Textile special equipment manufacturing process data analysis method and system

Abstract

The present application relates to the technical field of manufacturing data analysis, in particular to a textile special equipment manufacturing process data analysis method and system, comprising the following steps: collecting temperature feedback and quality index record to generate a basic sample set, scanning according to effective temperature characteristics to obtain continuity mutation and low value platform sequence to locate the physical failure boundary section, dividing the sample according to the boundary to determine the upper limit position of process risk combined with the quality threshold, generating a constrained temperature recommendation interval and analysis results based on the correlation modeling of the samples before the boundary, in the present application, the failure boundary after mutation tends to a low value platform is accurately located by calculating the regression slope and obtaining the characteristic sequence, the sample area is divided and the abnormal data exceeding the limit is removed to avoid indiscriminate mixed mining by statistical algorithm, the model is established in the effective physical interval of the material to eliminate the pseudo correlation error caused by high temperature damage, and the output recommendation interval provides the production process adjustment benchmark.

Description

Technical Field

[0001] This invention relates to the field of manufacturing data analysis technology, and in particular to a method and system for analyzing manufacturing process data of textile-specific equipment. Background Technology

[0002] In the data analysis stage of textile special equipment manufacturing, clarifying the relationship between key process parameters and final product performance is the foundation for resource allocation and process optimization. Existing technologies mainly collect process parameters and output quality indicators at different process nodes, and establish continuous correlation models through data mining. For example, analyzing the relationship between the temperature controller setpoint of heat treatment equipment and the yarn strength after subsequent cooling in order to find a suitable temperature control range. In long-term actual production records, in addition to normal process fluctuations, there are usually abnormal high temperature data that exceed the normal process tolerance limit due to temporary loss of equipment temperature control, operational errors, or extreme testing.

[0003] Existing analytical processes for establishing correlation models typically assume that all samples in the dataset follow a uniform physical mechanism, lacking prior determination of material physical failure boundaries. This operational mode of indiscriminately incorporating collected data into the analysis process directly merges conventional parameters with abnormal parameters exceeding permissible limits. When the analysis scope covers high-temperature regions exceeding process limits, changes such as thermal decomposition of materials occur, causing normal process correlation mechanisms to fail. Because yarn strength is generally lower and fluctuates more slowly at high temperatures due to material damage, statistical algorithms will fit the data characteristics within this region. This can lead to the model misinterpreting the destructive consequences of parameter changes as weak correlations between data, causing the output process judgments to deviate from the actual physical mechanism. This can mislead the direction of process optimization during resource allocation and affect the accuracy of production quality control. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology, and to this end, a data analysis method and system for the manufacturing process of textile special equipment is proposed.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a data analysis method for the manufacturing process of textile-specific equipment, comprising the following steps:

[0006] S1: Collect temperature feedback records, traction roller linear speed records, yarn breaking strength records, sensor status records, and data integrity records. Retrieve production batch identifiers and material specification identifiers. Align and aggregate the temperature feedback records to obtain effective temperature features. Bind quality indicators according to production batch identifiers to obtain a basic analysis sample set containing material specification identifiers, sensor status records, and data integrity records.

[0007] S2: Based on the aforementioned basic analysis sample set, group the samples according to material specification identifiers, remove invalid samples, and set observation marks for samples exceeding the limits to obtain the variety analysis sample set;

[0008] S3: For the sample set of the variety analysis, perform an ascending scan according to the effective temperature characteristics, calculate the regression slope within the window, obtain the continuous feature sequence based on the regression slope sequence within the window, obtain the mutation feature sequence based on the change of slope sign and negative absolute value of the slope of the regression slope sequence within the window, calculate the mean and standard deviation of the quality indicators within the window, and obtain the low-value plateau feature sequence.

[0009] S4: Based on the continuous feature sequence, the abrupt change feature sequence and the low-value platform feature sequence, locate the position of the abrupt change tending to the low-value platform, obtain the boundary positioning temperature value, obtain the thermal decomposition temperature reference value and the historical failure threshold, and verify the boundary positioning temperature value to determine the physical failure boundary segment.

[0010] S5: Based on the physical failure boundary segment, the sample set of the variety analysis is divided into a sample set before the boundary, a sample set near the boundary, and a sample set after the boundary. The quality qualification threshold is obtained. Based on the sample set near the boundary and the quality qualification threshold, the upper limit position of the process risk is obtained. An over-temperature failure cause identifier is set for the sample set after the boundary.

[0011] S6: Based on the pre-boundary sample set, perform correlation modeling to obtain an effective correlation model and an effective temperature feature recommendation interval. Based on the physical failure boundary segment and the upper limit position of the process risk, obtain the constrained effective temperature feature recommendation interval, and filter out the analysis samples with the over-temperature failure cause identifier to generate manufacturing process data analysis results.

[0012] As a further aspect of the present invention, the process of S1 is specifically as follows:

[0013] S111: Collect temperature feedback records of each temperature zone of the heat treatment equipment, traction roller linear speed records, sensor status records, data integrity records, and yarn breaking strength records of cooled yarn samples, and retrieve production batch identifiers and material specification identifiers.

[0014] S112: Determine the material heating time based on the traction roller linear speed record, and perform time-series alignment and weighted aggregation of the temperature feedback record of the temperature zone according to the material heating time to generate effective temperature characteristics;

[0015] S113: Call the effective temperature feature, use the yarn breaking strength record as a quality indicator, bind the effective temperature feature and the quality indicator to the production batch according to the production batch identifier, and associate the material specification identifier, the sensor status record and the data integrity record to the bound analysis sample according to the production batch identifier to obtain a basic analysis sample set containing the material specification identifier, sensor status record and data integrity record.

[0016] As a further aspect of the present invention, the process of performing time-series alignment and weighted aggregation of the temperature feedback records of the temperature zone according to the heating time of the material specifically includes:

[0017] The heating time of the material is calculated based on the linear speed record of the traction roller and the preset physical length of the heating box. The heating time period record matching the production batch identifier is extracted from the temperature feedback record of the temperature zone according to the production batch identifier. The heating time period record is weighted and summed according to the preset temperature zone weight to generate the effective temperature feature.

[0018] As a further aspect of the present invention, the process of S2 is specifically as follows:

[0019] S211: Call the basic analysis sample set, group the basic analysis sample set according to the material specification identifier to obtain the variety group sample set, and based on the variety group sample set, remove null value samples, zero value samples and garbled samples according to the sensor status record and data integrity record to obtain the effective sample set;

[0020] S212: Call the effective sample set, set observation marks for analysis samples whose effective temperature characteristics exceed the preset production setting range and whose data integrity records and sensor status records meet the preset acquisition conditions, and obtain the variety analysis sample set.

[0021] As a further aspect of the present invention, the process of S3 is specifically as follows:

[0022] S311: Call the variety analysis sample set, arrange the analysis samples in ascending order according to the effective temperature characteristics, keep the analysis samples corresponding to the observation marks within the sliding window scanning range, calculate the regression slope between the effective temperature characteristics and quality indicators within the window, and obtain the regression slope sequence within the window.

[0023] S312: Obtain a continuous feature sequence based on the regression slope sequence within the window, and obtain a mutation feature sequence based on the change in the slope sign and the negative absolute value of the slope of the regression slope sequence within the window;

[0024] S313: For the sliding window, calculate the mean and standard deviation of the quality indicators within the window, and obtain the low-value platform feature sequence based on the mean and standard deviation of the quality indicators within the window.

[0025] As a further aspect of the present invention, the process of S4 is specifically as follows:

[0026] S411: Call the continuous feature sequence, the abrupt change feature sequence and the low value plateau feature sequence, and filter continuous stable segments with consistent slope signs and absolute values ​​of adjacent slope differences less than a preset continuity threshold according to the continuous feature sequence;

[0027] S412: Extract the segment after the continuous stable segment, and filter the mutation window with a slope that changes from positive to negative and the absolute value of the negative slope is greater than the preset process fluctuation slope threshold according to the mutation feature sequence. Filter the low-value platform window with a mean quality index less than the preset low-value quality threshold and a standard deviation of the quality index less than the preset fluctuation threshold according to the low-value platform feature sequence.

