A data processing method

Abstract

The present invention discloses a data processing method, which relates to the field of industrial data processing technology and includes the following steps: First, various parameter information during the operation of textile equipment is collected according to a preset monitoring frequency. By dynamically adjusting the monitoring frequency, the present invention enables the system to more sensitively capture minor anomalies and avoid the omission of low-frequency problems in traditional fixed-frequency monitoring. Combined with a machine learning model, real-time analysis of loom tension fluctuations and cycle time changes can be performed to effectively prevent the accumulation of quality problems and improve product consistency and pass rate. At the same time, dynamic adjustment optimizes resource utilization, avoids computational waste caused by excessive monitoring, and improves monitoring efficiency and economy.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial data processing, and in particular to a data processing method. Background Art

[0002] Data processing refers to the process of organizing, transforming, analyzing, and optimizing raw data in order to extract valuable information or generate actionable results. This process typically includes several key steps: first, data acquisition, which is the collection of raw data through sensors, databases, or other sources; followed by data cleaning, which improves data quality by removing noise, erroneous data, or incomplete data; then data conversion, which adjusts the data format or structure to a form suitable for further analysis, such as data standardization, normalization, or encoding conversion; followed by data analysis, which uses statistical methods, algorithms, or machine learning models to deeply explore the data to find trends, patterns, or associations; finally, through data visualization or report generation, the analysis results are presented to decision makers or system users to support decision-making, optimize operations, or make predictions. The goal of data processing is to improve the availability of data and help make more accurate decisions in complex environments.

[0003] Data processing plays a vital role in textile production, particularly in the intelligent and automated development of textile machinery. By collecting and processing sensor data (such as temperature, humidity, speed, tension, and pressure) from various textile machines, including looms, dyeing machines, and finishing equipment, data processing can help monitor and optimize equipment performance and prevent potential failures. For example, data processing can analyze the operating status of a loom and identify abnormal vibrations or tension fluctuations, providing early warning of equipment failures and reducing downtime. During the dyeing process, variables such as dye concentration and temperature can be monitored in real time, and data corrections performed to ensure dyeing uniformity. Data analysis can also optimize production processes, such as adjusting machine speeds, reducing raw material waste, and improving production efficiency. The application of data processing in textile production not only improves the stability of equipment operation but also enhances product quality through precise control, ensuring efficient and accurate production processes.

[0004] The existing technology has the following deficiencies:

[0005] In the textile production process, existing technologies typically use a fixed monitoring frequency to identify abnormal loom operation. While this method effectively captures most abnormal operating modes during equipment operation, fixed-frequency monitoring may not detect minor operating anomalies in a timely manner, leading to serious consequences. Although minor anomalies may not immediately affect loom production efficiency, they can have a cumulative impact on fabric quality. For example, uneven fabric tension can lead to uneven thickness, texture deviations, or surface defects. Fixed-frequency monitoring systems struggle to detect these minor quality fluctuations in a timely manner, resulting in defective products. These defective products may remain undetected in the short term, but as production batches increase, quality issues gradually become apparent, ultimately affecting product yields and even leading to rework or scrap. Therefore, traditional methods that rely on fixed-frequency monitoring may not effectively ensure production stability and product quality consistency.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0007] The purpose of the present invention is to provide a data processing method that, by dynamically adjusting the monitoring frequency, enables the system to more sensitively identify minor anomalies, avoiding the problem of traditional fixed-frequency monitoring ignoring low-frequency anomalies. This flexible monitoring method, combined with a machine learning model, effectively captures key features such as loom tension fluctuations and cycle time changes, thereby preventing the accumulation of quality issues and improving product consistency and pass rate. Compared to fixed-frequency monitoring, dynamic adjustment not only increases sensitivity to potential problems but also optimizes resource utilization, avoiding excessive consumption of computing resources, thereby improving monitoring efficiency and economy, and addressing the problems mentioned above in the background technology.

