber textile processes according to claim 1, characterized in that, The process involves performing time-series trend analysis on the multidimensional energy consumption data to obtain current variation characteristics, and then reconstructing fiber physical indicators based on these current variation characteristics to obtain a fiber morphology distribution matrix, including: The fluctuation amplitude sequences of active power and reactive power are extracted from the multidimensional energy consumption data and first-order difference calculation is performed to obtain the active power change rate sequence and the reactive power change rate sequence. The active power rate sequence and the reactive power rate sequence at the same sampling time are combined into a two-dimensional current change feature vector to obtain the current change characteristics of the drawing machine under continuous operation. Based on the current change characteristics, a pre-established feature-index mapping table is queried to obtain multiple physical index values; The values ​​of multiple physical indicators are arranged in a predetermined order to obtain a fiber morphology distribution matrix that includes the values ​​of thickness uniformity.

5. The intelligent scheduling method for bicomponent fiber textile processes according to claim 1, characterized in that, The step of analyzing morphologically abnormal batches based on the fiber morphology distribution matrix to obtain statistical results of unqualified batches includes: Extract the thickness uniformity value from the fiber morphology distribution matrix. If the thickness uniformity value is lower than the preset uniformity threshold, obtain the corresponding abnormal batch traceability code. Based on the traceability code of the abnormal batch, the cumulative frequency increment statistics are performed to obtain the statistical results of the unqualified batches.

6. The intelligent scheduling method for bicomponent fiber textile processes according to claim 1, characterized in that, The step of matching the production line number based on the statistical results of the non-conforming batches, and retrieving pre-stored historical frequency records based on the production line number to perform trend extrapolation prediction to obtain the cumulative rework frequency prediction value includes: If the statistical result of the non-conforming batches is equal to zero, then the predicted value of the cumulative rework frequency is set to zero; If the statistical result of the non-conforming batch is greater than zero, the defect type identifier is extracted based on the abnormal batch traceability code corresponding to the statistical result of the non-conforming batch and matched with the corresponding production line number. Based on the production line number and the defect type identifier, read the historical rework frequency sequence from the pre-stored historical frequency record; The historical rework frequency sequence is extrapolated to predict the predicted value for the next statistical period. The predicted value is then rounded to obtain the cumulative rework frequency prediction value.

7. The intelligent scheduling method for bicomponent fiber textile processes according to claim 1, characterized in that, The step of using the predicted cumulative rework frequency as a constraint to perform sequence iterative optimization on the workshop task allocation sequence to generate the target process execution sequence includes: Using the predicted cumulative rework frequency as a constraint, the workshop task allocation sequence is subjected to population initialization processing to obtain the initial allocation population. The fitness of the initially assigned population is evaluated to obtain fitness values; The initial population is iteratively evolved based on the fitness value to generate the target process execution sequence.

8. The intelligent scheduling method for bicomponent fiber textile processes according to claim 2, characterized in that, The physical operating parameters include the real-time rotational speed, tension fluctuation value, surface temperature data, and operating frequency value of the drawing machine.

9. An intelligent scheduling system for a bicomponent fiber textile process, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the method described in any one of claims 1 to 8.

Images