[0028] S413: Call the mutation window and the low-value platform window. When the low-value platform window exists after the mutation window, obtain the boundary positioning temperature value according to the starting temperature of the mutation window, obtain the thermal decomposition temperature reference value and the historical failure threshold, calculate the absolute value of the difference between the boundary positioning temperature value and the thermal decomposition temperature reference value to obtain the first temperature deviation, and calculate the absolute value of the difference between the boundary positioning temperature value and the historical failure threshold to obtain the second temperature deviation.

[0029] S414: When either the first temperature deviation or the second temperature deviation is within a preset engineering error range, the starting temperature of the sudden change window is taken as the starting temperature of the physical failure boundary section, and the starting temperature of the low value platform window is taken as the ending temperature of the physical failure boundary section. Based on the starting temperature and ending temperature of the physical failure boundary section, a physical failure boundary section is established.

[0030] As a further aspect of the present invention, the process of S5 is specifically as follows:

[0031] S511: Invoke the physical failure boundary section and the variety analysis sample set;

[0032] The analysis samples with effective temperature characteristics lower than the starting temperature of the physical failure boundary section are divided into the pre-boundary sample set.

[0033] The analysis samples with effective temperature characteristics that are not less than the starting temperature of the physical failure boundary section and not greater than the ending temperature of the physical failure boundary section are divided into a sample set near the boundary.

[0034] The analytical samples whose effective temperature characteristics are greater than the termination temperature of the physical failure boundary section are divided into a post-boundary sample set.

[0035] S512: Call the sample set near the boundary and the quality qualification threshold, calculate the probability distribution of the quality index in the sample set near the boundary being lower than the quality qualification threshold, select the upper limit of the effective temperature feature from the probability distribution according to the preset safety margin, obtain the upper limit position of the process risk, and set the over-temperature failure cause identifier for the analysis samples in the sample set after the boundary.

[0036] As a further aspect of the present invention, the process of S6 is specifically as follows:

[0037] S611: Call the pre-boundary sample set, remove outlier samples from the process, perform weighted regression fitting on the effective temperature features and quality indicators, establish an effective correlation model when the model parameters converge, and obtain the recommended interval of the effective temperature features.

[0038] S612: Determine the physical effective analysis interval based on the physical failure boundary segment, and impose an upper limit constraint on the recommended effective temperature feature interval based on the upper limit position of the process risk, to obtain the constrained recommended effective temperature feature interval;

[0039] S613: Call the effective correlation model, the recommended range of constrained effective temperature features, the physical effective analysis range, and the over-temperature failure cause identifier, write the over-temperature failure cause identifier into the correlation modeling sample screening conditions, filter out the analysis samples with the over-temperature failure cause identifier according to the correlation modeling sample screening conditions, and generate manufacturing process data analysis results.

[0040] As a further aspect of the present invention, the process of performing weighted regression fitting on the effective temperature characteristics and quality indicators, establishing an effective correlation model when the model parameters converge, and obtaining the recommended interval for the effective temperature characteristics specifically involves:

[0041] A preset initial weight is set for the analysis sample after removing the outlier samples of the process. The regression curve between the effective temperature feature and the quality index is fitted according to the preset initial weight. The quality index residual of the analysis sample is calculated. The absolute value of the quality index residual is compared with the preset residual threshold. The sample weight is updated according to the preset weight update table.

[0042] The regression curve is fitted again based on the updated sample weights. When the absolute value of the difference between the corresponding model parameters of two adjacent regression curves is less than the corresponding preset convergence threshold, the effective correlation model is established.

[0043] Based on the effective temperature feature range corresponding to the quality index values ​​in the effective correlation model, the recommended interval for the effective temperature feature is obtained.

[0044] A data analysis system for the manufacturing process of textile special equipment, the system being used to execute the aforementioned data analysis method for the manufacturing process of textile special equipment, the system comprising:

[0045] Temperature zone data integration module: Collects temperature feedback records, traction roller linear speed records, yarn breaking strength records, sensor status records, and data integrity records; retrieves production batch identifiers and material specification identifiers; aligns and aggregates temperature feedback records to obtain effective temperature features; and binds quality indicators according to production batch identifiers to obtain a basic analysis sample set containing material specification identifiers, sensor status records, and data integrity records.

[0046] Variety cleaning and marking module: Based on the basic analysis sample set, the sample set is grouped according to material specification identification, invalid samples are removed, and observation marks are set for samples exceeding the limit to obtain the variety analysis sample set;

[0047] Degradation Feature Identification Module: For the sample set of the variety analysis, the module performs an ascending scan based on the effective temperature feature, calculates the regression slope within the window, obtains the continuous feature sequence based on the regression slope sequence within the window, obtains the abrupt change feature sequence based on the change in the slope sign and the negative absolute value of the slope, calculates the mean and standard deviation of the quality indicators within the window, and obtains the low-value plateau feature sequence.

[0048] Failure boundary verification module: Based on the continuous feature sequence, the abrupt change feature sequence and the low value plateau feature sequence, locate the position of the abrupt change tending to the low value plateau, obtain the boundary location temperature value, acquire the thermal decomposition temperature reference value and historical failure threshold, verify the boundary location temperature value, and determine the physical failure boundary segment;

[0049] Risk zoning and identification module: Based on the physical failure boundary segment, the sample set of the product analysis is divided into a sample set before the boundary, a sample set near the boundary, and a sample set after the boundary. The quality qualification threshold is obtained. Based on the sample set near the boundary and the quality qualification threshold, the upper limit position of the process risk is obtained. The over-temperature failure cause identification is set for the sample set after the boundary.

[0050] Modeling and Interval Recommendation Module: Based on the pre-boundary sample set, perform correlation modeling to obtain an effective correlation model and effective temperature feature recommendation intervals. Based on the physical failure boundary segment and the upper limit position of the process risk, obtain constrained effective temperature feature recommendation intervals, and filter out analysis samples with the over-temperature failure cause identifier to generate manufacturing process data analysis results.

[0051] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0052] In this invention, for the analysis sample generated by aggregating temperature feedback and quality indicators, the regression slope within the calculation window is obtained by ascending scan to obtain continuity and abrupt change characteristics. The low-value plateau features are extracted by combining the mean and standard deviation. Multidimensional features are used to accurately locate the position of the low-value plateau after abrupt change, thereby locking the physical failure boundary segment. Based on this boundary, the sample is divided into different regions and the upper limit of process risk is determined by combining the quality threshold. Abnormal data exceeding the limit is eliminated to avoid indiscriminate mixing and mining by statistical algorithms. This ensures that a correlation model is established within the effective range of the material to eliminate spurious correlation errors caused by damage. The constrained temperature recommendation range is output and a reliable process adjustment benchmark is provided. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the manufacturing process data analysis flow of the present invention;

[0054] Figure 2 This is a schematic diagram of the manufacturing process data analysis workflow of the present invention;

[0055] Figure 3 This is a schematic diagram illustrating the failure boundary location and verification principle of the present invention;

[0056] Figure 4 This is a schematic diagram illustrating the principle of constrained interval recommendation in this invention;

[0057] Figure 5 This is a schematic diagram of the manufacturing process data analysis system for textile-specific equipment of the present invention. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0059] This embodiment establishes a unified data analysis chain based on heat treatment data, traction data, and cooled yarn quality data during the manufacturing process of textile-specific equipment. Temperature feedback records are time-series records generated by the feedback from each temperature zone of the heat treatment equipment during the production batch operation, used to describe the thermal effects on the material as it passes through different temperature zones; traction roller linear speed records are the running speed records of the traction rollers within the same production batch, used to determine the continuous state of the material's residence and heating within the heating chamber; yarn breaking strength records are quality data generated by strength testing of cooled yarn samples, used as a quality indicator in this embodiment; sensor status records indicate whether the sensors in the acquisition chain are in a usable state; data integrity records indicate whether there are missing, null, garbled, or incomplete data in the same production batch; production batch identifiers are used to correlate process data and quality data within the same batch; material specification identifiers are used to distinguish different varieties or specifications of yarn materials, ensuring that the thermal effects and quality responses under different material specifications are not mixed in the analysis.