[0008] In order to achieve the above object, the present invention provides the following technical solution: a data processing method comprising the following steps:

[0009] First, various parameter information during the operation of the textile equipment is collected according to the preset monitoring frequency;

[0010] Collected textile equipment operating parameters are aggregated in real time to form a data set, and key features reflecting potential subtle anomalies are extracted from the data set;

[0011] Within the detection window, the extracted key features are analyzed and processed in detail, and the analyzed and processed key features are input into a pre-trained machine learning model, which then performs an intelligent assessment of the abnormal operating status of the equipment.

[0012] Based on the evaluation results of the machine learning model, the abnormal operating status of the equipment is classified into two categories: normal operating status and minor abnormal operating status;

[0013] For normal operation, continue to collect and analyze data at the preset monitoring frequency. By maintaining the existing frequency, ensure that significant equipment faults can be discovered and resolved in a timely manner to prevent the problem from escalating.

[0014] For minor abnormal operating conditions, the monitoring frequency is dynamically increased based on the evaluation results of the machine learning model to capture more production data points, thereby enhancing sensitivity to minor anomalies and promptly identifying their potential impacts.

[0015] Preferably, key features reflecting potential low-frequency anomalies are extracted from the data set, and the extracted features include the fluctuation amplitude of the loom tension and the change of the cycle time of the loom during the production process. Under the detection window, the extracted fluctuation amplitude of the loom tension and the change of the cycle time of the loom during the production process are analyzed and processed to generate a loom tension change index and a loom operation cycle time index. The loom tension change index reflects the tension unevenness of the fabric during the weaving process by quantifying the amplitude and frequency of the tension fluctuation of the loom during the production process; the loom operation cycle time index quantifies the slight changes in the cycle time during the loom production process.

[0016] Preferably, after obtaining the loom tension change index and the loom operation cycle time index generated after analyzing and processing the key features, the loom tension change index and the loom operation cycle time index are input into a pre-learned machine learning model, and an abnormality recognition coefficient is generated by the machine learning model, and the equipment operation status is intelligently evaluated by the abnormality recognition coefficient.

[0017] Preferably, the abnormality identification coefficient generated after analyzing and processing the extracted key features is compared with a preset abnormality identification coefficient reference threshold value to divide the equipment operation status. The division steps are as follows:

[0018] If the abnormality recognition coefficient is greater than or equal to the preset abnormality recognition coefficient reference threshold, the operating state of the equipment is classified as an unrecognizable minor abnormality;

[0019] If the abnormality recognition coefficient is less than a preset abnormality recognition coefficient reference threshold, the operating state of the equipment is classified as a recognizable abnormality.

[0020] Preferably, the specific steps for dynamically increasing the monitoring frequency based on the evaluation results of the machine learning model to capture more production data points, thereby enhancing sensitivity to minor anomalies and promptly identifying their potential impact are as follows:

[0021] The generated abnormality identification coefficient ADC is compared with the preset abnormality identification coefficient reference threshold to obtain the degree of deviation. The calculation expression is as follows:

[0022]

[0023] , where ΔADC is the abnormal deviation value, ADC ref is the reference threshold of the abnormal identification coefficient;

[0024] Based on the abnormal deviation value, if a minor abnormality is detected, the monitoring frequency will be dynamically adjusted. In order to enhance the sensitivity to minor abnormalities, the monitoring frequency will be adjusted according to the deviation degree of the current abnormal recognition coefficient. The calculation expression is as follows:

[0025]

[0026] , where DFR new is the adjusted detection frequency, DFR base is the preset monitoring frequency, τ is the frequency adjustment coefficient;

[0027] After dynamically adjusting the monitoring frequency, more data points are collected at a higher frequency. Each collected data point will be further used to analyze the operating status of the equipment. At this time, the updated anomaly recognition coefficient is calculated based on each new data point. The calculation expression is as follows:

[0028]

[0029] , where X y is the yth collected data point, indicating the yth production data point collected at the new frequency. X is the total number of data points collected at the adjusted monitoring frequency. mean is the mean of the current collected data set, σ X is the standard deviation of the data set, ADC new is the updated anomaly identification coefficient;