[0060] In this embodiment, the preset heating chamber physical length, preset temperature zone weights, preset production setting range, preset data acquisition conditions, preset continuity threshold, preset process fluctuation slope threshold, preset low-value quality threshold, preset fluctuation threshold, preset engineering error range, quality qualification threshold, preset safety margin, preset initial weight, preset residual threshold, preset weight update table, and preset convergence threshold are all obtained from process configuration records, equipment structure records, material specification association records, quality judgment records, historical failure records, or data acquisition rule records before the analysis is performed, and a correspondence is established with the production batch identifier and material specification identifier. The preset heating chamber physical length comes from the heat treatment equipment structure record or assembly verification record, and is used to determine the material heating time together with the traction roller linear speed record; the preset temperature zone weight comes from the temperature zone layout, material heating path, and process configuration records, and is used to explain the contribution relationship of different temperature zones in the comprehensive thermal effect; various thresholds and judgment conditions come from the process configuration record, quality judgment record, and historical failure record corresponding to the material specification, and are used to complete the screening, comparison, verification, and convergence judgment in different steps. The above records only provide judgment rules and processing basis, and no specific values ​​are introduced in the text.

[0061] Please see Figure 1 and Figure 2S1: Collect temperature feedback records from the temperature zones, traction roller linear speed records, yarn breaking strength records, sensor status records, and data integrity records. Retrieve production batch identifiers and material specification identifiers. Align and aggregate the temperature feedback records from the temperature zones to obtain effective temperature characteristics. Bind quality indicators according to the production batch identifier to obtain a basic analysis sample set containing material specification identifiers, sensor status records, and data integrity records. The purpose of this step is to organize the data scattered in the heat treatment equipment, traction mechanism, quality inspection links, and collected status records into comparable analysis samples within the same production batch. Since the yarn material is not statically heated in a single temperature zone during heat treatment, but passes through multiple temperature zones sequentially as the traction roller moves, using only the instantaneous feedback temperature of a certain temperature zone cannot represent the comprehensive thermal effects experienced by the batch of material. The heating time of the material is determined by recording the linear speed of the traction roller and the preset physical length of the heating chamber. Then, the temperature feedback records of the temperature zones are time-sequentially aligned according to the heating time, ensuring that the temperature records correspond to the actual time the material passes through the heating chamber. Furthermore, the matched heating time records of each temperature zone are weighted and aggregated to form an effective temperature characteristic describing the overall thermal effect level of the batch of material. In this embodiment, the effective temperature characteristic is a batch-level temperature characteristic jointly determined by the temperature feedback records of the temperature zones, the linear speed records of the traction roller, the preset physical length of the heating chamber, and the preset temperature zone weights. It is not a single measurement point temperature, but rather an analytical input formed after the heating time period is truncated, time-sequentially aligned, and weighted by the temperature zones. Subsequently, the effective temperature characteristic is bound to the yarn breaking strength record after cooling according to the production batch identifier, establishing a correspondence between the process input and quality output of the same batch. The material specification identifier, sensor status records, and data integrity records are also synchronously associated with this analytical sample. If, during the data collection process, batch identifiers are missing, material specification identifiers cannot be retrieved, temperature records and traction records cannot be matched by batch, quality records cannot be traced back to the same production batch, sensor status records cannot be parsed, or data integrity records indicate incomplete data, then the sample will not proceed directly to subsequent feature identification. Instead, its status information will be retained in the sample for the next step of cleaning and judgment, thus avoiding data of unknown origin or with invalid matching relationships being mistakenly identified as valid samples.

[0062] S111: Collect temperature feedback records for each temperature zone of the heat treatment equipment, traction roller linear speed records, sensor status records, data integrity records, and yarn breaking strength records from cooled yarn samples. Retrieve production batch identifiers and material specification identifiers. This sub-step first establishes the data source for the same production batch. Temperature feedback records carry the feedback temperature status of the heat treatment equipment in temperature zone and time sequence; traction roller linear speed records carry the conveying status of materials through the heating chamber; sensor status records carry the status of temperature, traction, and other acquisition terminals to ensure effective operation; data integrity records carry the inspection results of whether the collected data is complete and whether there are any null values ​​or abnormal formats; yarn breaking strength records are from cooled yarn samples, representing the quality response after heat treatment. Production batch identifiers serve as batch indexes, placing process data and quality inspection data in the same matching link; material specification identifiers serve as variety indexes for subsequent grouping by material specification. During data collection, first check whether each record has resolvable batch association information, and then check whether each record is in a readable state. For data that cannot be parsed, cannot be matched, or whose collection status is uncertain, it is not deleted directly in this step, but is entered into the basic sample formation process along with the sensor status record and data integrity record, so that the subsequent cleaning steps can be processed based on the clear status.

[0063] S112: Determine the material heating time based on the traction roller linear speed record. Perform time-series alignment and weighted aggregation of the temperature feedback records for the temperature zone according to the material heating time to generate effective temperature characteristics. The traction roller linear speed record and the preset physical length of the heating chamber jointly determine the duration of the material's thermal effect in the heating chamber. The faster the material moves, the shorter the dwell time over the same heating chamber length; the slower the material moves, the longer the dwell time. The preset physical length of the heating chamber is obtained from the equipment structure record or assembly verification record and corresponds to the production batch record of the current heat treatment equipment before use. This sub-step first determines the material heating time based on the traction roller linear speed record and the preset physical length of the heating chamber, and then extracts the heating period record matching the batch from the temperature feedback temperature record according to the production batch identifier. The extraction process means selecting only the temperature feedback segment where the material is actually within the heating chamber's thermal effect range, excluding temperature records before the start of the batch, after the end of the batch, or those not belonging to the batch's heating process. Subsequently, the heating period records are weighted and summed according to the preset temperature zone weights. The preset temperature zone weights are obtained from the process configuration records corresponding to the material specifications, and are used to reflect the differences in the thermal impact of different temperature zone locations or stages on the material. After weighted aggregation, the effective temperature feature is obtained, which serves as the common temperature input for subsequent sample grouping, degradation scanning, boundary location, and interval recommendation. If the temperature feedback record of the temperature zone contains unmatched segments within the heating period, or if the traction roller linear speed record cannot support the determination of the heating duration, the effective temperature feature of that batch will not be used as a reliable input, but will instead be combined with data integrity records and sensor status records for subsequent validity screening.

[0064] S113: The effective temperature feature is invoked, and the yarn breaking strength record is used as a quality indicator. Based on the production batch identifier, the effective temperature feature and quality indicator are batch-bound. Then, the material specification identifier, sensor status record, and data integrity record are associated with the bound analysis sample according to the production batch identifier, resulting in a basic analysis sample set containing the material specification identifier, sensor status record, and data integrity record. In this embodiment, the quality indicator is the yarn breaking strength record of the cooled yarn sample, used to represent the quality result after heat treatment. The batch binding process is as follows: first, the batch corresponding to the effective temperature feature is retrieved by the production batch identifier; then, the quality indicator of the same batch is retrieved; and finally, the material specification identifier, sensor status record, and data integrity record are linked to the same analysis sample. This forms a basic analysis sample set containing temperature input, quality output, material specifications, and data reliability status. Subsequent steps can complete grouping, cleaning, degradation feature identification, and boundary analysis within the same data structure. If multiple source records appear in the same production batch but a unique correspondence cannot be determined, the traceable batch relationship is retained, and samples that do not meet the complete matching condition are handled by subsequent rejection rules to prevent incorrect binding from causing a mismatch between the temperature feature and the quality indicator.

[0065] Please see Figure 1 and Figure 2S2: Based on the basic analysis sample set, group samples according to material specification identifiers, remove invalid samples, and set observation marks for samples exceeding limits to obtain the product analysis sample set. The purpose of this step is to process samples corresponding to different material specifications separately and remove invalid data without disrupting the continuity of boundary analysis. Different material specifications have different temperature responses and quality changes during heat treatment. If different specifications are mixed in the same analysis sequence, the relationship between temperature characteristics and quality indicators will be interfered with by product differences. Therefore, the basic analysis sample set is first split into product group sample sets according to material specification identifiers. Then, null value samples, zero value samples, and garbled samples are identified based on sensor status records and data integrity records. Null value samples indicate that the fields required for analysis are missing, zero value samples indicate invalid values ​​appearing under conditions that do not conform to the actual acquisition logic, and garbled samples indicate that the record format cannot be parsed normally. These types of samples cannot reliably participate in the judgment of regression slope, quality mean, or quality fluctuation, and are therefore removed. For samples whose effective temperature characteristics exceed the preset production setting range but whose sensor status records and data integrity records meet the preset acquisition conditions, they are not directly removed, but observation marks are set. The preset production setting range is obtained from the process configuration record corresponding to the material specification, and is used to describe the temperature setting boundary of the material specification under the current production conditions. The preset acquisition conditions are obtained from the acquisition rule record, and are used to determine whether the sensor status record and data integrity record support the sample to continue participating in the analysis. The observation mark is the status information attached to the analysis sample, which is used to indicate that the sample is outside the production setting range, but the acquisition link is valid and the data structure is complete. Retaining this type of sample allows subsequent sliding window scans to cover sections where thermal failure boundaries may occur, and at the same time, the mark distinguishes its status from that of samples within the normal production range. If the data integrity record or sensor status record does not meet the acquisition conditions, the out-of-limit sample will not be marked with an observation mark and will be treated as an invalid sample to avoid misjudging acquisition anomalies as actual process over-limits.