[0030] After collecting new data, we can determine whether the monitoring frequency needs to be further adjusted by calculating the potential impact of anomalies on production in order to promptly identify and address these potential problems. The calculation expression is as follows:

[0031]

[0032] , where Impact postential Is the potential impact, ADC y is the anomaly recognition coefficient of the y-th data point, δ y is the weight coefficient of the y-th data point, which is used to indicate the importance of the y-th data point to the final impact, 1+μ×(ΔADC) 2is the nonlinear adjustment term of the influence, and μ is the adjustment coefficient, which is used to control the nonlinear relationship of the influence;

[0033] If Impact potential Greater than the preset impact threshold Impact threshold , the feedback mechanism is triggered to adjust the operating status of the equipment and further adjust the monitoring frequency to ensure that minor anomalies can be identified and eliminated in a timely manner to prevent them from causing long-term impact on production.

[0034] Preferably, within the detection window, the specific steps of analyzing and processing the fluctuation amplitude of the loom tension to generate the loom tension variation index are as follows:

[0035] Within the detection window, the intensity of the loom tension fluctuation is first quantified by calculating the tension variation amplitude factor. The tension variation amplitude factor measures the change in the tension signal at adjacent time points and is analyzed in combination with the maximum deviation of the fluctuation to effectively reflect the amplitude of the tension fluctuation. The calculation expression is as follows:

[0036]

[0037] , where TAF is the tension variation factor, T(t i ) is the time point t i The loom tension signal value at time t i is the i-th time point t, T(t i-1 ) is the previous time point t i-1 The loom tension signal value at time N is the total number of time points, and α is the weight factor;

[0038] Next, we extract nonlinear dynamic features from the tension fluctuations. We use the tension fluctuation index to quantify the nonlinear change trend of the tension signal. Through Fourier transform, we decompose the tension signal into different frequency components to capture the periodic fluctuation pattern. The calculation expression is as follows:

[0039]

[0040] , where TFI is the tension fluctuation index, is the tension signal component after Fourier transform, that is, the Fourier transform result of the tension signal T(t) at the frequency domain frequency f, β is the adjustment index, F is the maximum frequency in the frequency domain, is the high-frequency component attenuation factor, which is used to attenuate the influence of high-frequency components, and λ is the adjustment parameter of the attenuation factor;

[0041] The complexity of tension fluctuation is calculated based on the dynamic model. The tension non-stationarity coefficient is used to measure the variation pattern of loom tension. The tension non-stationarity coefficient is used to capture the time dependence of tension fluctuation and reflect the fluctuation difference of loom tension in different time intervals. The calculation expression is as follows:

[0042]

[0043] , where TNSc is the tension non-stationary coefficient, is the rate of change of tension, i.e. the rate of change of loom tension over time, γ represents the nonlinear index, which is an adjustment index, e is the natural base, δ is the attenuation factor, and T is the total time length of the detection window;

[0044] The tension variation index of the loom is generated by integrating the tension variation amplitude factor TAF, the tension fluctuation index TFI and the tension non-stationarity coefficient TNSc. The calculation expression is as follows:

[0045]

[0046] , where TVI is the loom tension variation index.

[0047] Preferably, the specific steps of analyzing and processing the cycle time variation of the loom during the production process to generate the loom operation cycle time index are as follows:

[0048] In the detection window, first define the cycle time of the loom, that is, the time required for the loom to complete one cycle of production. The calculation expression is as follows:

[0049]

[0050] , where T cycle (t) is the cycle time at time point t, T production (t) is the total production time at time t, N units (t) is the number of units produced by the loom at time t;

[0051] Through the cycle time T cycle (t) Calculate the variation rate of the cycle time. The calculation expression is as follows:

[0052]

[0053] , where CTV is the cycle time variation rate, T cycle (t-1) is the cycle time at the previous time point t-1;

[0054] After analyzing the variation rate of the cycle time, the loom operation cycle time index is generated based on the cycle time variation rate. The calculation expression is as follows:

[0055]

[0056] , where F cycle is the loom operation cycle time index, ω is the magnification coefficient, e is the natural base number, is the attenuation coefficient;

[0057] To prevent the cycle time index from fluctuating sharply in the short term and affecting the entire production process, we then perform a time window weighted average on the cycle time index to retain trend changes over a longer period of time and suppress occasional abnormal fluctuations in the short term. The expression is as follows:

[0058]

[0059] , where I cycle is the weighted cycle time index, F cycle (tj) refers to the loom operation cycle time index at the previous time point tj, M is the total number of data points, j refers to the jth data point, and represents the offset step from time point t to the past time point, w j is the weight coefficient of each loom operation cycle time index;

[0060] Based on the weighted cycle time index I cycle , the final loom operation cycle time index of the loom is obtained, and the expression is as follows:

[0061]

[0062] , where CTDI is the loom cycle time index and θ is the adjustment coefficient used to scale the index according to actual production needs.

[0063] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0064] The present invention can significantly improve the ability to capture minor anomalies by introducing a mechanism for dynamically adjusting the monitoring frequency. Traditional fixed-frequency monitoring methods often ignore smaller, low-frequency abnormal changes, causing these minor fluctuations to accumulate into quality problems without being identified in time. By combining the intelligent evaluation of the machine learning model and extracting key features such as the loom tension change index and the cycle time change index, the system can keenly capture minor abnormal fluctuations. For example, the tension unevenness of the fabric may not be discovered in the early stages, but as production progresses, these minor fluctuations gradually affect the uneven thickness, texture deviation or surface defects of the fabric. If they can be discovered and adjusted in time at the early stage, the accumulation of quality problems can be effectively avoided, thereby improving the consistency and pass rate of products and avoiding the risk of defective products, rework or scrap.

[0065] The present invention dynamically adjusts the monitoring frequency, and the system can proactively increase the monitoring frequency when it identifies unrecognizable subtle anomalies. This can enhance sensitivity to potential problems while ensuring that resources are not over-consumed. Compared with the traditional method of continuous monitoring at a fixed frequency, this flexible adjustment method can be optimized according to the actual needs of the production environment. For significant faults, the system continues to use the traditional frequency for monitoring to ensure that problems are discovered and resolved in a timely manner; for minor anomalies, the system obtains more data points by increasing the monitoring frequency, thereby accurately capturing these potential impacts and preventing the waste of computing resources caused by overly intensive monitoring data. This method can ensure that the operating status of the equipment is effectively monitored while avoiding unnecessary monitoring burdens and waste of computing resources, thereby improving the efficiency and economy of production monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0067] Figure 1 The present invention is a flowchart of a data processing method. DETAILED DESCRIPTION

[0068] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0069] The present invention provides Figure 1 A data processing method shown includes the following steps:

[0070] First, various parameter information during the operation of the textile equipment is collected according to the preset monitoring frequency;

[0071] A preset monitoring frequency refers to the regular collection and recording of equipment operating parameter data at pre-set intervals or data point collection frequencies during the operation of textile equipment. This frequency is typically set based on the equipment's operating characteristics and production requirements, and can be every second, every minute, or at other intervals. For example, for certain fast-running looms, the monitoring frequency might be set to once per second to quickly respond to changes; whereas for more stable equipment, the monitoring frequency can be appropriately reduced.

[0072] The primary purpose of a preset monitoring frequency is to ensure the system can efficiently collect and process equipment operating data, especially in scenarios where batch data processing is required. By setting a fixed monitoring frequency, the system can regularly collect data and identify potential equipment issues or anomalies without placing an excessive burden on equipment performance. For example, when certain equipment operating parameters such as temperature, speed, and tension change, the system can use this fixed frequency to promptly capture and analyze data, providing early warning of potential failures or quality issues.