[0066] S211: The basic analysis sample set is retrieved and grouped according to material specification identifiers to obtain product-grouped sample sets. Based on these product-grouped sample sets, null, zero, and garbled samples are removed according to sensor status records and data integrity records to obtain a valid sample set. This sub-step first uses material specification identifiers as the grouping basis, ensuring that valid temperature characteristics and quality indicators within the same material specification are included in the same analysis set. The product-grouped sample set is a sample set formed according to material specifications. Samples within each set have the same material specification identifier and retain the corresponding temperature characteristics, quality indicators, and acquisition status information. When removing invalid samples, not only are field contents checked for null or unparseable values, but sensor status records are also used to determine the reliability of the data source, and data integrity records are used to determine the completeness of the required data for the same batch. For null samples, the correlation between temperature and quality cannot be calculated subsequently; for zero-value samples, deviations from the actual heat treatment and quality inspection process will occur; for garbled samples, their actual meaning cannot be confirmed. Therefore, this sub-step forms a valid sample set, ensuring that subsequent scanning and modeling are performed only on parsable, matchable data with a complete record basis.

[0067] S212: Call the valid sample set. For analytical samples whose effective temperature characteristics exceed the preset production setting range, but whose data integrity records and sensor status records meet the preset acquisition conditions, set observation markers to obtain the product analysis sample set. The preset production setting range is obtained from the process configuration record corresponding to the current material specification, and the preset acquisition conditions are obtained from the acquisition rule record. An effective temperature characteristic exceeding this range does not necessarily indicate invalid data, because when the sensor status is normal and the data is complete, this out-of-range state may reflect that the material has actually experienced thermal effects outside the production setting range. This sub-step first determines whether the effective temperature characteristic exceeds the preset production setting range, and then determines whether the data integrity records and sensor status records meet the preset acquisition conditions. Observation markers are only set for analytical samples when both the out-of-range condition and the acquisition reliability are met. Observation markers do not change the temperature characteristics and quality indicators of the samples; they only serve as a basis for retention during subsequent sliding window scanning, ensuring that data near the failure zone during heat treatment is not prematurely excluded during the cleaning stage. After marking, the product analysis sample set is obtained, which is then called by S3 for ascending-order scanning of effective temperature characteristics.

[0068] Please see Figure 1 and Figure 2S3: For the sample set of product analysis, perform an ascending scan based on effective temperature characteristics, calculate the regression slope within the window, obtain the continuous characteristic sequence based on the regression slope sequence within the window, and obtain the abrupt change characteristic sequence based on the change in the sign of the slope and the negative absolute value of the slope. Calculate the mean and standard deviation of the quality indicators within the window to obtain the low-value plateau characteristic sequence. The purpose of this step is to identify the continuous trend, abrupt trend, and low-value plateau state of quality indicators with effective temperature characteristics from the samples arranged by temperature. The ascending scan of effective temperature characteristics arranges the analytical samples under the same material specification from low to high according to the intensity of thermal effect, so that subsequent sliding windows can observe quality changes segment by segment along the temperature direction. The regression slope within the window indicates the direction and degree of change of the quality indicators with effective temperature characteristics within the local temperature range covered by the current sliding window. When the slope sign remains consistent and the change is gradual, it indicates that the quality response within the local interval has a continuous changing state. When the slope changes from positive to negative and the absolute value of the negative direction meets the preset judgment condition, it indicates that the quality response has shifted from an upward or stable relationship to a downward relationship, possibly entering a sudden stage of quality degradation caused by thermal effects. The mean of the quality index within the window is used to describe the concentration state of the quality level within the local interval, and the standard deviation of the quality index within the window is used to describe the fluctuation state of the quality index within the interval. When the mean of the quality index is at the preset low-value quality judgment state and the fluctuation meets the preset stable judgment state, the interval forms a low-value plateau characteristic. The continuous feature sequence, the sudden change feature sequence, and the low-value plateau feature sequence are jointly passed to S4 to locate the position of the low-value plateau after the mutation. If a certain window in the variety analysis sample set lacks effective samples that can be used for regression or quality statistics, the window will not output effective features. The scanning process continues to advance in the order of effective temperature features, and the unusable state of the window is retained to prevent subsequent misidentification of the breakpoint formed by the missing feature as a real mutation.

[0069] S311: The sample set for variety analysis is retrieved, and the samples are arranged in ascending order of effective temperature characteristics. The samples corresponding to the observation markers are kept within the sliding window scanning range. The regression slope between the effective temperature characteristics and quality indicators within the window is calculated to obtain the regression slope sequence within the window. This sub-step first sorts the samples under the same material specification from low to high effective temperature characteristics, ensuring the scanning direction aligns with the direction of increased thermal effect. Although the samples corresponding to the observation markers exceed the preset production setting range, their acquisition status and data integrity have already been confirmed in S2, so they are retained within the sliding window scanning range to observe whether quality degradation or failure boundaries occur outside the production setting range. The sliding window covers adjacent samples after sorting, and the corresponding changes in effective temperature characteristics and quality indicators are compared within each window to obtain the regression slope within the window. In this embodiment, the regression slope represents the trend of quality indicators relative to effective temperature characteristics within a local temperature range. A positive slope indicates that the quality indicators increase with increasing temperature characteristics, a negative slope indicates that the quality indicators decrease with increasing temperature characteristics, and a near-stable slope indicates that the quality response changes slowly within the local range. As the window moves along the ascending sample sequence, the regression slopes within multiple windows are arranged in temperature order, forming a sequence of regression slopes within the window. This sequence is used by S312 to determine continuity and abrupt changes. If there is a sample within the window whose quality indicators or effective temperature characteristics cannot be read, that sample will not participate in the slope formation of that window, and its abnormal state will be retained in the sample record.

[0070] S312: Obtain the continuous feature sequence based on the regression slope sequence within the window, and obtain the abrupt change feature sequence based on the change in slope sign and the absolute value of the negative slope of the regression slope sequence within the window. The continuous feature sequence records whether the slope change is smooth between adjacent windows. It is generated based on whether the slope sign is consistent and whether the difference between adjacent slopes meets the judgment condition corresponding to the preset continuity threshold. It is used to identify whether the quality response maintains continuous change in the ascending direction of the effective temperature characteristic. The preset continuity threshold is obtained from the process configuration record or historical analysis rules corresponding to the material specifications and is used to distinguish between continuous changes and relatively obvious local changes. The abrupt change feature sequence records the change in slope sign and the degree of negative change. It focuses on whether the slope changes from positive to negative and whether the absolute value of the negative change meets the judgment condition corresponding to the preset process fluctuation slope threshold. The preset process fluctuation slope threshold is obtained from the process configuration record and is used to exclude slope changes caused by ordinary process fluctuations. A change in slope sign indicates a change in the direction of the local relationship, and the absolute value of the negative change meeting the judgment condition indicates that the decline in quality indicators has a identifiable magnitude. By generating continuous and abrupt features separately, the subsequent S4 can first find the stable basic segment, and then look for the abrupt window where the quality changes from stable or improving to deteriorating, thus avoiding taking local noise directly as the failure boundary.

[0071] S313: For the sliding window, calculate the mean and standard deviation of the quality indicators within the window. Based on these values, obtain the low-value plateau feature sequence. The mean of the quality indicators represents the concentrated state of the quality level of the samples covered by the sliding window, while the standard deviation represents the discrete state of the quality results within the window. This sub-step performs qualitative statistics on the concentrated and fluctuating states of the quality indicators within each sliding window, and then identifies the low-value plateau characteristics according to the judgment conditions corresponding to the preset low-value quality threshold and preset fluctuation threshold. The preset low-value quality threshold is obtained from the quality judgment record and is used to determine whether the quality level within the window has entered a low-value state; the preset fluctuation threshold is obtained from the process configuration record or the quality fluctuation judgment rule and is used to determine whether the low-value state has local stability. A low-value plateau is not a single low-quality sample, but a locally stable state formed by a segment of quality indicators within a window that are in a low-value state and have small fluctuations. This processing can distinguish between occasional quality declines and persistent low-value plateaus, because occasional declines are often accompanied by fluctuations within the window, while low-value plateaus simultaneously exhibit low quality levels and limited local fluctuations. The low-value platform feature sequences are arranged in ascending order of effective temperature features and are available for S4 to use along with the continuous feature sequences and the abrupt change feature sequences.

[0072] Please see Figure 1 , Figure 2 and Figure 3S4: Based on the continuous characteristic sequence, abrupt change characteristic sequence, and low-value plateau characteristic sequence, locate the position of the low-value plateau after the abrupt change, obtain the boundary location temperature value, acquire the thermal decomposition temperature reference value and historical failure threshold, and verify the boundary location temperature value to determine the physical failure boundary segment. The purpose of this step is to combine the local trend changes in the statistical sequence with the physical meaning of material thermal failure to form a physical failure boundary segment that can be used for subsequent zoning and recommended constraints. The continuous characteristic sequence is used to identify the stable segment of the quality response as the thermal effect gradually increases; the abrupt change characteristic sequence is used to identify the window where the quality response undergoes a directional decline; and the low-value plateau characteristic sequence is used to identify the window where the quality index enters a low and stable state after a decline. The mere appearance of an abrupt change window cannot directly determine the physical failure boundary, because local process fluctuations may also cause slope changes; the mere appearance of a low-value plateau cannot directly determine the boundary either, because the plateau position may come from a stable low-quality state after the failed interval. This step requires that a low-value plateau window exists after the abrupt change window, indicating that the quality index has gone from a stable or improving state to a declining state, and further tends to a low-value stable state. The boundary location temperature value is derived from the starting temperature of the abrupt change window, indicating the location where thermal changes begin to cause a directional change in the quality response. Subsequently, a thermal decomposition temperature reference value and a historical failure threshold are introduced for verification. The thermal decomposition temperature reference value is obtained from the thermal decomposition reference record associated with the material specification, the historical failure threshold is obtained from historical failure records, and the preset engineering error range is obtained from engineering verification rules. These three are used to confirm whether the boundary location temperature value is consistent with physical failure perception or historical failure experience. If the verification does not meet the preset engineering error range, a physical failure boundary segment is not established, and the process reverts to the joint screening process of the abrupt change window and the low-value plateau window, avoiding the solidification of statistical fluctuations without physical correspondence as boundaries.

[0073] S411: Invoke the continuous feature sequence, abrupt change feature sequence, and low-value plateau feature sequence. Based on the continuous feature sequence, filter for continuous stable segments with consistent slope signs and adjacent slope differences less than a preset continuity threshold. A continuous stable segment is a segment where the quality response trend remains consistent in the ascending direction of the effective temperature feature and adjacent changes meet preset continuity criteria. This sub-step first reads the slope sign state of each window in the continuous feature sequence and the slope difference state of adjacent windows, then connects adjacent windows with consistent slope signs and adjacent slope differences within the preset continuity threshold constraint to form a continuous stable segment. This segment provides a reference position for subsequent abrupt change identification because the true failure boundary manifests as a change in trend direction after a relatively stable quality response, rather than arbitrarily selecting a descending window from disordered fluctuations. After the continuous stable segment is formed, its subsequent segments are invoked by S412.

[0074] S412: After extracting the segment following the continuous stable section, select abrupt change windows based on the abrupt change characteristic sequence where the slope changes from positive to negative and the absolute value of the negative slope is greater than a preset process fluctuation slope threshold. Select low-value plateau windows based on the low-value plateau characteristic sequence where the mean quality index within the window is less than a preset low-value quality threshold and the standard deviation of the quality index within the window is less than a preset fluctuation threshold. This sub-step only performs selection after the continuous stable section, ensuring a temporal relationship between the abrupt change windows and the low-value plateau windows: first stable, then abrupt, then stable at low values. The selection criteria for abrupt change windows are based on the abrupt change characteristic sequence; a slope changing from positive to negative indicates a reversal in the local quality response direction, and the absolute value of the negative slope meeting the preset process fluctuation slope threshold indicates that the degree of decline exceeds the judgment boundary of ordinary process fluctuation. The selection criteria for low-value plateau windows are based on the low-value plateau characteristic sequence; the mean quality index within the window meeting the preset low-value quality threshold indicates that the quality level of the window is in a low-value state, and the standard deviation of the quality index within the window meeting the preset fluctuation threshold indicates that the low-value state has local stability. By simultaneously screening mutation windows and low-value plateau windows, simple slope inversions, single-point low values, and persistent failure plateaus can be distinguished. If no mutation window and low-value plateau window that meet the criteria are found simultaneously after a continuous stationary segment, no boundary candidate is output for that segment, and the scanning process continues to check subsequent available segments.

[0075] S413: Invoke the mutation window and low-value plateau window. When a low-value plateau window exists after the mutation window, obtain the boundary positioning temperature value based on the starting temperature of the mutation window, acquire the thermal decomposition temperature reference value and the historical failure threshold, calculate the absolute value of the difference between the boundary positioning temperature value and the thermal decomposition temperature reference value to obtain the first temperature deviation, and calculate the absolute value of the difference between the boundary positioning temperature value and the historical failure threshold to obtain the second temperature deviation. The existence of a low-value plateau window after the mutation window indicates that the local decline in the quality index is not an isolated fluctuation, but is connected to the subsequent low-value stable state. This sub-step uses the starting temperature of the mutation window as the boundary positioning temperature value because the starting position of the mutation window corresponds to the position where the quality response direction begins to change, which can serve as a candidate starting point for the physical failure boundary. The thermal decomposition temperature reference value comes from the thermal decomposition reference record associated with the material specification, and the historical failure threshold comes from the historical failure record. The first temperature deviation is used to describe the degree of deviation between the boundary positioning temperature value and the thermal decomposition temperature reference value, and the second temperature deviation is used to describe the degree of deviation between the boundary positioning temperature value and the historical failure threshold. Both are expressed as absolute deviations, regardless of positive or negative directions, because the verification focuses on the proximity between the candidate boundary and the reference failure position. If the order of the mutation window and the low-value plateau window is not established, then the boundary location temperature value will not be generated to avoid misjudging the interval before the low-value plateau, where there is no clear mutation, as a boundary segment.

[0076] S414: When either the first temperature deviation or the second temperature deviation is within a preset engineering error range, the starting temperature of the mutation window is used as the starting temperature of the physical failure boundary segment, and the starting temperature of the low-value plateau window is used as the ending temperature of the physical failure boundary segment. A physical failure boundary segment is established based on the starting and ending temperatures of the physical failure boundary segment. The preset engineering error range is used to determine whether the deviation between the boundary positioning temperature value and the thermal decomposition temperature reference value or the historical failure threshold is within an acceptable engineering consistency state. If either the first temperature deviation or the second temperature deviation meets this range, it indicates that the boundary candidate obtained by statistical positioning can be mutually verified with at least one piece of information from the physical thermal decomposition reference or historical failure experience. Subsequently, the starting temperature of the mutation window is used as the starting temperature of the physical failure boundary segment, and the starting temperature of the low-value plateau window is used as the ending temperature of the physical failure boundary segment, forming a segment from the start of the quality response mutation to the start of the low-value plateau. In this embodiment, the physical failure boundary segment is interval information in the direction of the effective temperature characteristic, used in S5 to divide the sample set before the boundary, the sample set near the boundary, and the sample set after the boundary. If neither of the two temperature deviations is within the preset engineering error range, then no physical failure boundary section will be established, and the candidate results will be returned to the screening process to re-examine other combinations of mutation windows and low-value plateau windows.