[0073] Collected textile equipment operating parameters are aggregated in real time to form a data set, and key features reflecting potential subtle anomalies are extracted from the data set;

[0074] The specific steps for collecting the textile equipment operating parameters in real time to form a data set are as follows:

[0075] First, textile equipment continuously collects various operating parameters, such as loom tension, speed, temperature, humidity, and vibration, through sensors. Whenever new parameter data is collected, the system stores it in a temporary cache in real time, based on timestamp and parameter type. Subsequently, this data is aggregated and transmitted to a central database or data storage platform, either periodically or based on event triggers. During this process, the data requires preliminary cleaning and standardization to remove noise and outliers to ensure the accuracy of subsequent analysis. Each set of data is categorized and labeled according to the different operating states of the equipment, forming a systematic and structured data set. This data set contains information on the various operating states of the equipment, facilitating the extraction of meaningful features for in-depth analysis.

[0076] The primary purpose of aggregating textile equipment operating parameters in real time into a data set is to provide a foundation for subsequent data analysis and feature extraction. This aggregated data set systematically stores and organizes equipment operating information, laying the data foundation for subsequent machine learning model training, anomaly detection, and fault prediction. Data sets efficiently store various operating parameters while supporting multi-dimensional analysis, helping to extract hidden patterns and potential minor anomalies from large amounts of data. This structured data not only supports retrospective analysis of historical data but also enables real-time monitoring of equipment operating status, promptly identifying early signs of quality fluctuations, equipment anomalies, or failures, thereby providing data support for improving production stability and product quality.

[0077] Within the detection window, the extracted key features are analyzed and processed in detail, and the analyzed and processed key features are input into a pre-trained machine learning model, which then performs an intelligent assessment of the abnormal operating status of the equipment.

[0078] Key features reflecting potential low-frequency anomalies are extracted from the data set. The extracted features include the fluctuation amplitude of the loom tension and the change of the cycle time of the loom during the production process. Within the detection window, the extracted fluctuation amplitude of the loom tension and the change of the cycle time of the loom during the production process are analyzed and processed to generate the loom tension variation index and the loom operation cycle time index. The loom tension variation index reflects the tension unevenness of the fabric during the weaving process by quantifying the amplitude and frequency of the tension fluctuation of the loom during the production process; the loom operation cycle time index quantifies the slight changes in the cycle time during the loom production process.

[0079] After obtaining the loom tension variation index and loom operation cycle time index generated by analyzing and processing the key features, the loom tension variation index and loom operation cycle time index are input into a pre-learned machine learning model, and an abnormality recognition coefficient is generated by the machine learning model. The abnormality recognition coefficient is used to perform an intelligent assessment of the equipment operation status;

[0080] A pre-learned machine learning model is a model built using historical data and a training process. It can automatically identify and predict potential anomalies in equipment operation based on input feature data. This model is typically trained by data scientists or engineers using a large amount of equipment operation data (including data from both normal and abnormal conditions). During training, the model analyzes the relationships between various features in this data to learn the distribution patterns and changing trends of features under different operating conditions. Model training includes steps such as data preprocessing, feature extraction, labeling, model selection, and parameter tuning. After training, the model can automatically identify various anomalies that may occur during loom production based on the existing dataset, including subtle tension fluctuations and cycle time variations. These models can be supervised learning models such as support vector machines (SVMs), decision trees, and random forests, or deep learning models such as neural networks. The specific choice depends on the complexity of the problem and the characteristics of the data.

[0081] Once the machine learning model is trained and optimized, it can assess the loom's operating status in real time during actual production. In practice, the monitoring system will collect characteristic data, such as the loom tension variation index and the loom cycle time index, and feed these data into the pre-trained machine learning model. The model compares the real-time input data with patterns in historical training data to determine whether the current equipment status falls within the abnormal range. If the model identifies a similarity between the current operating status and a known abnormal pattern, it generates an abnormality identification coefficient to quantify the difference between the current equipment status and normal operation. A higher abnormality identification coefficient indicates a greater deviation from normal equipment operation, indicating a potential failure or quality issue. Through this intelligent assessment, the system can proactively identify and warn of potential equipment failures and provide appropriate adjustment or maintenance recommendations, thereby ensuring production stability and product quality. This machine learning-based abnormality assessment not only identifies significant failures but also detects low-frequency, subtle abnormal changes, significantly improving monitoring efficiency and accuracy during production.