[0077] Please see Figure 1 and Figure 2S5: Based on the physical failure boundary segment, the product analysis sample set is divided into a pre-boundary sample set, a near-boundary sample set, and a post-boundary sample set. A quality pass threshold is obtained. Based on the near-boundary sample set and the quality pass threshold, the upper limit of process risk is determined. An over-temperature failure cause identifier is set for the post-boundary sample set. The purpose of this step is to transform the boundary segment formed in S4 into a sample region that can be used for modeling screening and constraint recommendation. The physical failure boundary segment includes the starting temperature and the ending temperature. The product analysis sample set is divided into three categories along the effective temperature characteristic direction: pre-boundary, near-boundary, and post-boundary. The pre-boundary sample set corresponds to samples with effective temperature characteristics lower than the boundary starting temperature, representing a segment that has not yet entered the physical failure boundary and will be used to establish an effective correlation model. The near-boundary sample set corresponds to samples with effective temperature characteristics falling within the boundary segment, representing a segment where quality risk transitions from an acceptable state to a failure state, and will be used to calculate the upper limit of process risk. The post-boundary sample set corresponds to samples with effective temperature characteristics higher than the boundary ending temperature, representing samples that have crossed the physical failure boundary, and will be used to set an over-temperature failure cause identifier. The quality acceptance threshold is obtained from the quality judgment record corresponding to the material specification and is used to determine whether the quality index is below the acceptance requirement. By analyzing the probability distribution of quality indexes below the quality acceptance threshold in the sample set near the boundary, the upper bound position where the risk becomes unacceptable can be found along the direction of the effective temperature feature; then, the upper bound of the effective temperature feature is selected according to the preset safety margin to obtain the upper limit position of the process risk. The preset safety margin is obtained from the process configuration record or quality risk control rules and is used to reserve a risk margin when the recommended upper limit is formed. The upper limit position of the process risk is then used by S6 to constrain the upper limit of the recommended interval, preventing the recommended interval from extending to the temperature range close to or beyond the failure risk. If a sample in the product analysis sample set cannot be compared with the physical failure boundary section due to the lack of effective temperature features, the sample will not participate in the three-category classification and the abnormal state will be retained for exclusion when generating the results.

[0078] S511: Retrieve the physical failure boundary segment and product analysis sample set; classify the analytical samples with effective temperature characteristics lower than the starting temperature of the physical failure boundary segment into the pre-boundary sample set; classify the analytical samples with effective temperature characteristics not lower than the starting temperature and not higher than the ending temperature of the physical failure boundary segment into the near-boundary sample set; classify the analytical samples with effective temperature characteristics higher than the ending temperature of the physical failure boundary segment into the post-boundary sample set. The pre-boundary, near-boundary, and post-boundary sample sets all use the sample structure from the product analysis sample set formed in S2, retaining effective temperature characteristics, quality indicators, material specification identifiers, sensor status records, data integrity records, and observation markers. This sub-step first reads the starting and ending temperatures of the physical failure boundary segment, then compares the relationship between the effective temperature characteristics of each analytical sample and the two boundary positions. Samples below the starting temperature enter the pre-boundary sample set, samples between the starting and ending temperatures enter the near-boundary sample set, and samples above the ending temperature enter the post-boundary sample set. This classification method stratifies samples of the same variety according to their physical failure boundary segments. This allows subsequent modeling to extract the effective correlation between effective temperature features and quality indicators only on the sample set before the boundary, while the risk upper limit calculation is concentrated on the sample set near the boundary. If the physical failure boundary segment is not established or the start-end relationship cannot form an effective segment, the three-class sample classification is not performed, and the boundary establishment results in S4 are returned for verification.

[0079] S512: Using the sample set near the boundary and the quality pass threshold, calculate the probability distribution of quality indicators below the quality pass threshold in the sample set near the boundary. Select the upper bound of the effective temperature feature from the probability distribution according to a preset safety margin to obtain the upper limit position of the process risk. Set an over-temperature failure cause identifier for the analytical samples in the sample set after the boundary. The sample set near the boundary covers the transition section from abrupt change to a low-value plateau in the quality response. Therefore, comparing the quality indicators with the quality pass threshold within this set yields the distribution relationship of the quality non-conforming state along the effective temperature feature direction. In this embodiment, the probability distribution is temperature-direction risk information formed by statistically analyzing the occurrence of quality indicators below the quality pass threshold in the sample set near the boundary. It describes the occurrence of quality below-pass judgment conditions corresponding to different effective temperature feature positions. The preset safety margin is a preset processing rule for conservatively constraining the risk position. Select the upper bound of the effective temperature feature from the probability distribution according to this rule to obtain the upper limit position of the process risk. The upper limit position of the process risk is not the physical failure boundary itself, but a recommended constraint upper limit obtained by combining the quality pass threshold and safety margin within the section near the boundary. Subsequently, an over-temperature failure cause identifier is set for the analytical samples in the sample set after the boundary. The over-temperature failure cause identifier is a status identifier written into the sample, indicating that the sample has reached the termination temperature of the physical failure boundary section and should be excluded according to the sample screening criteria during subsequent modeling. In this way, S6 will not include samples that have already crossed the failure boundary in the valid association relationship when building an effective association model and generating recommended intervals.

[0080] Please see Figure 1 , Figure 2 and Figure 4S6: Based on the pre-boundary sample set, perform correlation modeling to obtain an effective correlation model and recommended intervals for effective temperature features. Based on the physical failure boundary segment and the upper limit of process risk, obtain constrained recommended intervals for effective temperature features, and filter out analytical samples with over-temperature failure cause identifiers to generate manufacturing process data analysis results. The purpose of this step is to establish the correlation between effective temperature features and quality indicators using only samples before the physical failure boundary, and to limit the recommended intervals using the physical failure boundary segment and the upper limit of process risk. The pre-boundary sample set is located before the starting temperature of the physical failure boundary segment; the samples have not yet entered the failure boundary, making it suitable for establishing an effective correlation model. During correlation modeling, outlier samples from the process are first removed, and then weighted regression fitting is performed on the effective temperature features and quality indicators. Outlier samples from the process are those in the pre-boundary sample set whose relationship with the process data is inconsistent and interferes with the regression relationship. Removing them allows the model to primarily reflect the relationship between effective thermal effects before the boundary and quality indicators. Weighted regression fitting adjusts the impact of samples with large residuals on the regression curve according to a pre-defined weight update table by setting and updating weights for different samples. When the differences in model parameters corresponding to adjacent fitting results meet pre-defined convergence criteria, an effective correlation model is established. This effective correlation model is based on the relationship between effective temperature characteristics and quality indicators in the pre-boundary sample set, used to determine the effective temperature characteristic range corresponding to the quality indicator values. Subsequently, a recommended effective temperature characteristic interval is obtained based on this range, and further constrained by upper limits based on the physical failure boundary segment and the upper limit of process risk, forming a constrained recommended effective temperature characteristic interval. Finally, the over-temperature failure cause identifier is written into the correlation modeling sample selection criteria, ensuring that samples with this identifier after the boundary are explicitly excluded from the analysis result output layer (the pre-boundary sample set itself no longer contains such samples), thereby generating the manufacturing process data analysis results. The manufacturing process data analysis results carry the effective correlation model, the constrained recommended effective temperature characteristic interval, the physical effective analysis interval, and the sample selection criteria, enabling subsequent process analysis to simultaneously obtain quality correlation relationships and failure risk constraints.

[0081] S611: Call the pre-failure sample set, remove outlier samples from the process, perform weighted regression fitting on the effective temperature characteristics and quality indicators, and establish an effective correlation model when the model parameters converge to obtain the recommended interval for the effective temperature characteristics. This sub-step first reads the pre-failure sample set, as this set is located before the physical failure boundary and is suitable for describing the relationship between effective temperature characteristics and quality indicators before entering the failure boundary. When removing outlier samples, identification is based on the fitting residual state between the effective temperature characteristics and quality indicators to avoid deviations caused by local abnormal process states dominating the regression relationship. Preset initial weights are set for the analysis samples after removing outlier samples. These preset initial weights are obtained from the modeling configuration record and represent the initial influence assigned to the samples before fitting. A regression curve between the effective temperature characteristics and quality indicators is fitted based on the preset initial weights. The regression curve describes the continuous correspondence between temperature characteristic changes and quality indicator changes in the pre-failure sample. Subsequently, the quality indicator residual of the analysis sample is calculated. The quality indicator residual represents the degree of deviation between the actual quality indicator of the sample and the corresponding quality indicator on the regression curve. The absolute value of the quality index residuals is compared with a preset residual threshold, and the sample weights are updated according to a preset weight update table. The preset residual threshold and preset weight update table are obtained from the modeling configuration record and are used to explain the adjustment rules for sample weights under different deviation states. The regression curve is fitted again based on the updated sample weights, and the absolute value of the difference between the corresponding model parameters of two adjacent regression curves is compared. A preset convergence threshold, also obtained from the modeling configuration record, is used to determine whether the model parameter changes have reached a stable state. When the absolute value of the difference between the corresponding model parameters is less than the corresponding preset convergence threshold, it indicates that the model parameter changes after repeated fitting have met the stability criteria, and an effective correlation model is established. Based on the effective temperature feature range corresponding to the quality index values ​​in the effective correlation model, a recommended interval for the effective temperature feature is obtained. If the model parameters do not meet the convergence criteria, fitting continues based on the updated sample weights, and the final recommended interval is not output; if the samples available for fitting in the sample set before the boundary are insufficient to support correlation modeling after removal, the insufficient sample state is retained, and an effective correlation model is not established.

[0082] S612: Determine the physically effective analysis interval based on the physical failure boundary segment, and impose an upper limit constraint on the recommended interval of effective temperature characteristics based on the upper limit position of process risk, thus obtaining the constrained recommended interval of effective temperature characteristics. The physically effective analysis interval is the analyzable range of effective temperature characteristics determined based on the physical failure boundary segment, used to indicate that the recommended analysis should not exceed the physical failure boundary. This sub-step first reads the start and end temperatures of the physical failure boundary segment and determines the physically effective analysis interval accordingly. Then, it reads the upper limit position of process risk formed in S512 and compares the upper limit of the recommended interval of effective temperature characteristics with the upper limit position of process risk. If the upper limit of the recommended interval of effective temperature characteristics extends beyond the upper limit position of process risk, the recommended interval is shrunk according to the upper limit constraint rule; if the recommended interval is already before the upper limit position of process risk, its consistency with the physically effective analysis interval is maintained. Through this process, the constrained recommended interval of effective temperature characteristics is simultaneously constrained by the effective correlation model, the physical failure boundary segment, and the upper limit position of process risk, and will not extend to the failure risk area solely based on the pre-boundary fitting relationship.

[0083] S613: This step calls upon the effective correlation model, the recommended range of constrained effective temperature features, the physically effective analysis range, and the over-temperature failure cause identifier. It then writes the over-temperature failure cause identifier into the correlation modeling sample screening conditions. Based on these conditions, it filters out analysis samples with over-temperature failure cause identifiers, generating the manufacturing process data analysis results. The correlation modeling sample screening conditions are rules used to limit which samples are allowed to enter correlation modeling and result output. This sub-step writes the over-temperature failure cause identifier into these screening conditions, excluding over-temperature failure samples in the sample set after the boundary from the model usage and result generation. The filtering process does not delete the original acquisition records; instead, it prevents such samples from participating in the interpretation of the effective correlation model and the determination of the recommended range in the analysis chain, thus maintaining consistency between the model relationship and the effective process range before the boundary. The final generated manufacturing process data analysis results include the correspondence between the effective correlation model, the recommended range of constrained effective temperature features, the physically effective analysis range, and the sample screening conditions. If any result in the effective correlation model, the recommended range of constrained effective temperature features, or the physical effective analysis range fails to be generated, the manufacturing process data analysis results will not output complete recommended content, but will retain the corresponding missing state, indicating that the preceding boundary positioning, sample division, or correlation modeling steps need to be re-verified.

[0084] In this embodiment, the data transmission and parsing process uses production batch identifiers, material specification identifiers, sensor status records, and data integrity records as verification criteria at each stage. When input is missing, format is inconsistent, batches cannot be matched, record parsing fails, acquisition status is unconfirmed, or data integrity does not meet the judgment criteria, the corresponding sample is not directly included in the degradation scan, boundary establishment, or association modeling. Instead, it is categorized into rejection, retention marking, or result missing status based on its source of the anomaly. In this way, temperature feedback records, traction roller linear speed records, yarn breaking strength records, sensor status records, and data integrity records form a closed loop within the same data stream. All preceding results have a clear source, and subsequent inputs are generated from the preceding steps.

[0085] Please see Figure 5 A data analysis system for the manufacturing process of textile-specific equipment includes: a temperature zone data integration module, which performs multi-source record acquisition, temperature alignment and aggregation, quality index binding, and basic analysis sample set formation in S1; a variety cleaning and marking module, which performs material specification grouping, invalid sample removal, and observation mark setting in S2; a degradation feature identification module, which performs ascending scan, regression slope sequence, continuous feature sequence, abrupt change feature sequence, and low-value plateau feature sequence acquisition in S3; a failure boundary verification module, which performs boundary positioning temperature value determination, thermal decomposition temperature reference value and historical failure threshold verification, and physical failure boundary segment establishment in S4; a risk zoning and identification module, which performs three types of sample division, determination of process risk upper limit position, and over-temperature failure cause identification setting in S5; and a modeling and interval recommendation module, which performs association modeling, generation of constrained effective temperature feature recommendation intervals, over-temperature failure sample screening, and generation of manufacturing process data analysis results in S6.

[0086] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for analyzing manufacturing process data of textile-specific equipment, characterized in that, Includes the following steps: S1: Collect temperature feedback records, traction roller linear speed records, yarn breaking strength records, sensor status records, and data integrity records. Retrieve production batch identifiers and material specification identifiers. Align and aggregate the temperature feedback records to obtain effective temperature features. Bind quality indicators according to production batch identifiers to obtain a basic analysis sample set containing material specification identifiers, sensor status records, and data integrity records. S2: Based on the aforementioned basic analysis sample set, group the samples according to material specification identifiers, remove invalid samples, and set observation marks for samples exceeding the limits to obtain the variety analysis sample set; S3: For the sample set of the variety analysis, perform an ascending scan according to the effective temperature characteristics, calculate the regression slope within the window, obtain the continuous feature sequence based on the regression slope sequence within the window, obtain the mutation feature sequence based on the change of slope sign and the negative absolute value of the slope of the regression slope sequence within the window, calculate the mean of the quality index within the window and the standard deviation of the quality index within the window, and obtain the low value plateau feature sequence. S4: Based on the continuous feature sequence, the abrupt change feature sequence and the low-value platform feature sequence, locate the position of the abrupt change tending to the low-value platform, obtain the boundary positioning temperature value, obtain the thermal decomposition temperature reference value and the historical failure threshold, and verify the boundary positioning temperature value to determine the physical failure boundary segment. S5: Based on the physical failure boundary segment, the sample set of the variety analysis is divided into a sample set before the boundary, a sample set near the boundary, and a sample set after the boundary. The quality qualification threshold is obtained. Based on the sample set near the boundary and the quality qualification threshold, the upper limit position of the process risk is obtained. An over-temperature failure cause identifier is set for the sample set after the boundary. S6: Based on the pre-boundary sample set, perform correlation modeling to obtain an effective correlation model and an effective temperature feature recommendation interval. Based on the physical failure boundary segment and the upper limit position of the process risk, obtain the constrained effective temperature feature recommendation interval, and filter out the analysis samples with the over-temperature failure cause identifier to generate manufacturing process data analysis results.

2. The data analysis method for the manufacturing process of textile special equipment according to claim 1, characterized in that, The process of S1 is as follows: S111: Collect temperature feedback records of each temperature zone of the heat treatment equipment, traction roller linear speed records, sensor status records, data integrity records, and yarn breaking strength records of cooled yarn samples, and retrieve production batch identifiers and material specification identifiers. S112: Determine the material heating time based on the traction roller linear speed record, and perform time-series alignment and weighted aggregation of the temperature feedback record of the temperature zone according to the material heating time to generate effective temperature characteristics; S113: Call the effective temperature feature, use the yarn breaking strength record as a quality indicator, bind the effective temperature feature and the quality indicator to the production batch according to the production batch identifier, and associate the material specification identifier, the sensor status record and the data integrity record to the bound analysis sample according to the production batch identifier to obtain a basic analysis sample set containing the material specification identifier, sensor status record and data integrity record.