[0082] Gradual fluctuations in loom tension amplitude do indicate a minor, progressive anomaly in loom operation. This type of fluctuation is typically caused by factors such as gradual wear, loosening, or poor lubrication of internal components. Because these issues often occur very slowly and in small increments, they do not significantly impact loom efficiency in the short term. However, over time, these gradual fluctuations in tension can accumulate, leading to deviations in fabric quality, such as uneven tension and inconsistent thickness. If these subtle, progressive anomalies are not promptly detected and corrected, they can have long-term, insidious consequences for product quality, ultimately resulting in large quantities of fabric failing to meet quality standards. Therefore, while these gradual fluctuations may be difficult to detect initially, the underlying anomaly, if not effectively monitored and corrected, can pose a serious threat to production stability and fabric quality consistency.

[0083] In the detection window, the specific steps for analyzing and processing the fluctuation amplitude of the loom tension and generating the loom tension change index are as follows:

[0084] Within the detection window, the intensity of the loom tension fluctuation is first quantified by calculating the tension variation amplitude factor. The tension variation amplitude factor measures the change in the tension signal at adjacent time points and is analyzed in combination with the maximum deviation of the fluctuation to effectively reflect the amplitude of the tension fluctuation. The calculation expression is as follows:

[0085]

[0086] , where TAF is the tension variation factor, T(t i ) is the time point ti The loom tension signal value at time t i is the i-th time point t, T(t i-1 ) is the previous time point t i-1 The loom tension signal value at time N is the total number of time points, and α is the weight factor;

[0087] The weighting factor α plays a role in adjusting the sensitivity to tension fluctuations in the calculation of the tension amplitude factor TAF(t). By setting the weighting factor, the weighted effect of tension fluctuations can be controlled. Specifically, a larger α value amplifies the impact of large fluctuations, making larger fluctuations contribute more significantly to the final result; while a smaller α value weakens the impact of fluctuations and emphasizes more stable changes. Therefore, the weighting factor serves to adjust the importance of tension changes of different amplitudes in the overall fluctuation assessment, thereby making the calculated results more consistent with the sensitivity requirements for small fluctuations in actual production environments. By properly adjusting α, small anomalies or fluctuations in loom tension can be more accurately captured, helping to improve the accuracy of anomaly detection.

[0088] The tension variation amplitude factor TAF(t) is an indicator that quantifies the intensity of loom tension fluctuations. It reflects the magnitude and frequency of tension fluctuations during loom operation by calculating the amplitude of the change in loom tension between different time points. By analyzing the tension differences between adjacent time points and weighting these differences, TAF can provide a comprehensive assessment of loom tension fluctuations. Its significance lies in that the tension variation amplitude factor helps us more accurately capture small fluctuations in loom operation, especially those that may not cause immediate failures but may affect fabric quality. This is crucial for preventing quality problems in production (such as uneven thickness of the fabric, texture deviation, etc.), because these small tension fluctuations may gradually accumulate during the production process and eventually have a negative impact on fabric quality. Through quantitative analysis of tension fluctuations, potential quality problems can be discovered in a timely manner and adjusted, thereby improving production stability and product quality consistency.

[0089] Next, we extract nonlinear dynamic features from the tension fluctuations. We use the tension fluctuation index to quantify the nonlinear change trend of the tension signal. Through Fourier transform, we decompose the tension signal into different frequency components to capture the periodic fluctuation pattern. The calculation expression is as follows:

[0090]

[0091] , where TFI is the tension fluctuation index, is the tension signal component after Fourier transform, that is, the Fourier transform result of the tension signal T(t) at the frequency domain frequency f, β is the adjustment index, which is used to control the weight of each frequency component in the calculation of the tension fluctuation index, F is the maximum frequency in the frequency domain, that is, the maximum frequency range that can be considered in the frequency domain after Fourier transform, It is the high-frequency component attenuation factor, which is used to attenuate the influence of high-frequency components. λ is the adjustment parameter of the attenuation factor. As the frequency f increases, the attenuation factor will reduce the contribution of high-frequency components, which can effectively suppress the high-frequency influence caused by noise or instantaneous fluctuations.