3. The data analysis method for the manufacturing process of textile special equipment according to claim 2, characterized in that, The process of aligning and weighting the temperature feedback records of the temperature zone according to the heating time of the material is as follows: The heating time of the material is calculated based on the linear speed record of the traction roller and the preset physical length of the heating box. The heating time period record matching the production batch identifier is extracted from the temperature feedback record of the temperature zone according to the production batch identifier. The heating time period record is weighted and summed according to the preset temperature zone weight to generate the effective temperature feature.

4. The data analysis method for the manufacturing process of textile special equipment according to claim 1, characterized in that, The process of S2 is as follows: S211: Call the basic analysis sample set, group the basic analysis sample set according to the material specification identifier to obtain the variety group sample set, and based on the variety group sample set, remove null value samples, zero value samples and garbled samples according to the sensor status record and data integrity record to obtain the effective sample set; S212: Call the effective sample set, set observation marks for analysis samples whose effective temperature characteristics exceed the preset production setting range and whose data integrity records and sensor status records meet the preset acquisition conditions, and obtain the variety analysis sample set.

5. The data analysis method for the manufacturing process of textile-specific equipment according to claim 1, characterized in that, The process of S3 is as follows: S311: Call the variety analysis sample set, arrange the analysis samples in ascending order according to the effective temperature characteristics, keep the analysis samples corresponding to the observation marks within the sliding window scanning range, calculate the regression slope between the effective temperature characteristics and quality indicators within the window, and obtain the regression slope sequence within the window. S312: Obtain a continuous feature sequence based on the regression slope sequence within the window, and obtain a mutation feature sequence based on the change in the slope sign and the negative absolute value of the slope of the regression slope sequence within the window; S313: For the sliding window, calculate the mean and standard deviation of the quality indicators within the window, and obtain the low-value platform feature sequence based on the mean and standard deviation of the quality indicators within the window.

6. The method for analyzing manufacturing process data of textile-specific equipment according to claim 1, characterized in that, The process of S4 is as follows: S411: Call the continuous feature sequence, the abrupt change feature sequence and the low value plateau feature sequence, and filter continuous stable segments with consistent slope signs and absolute values ​​of adjacent slope differences less than a preset continuity threshold according to the continuous feature sequence; S412: Extract the segment after the continuous stable segment, and filter the mutation window with a slope that changes from positive to negative and the absolute value of the negative slope is greater than the preset process fluctuation slope threshold according to the mutation feature sequence. Filter the low-value platform window with a mean quality index less than the preset low-value quality threshold and a standard deviation of the quality index less than the preset fluctuation threshold according to the low-value platform feature sequence. S413: Call the mutation window and the low-value platform window. When the low-value platform window exists after the mutation window, obtain the boundary positioning temperature value according to the starting temperature of the mutation window, obtain the thermal decomposition temperature reference value and the historical failure threshold, calculate the absolute value of the difference between the boundary positioning temperature value and the thermal decomposition temperature reference value to obtain the first temperature deviation, and calculate the absolute value of the difference between the boundary positioning temperature value and the historical failure threshold to obtain the second temperature deviation. S414: When either the first temperature deviation or the second temperature deviation is within a preset engineering error range, the starting temperature of the sudden change window is taken as the starting temperature of the physical failure boundary section, and the starting temperature of the low value platform window is taken as the ending temperature of the physical failure boundary section. Based on the starting temperature and ending temperature of the physical failure boundary section, a physical failure boundary section is established.

7. The method for analyzing manufacturing process data of textile-specific equipment according to claim 1, characterized in that, The process of S5 is as follows: S511: Invoke the physical failure boundary section and the variety analysis sample set; The analysis samples with effective temperature characteristics lower than the starting temperature of the physical failure boundary section are divided into the pre-boundary sample set. The analysis samples with effective temperature characteristics that are not less than the starting temperature of the physical failure boundary section and not greater than the ending temperature of the physical failure boundary section are divided into a sample set near the boundary. The analysis samples whose effective temperature characteristics are greater than the termination temperature of the physical failure boundary section are divided into a post-boundary sample set. S512: Call the sample set near the boundary and the quality qualification threshold, calculate the probability distribution of the quality index in the sample set near the boundary being lower than the quality qualification threshold, select the upper limit of the effective temperature feature from the probability distribution according to the preset safety margin, obtain the upper limit position of the process risk, and set the over-temperature failure cause identifier for the analysis samples in the sample set after the boundary.

8. The method for analyzing manufacturing process data of textile-specific equipment according to claim 1, characterized in that, The process of S6 is as follows: S611: Call the pre-boundary sample set, remove outlier samples from the process, perform weighted regression fitting on the effective temperature features and quality indicators, establish an effective correlation model when the model parameters converge, and obtain the recommended interval of the effective temperature features. S612: Determine the physical effective analysis interval based on the physical failure boundary segment, and impose an upper limit constraint on the recommended effective temperature feature interval based on the upper limit position of the process risk, to obtain the constrained recommended effective temperature feature interval; S613: Call the effective correlation model, the recommended range of constrained effective temperature features, the physical effective analysis range, and the over-temperature failure cause identifier, write the over-temperature failure cause identifier into the correlation modeling sample screening conditions, filter out the analysis samples with the over-temperature failure cause identifier according to the correlation modeling sample screening conditions, and generate manufacturing process data analysis results.

9. The method for analyzing manufacturing process data of textile special equipment according to claim 8, characterized in that, The process of performing weighted regression fitting on effective temperature characteristics and quality indicators, establishing an effective correlation model when the model parameters converge, and obtaining the recommended interval for effective temperature characteristics is as follows: A preset initial weight is set for the analysis sample after removing the outlier samples of the process. The regression curve between the effective temperature feature and the quality index is fitted according to the preset initial weight. The quality index residual of the analysis sample is calculated. The absolute value of the quality index residual is compared with the preset residual threshold. The sample weight is updated according to the preset weight update table. The regression curve is fitted again based on the updated sample weights. When the absolute value of the difference between the corresponding model parameters of two adjacent regression curves is less than the corresponding preset convergence threshold, the effective correlation model is established. Based on the effective temperature feature range corresponding to the quality index values ​​in the effective correlation model, the recommended interval for the effective temperature feature is obtained.

10. A data analysis system for the manufacturing process of textile-specific equipment, characterized in that, The system is used to implement the textile special equipment manufacturing process data analysis method according to any one of claims 1-9, and the system includes: Temperature zone data integration module: Collects temperature feedback records, traction roller linear speed records, yarn breaking strength records, sensor status records, and data integrity records; retrieves production batch identifiers and material specification identifiers; aligns and aggregates temperature feedback records to obtain effective temperature features; and binds quality indicators according to production batch identifiers to obtain a basic analysis sample set containing material specification identifiers, sensor status records, and data integrity records. Variety cleaning and marking module: Based on the basic analysis sample set, the sample set is grouped according to material specification identification, invalid samples are removed, and observation marks are set for samples exceeding the limit to obtain the variety analysis sample set; Degradation Feature Identification Module: For the sample set of the variety analysis, the module performs an ascending scan based on the effective temperature feature, calculates the regression slope within the window, obtains the continuous feature sequence based on the regression slope sequence within the window, obtains the abrupt change feature sequence based on the change in the slope sign and the negative absolute value of the slope, calculates the mean and standard deviation of the quality indicators within the window, and obtains the low-value plateau feature sequence. Failure boundary verification module: Based on the continuous feature sequence, the abrupt change feature sequence and the low value plateau feature sequence, locate the position of the abrupt change tending to the low value plateau, obtain the boundary location temperature value, acquire the thermal decomposition temperature reference value and historical failure threshold, verify the boundary location temperature value, and determine the physical failure boundary segment; Risk zoning and identification module: Based on the physical failure boundary segment, the sample set of the product analysis is divided into a sample set before the boundary, a sample set near the boundary, and a sample set after the boundary. The quality qualification threshold is obtained. Based on the sample set near the boundary and the quality qualification threshold, the upper limit position of the process risk is obtained. The over-temperature failure cause identification is set for the sample set after the boundary. Modeling and Interval Recommendation Module: Based on the pre-boundary sample set, perform correlation modeling to obtain an effective correlation model and effective temperature feature recommendation intervals. Based on the physical failure boundary segment and the upper limit position of the process risk, obtain constrained effective temperature feature recommendation intervals, and filter out analysis samples with the over-temperature failure cause identifier to generate manufacturing process data analysis results.

Images