[0092] The complexity of tension fluctuation is calculated based on the dynamic model. The tension non-stationarity coefficient is used to measure the variation pattern of loom tension. The tension non-stationarity coefficient is used to capture the time dependence of tension fluctuation and reflect the fluctuation difference of loom tension in different time intervals. The calculation expression is as follows:

[0093]

[0094] , where TNSc is the tension non-stationary coefficient, is the tension change rate, that is, the rate of change of the loom tension over time, which describes the speed of tension change at an instant and can reveal the sharp fluctuation of loom tension change. γ represents the nonlinear index, which is a regulating index that controls the impact of the tension change rate on non-stationarity. By weighting the power of the change rate, γ can enhance the impact of large fluctuations or suppress small fluctuations. e is the natural base, δ is the attenuation factor, and T is the total time length of the detection window.

[0095] The attenuation factor δ is a parameter used to control the degree of time influence. In dynamic system calculations, it gradually reduces the impact of historical data on current results. Specifically, the attenuation factor determines the weight that fluctuations at more distant points in the time series have on the current calculation results. Over time, more distant fluctuations are given smaller weights, thereby reducing their impact on the overall assessment. In the calculation of loom tension changes, the attenuation factor is used to mitigate the impact of long-term tension fluctuations, ensuring that only important changes in the near future are focused. This approach helps to avoid the interference of historical fluctuations on the current non-stationarity assessment, making the model more sensitive to current and short-term tension changes and helping to capture subtle abnormal signals in a timely manner.

[0096] The tension variation index of the loom is generated by integrating the tension variation amplitude factor TAF, the tension fluctuation index TFI and the tension non-stationarity coefficient TNSc. The calculation expression is as follows:

[0097]

[0098] , where TVI is the loom tension variation index;

[0099] Within the detection window, the larger the loom tension variation index, generated after analyzing and processing the fluctuation amplitude of the loom tension, the more likely it is that a slight, gradual abnormality has occurred during the loom's operation. An increase in the tension fluctuation amplitude may indicate that problems such as wear, poor lubrication, or slight mechanical looseness of certain internal loom components are gradually occurring. These abnormal changes tend to accumulate with smaller amplitudes and lower frequencies, and are unlikely to cause significant production efficiency issues in the short term, but will gradually affect the quality consistency of the fabric. If the loom tension variation index is large, it indicates that the system has identified these tiny abnormal fluctuations, which may be a sign of unstable equipment operation. Conversely, a small loom tension variation index generally indicates that the loom is operating smoothly, the tension fluctuation amplitude is within the normal range, and there are no obvious abnormalities or faults.

[0100] Slight fluctuations in a loom's cycle time during production often indicate a minor, gradual abnormality in its operation. Cycle time is the time it takes a loom to complete a full weaving cycle. Any deviation from the normal cycle time can indicate a change in the loom's operating status. Even small fluctuations often indicate a gradual change in the performance of certain parts of the equipment, such as minor wear on mechanical parts, uneven loading, or unstable tension. These fluctuations may not immediately lead to serious failures, but if not detected and addressed promptly, these minor anomalies will gradually accumulate and amplify as production continues, ultimately leading to fabric quality issues, decreased production efficiency, and even more serious equipment failure. Slight fluctuations in cycle time indicate that the loom is gradually deviating from normal operating conditions. This gradual abnormality is often an early warning sign. Although the fluctuations are small, their long-term presence can affect loom stability and fabric uniformity. Therefore, timely detection and response to these subtle changes are crucial to ensuring stable loom operation and product quality.

[0101] In the detection window, the specific steps for analyzing and processing the cycle time changes of the loom during the production process and generating the loom operation cycle time index are as follows:

[0102] In the detection window, first define the loom cycle time, which is the time required for the loom to complete one production cycle. The cycle time is a key indicator of the loom's operating efficiency. The calculation expression is as follows:

[0103]

[0104] , where T cycle (t) is the cycle time at time point t, T production (t) is the total production time at time t, N units (t) is the number of units produced by the loom at time t;

[0105] Through the cycle time T cycle (t) Calculate the variation rate of the cycle time. The calculation expression is as follows:

[0106]

[0107] , where CTV is the cycle time variation rate, T cycle (t-1) is the cycle time at the previous time point t-1;

[0108] Cycle Time Variation (CTV) refers to the degree of fluctuation in a loom's cycle time within a specific timeframe, reflecting the stability of the loom's cycle time during production. Specifically, cycle time variation (CTV) is calculated by comparing the differences in cycle time within consecutive time periods to determine the percentage of relative change. It measures the inconsistency or instability of the loom's cycle time during operation. A low CTV indicates relatively stable loom operation, with production efficiency and fabric quality relatively under control. Conversely, a high CTV indicates that the loom may have experienced fluctuations during certain timeframes, such as uneven tension, equipment failure, or improper operation. Such fluctuations can directly affect fabric quality and may even cause production delays or downtime. Therefore, monitoring and analyzing CTV helps identify potential production issues early, providing a basis for production optimization and fault warnings. It is a key indicator for maintaining production efficiency and consistent product quality.

[0109] After analyzing the variation rate of the cycle time, the loom operation cycle time index is generated based on the cycle time variation rate. The calculation expression is as follows:

[0110]

[0111] , where F cycle is the loom operation cycle time index, ω is the magnification coefficient, e is the natural base number, is the attenuation coefficient;

[0112] Loom operation cycle time index F cycleThis comprehensive indicator, generated based on the loom's Cycle Time Variation (CTV), quantifies the stability and consistency of a loom's cycle time during production. This index reflects the degree of cycle time fluctuation over a specific period. Cycle time fluctuations can be affected by a variety of factors, such as equipment failure, process instability, or raw material issues. By calculating the Cycle Time Variation (CTV) and generating a Cycle Time Index, we can promptly identify minor fluctuations in loom operation and provide early warning of potential equipment failures or quality issues. The Cycle Time Index provides precise monitoring and assessment of loom production status. A high Cycle Time Index indicates significant cycle time fluctuations during production, potentially indicating equipment instability, uneven tension control, or other production anomalies that can impact fabric quality. A low index indicates stable loom operation with minimal cycle time fluctuations, ensuring greater production efficiency and product quality. This index provides production management with a basis for timely adjustments to production processes or equipment maintenance, helping to optimize production processes, improve fabric quality, and reduce the occurrence of defective products.

[0113] The amplification factor controls the impact of the cycle time variation on the cycle time index. A large amplification factor can lead to significant changes in the cycle time index even with relatively small cycle time fluctuations, making the monitoring system more sensitive to subtle abnormal fluctuations. This allows for earlier identification of minor production anomalies or equipment issues, enabling timely adjustments or interventions, helping to improve production stability and fabric quality.

[0114] The attenuation coefficient determines the degree of nonlinearity in the response of the cycle time variation rate to the cycle time index. A larger attenuation coefficient makes fluctuations in the cycle time variation rate more dramatic, especially when the impact of short-term variations on the index is magnified. Conversely, a smaller attenuation coefficient makes the impact of variation on the index more gradual, making the system less sensitive to cycle time fluctuations. This setting helps adjust the speed and degree of the system's response to cycle time fluctuations, allowing the monitoring system to more accurately identify abnormal fluctuations or potential problems in production.

[0115] To prevent the cycle time index from fluctuating sharply in the short term and affecting the entire production process, we then perform a time window weighted average on the cycle time index to retain trend changes over a longer period of time and suppress occasional abnormal fluctuations in the short term. The expression is as follows:

[0116]

[0117] , where I cycle is the weighted cycle time index, F cycle(tj) refers to the loom operation cycle time index at the previous time point tj, M is the total number of data points, j refers to the jth data point, and represents the offset step from the time point t to the past time point, w j is the weight coefficient of each loom operation cycle time index;

[0118] Weight coefficient w of loom operation cycle time index j The role of is to determine the degree of influence of historical cycle time data on the current cycle time index. Specifically, the weight coefficient w j It is used to adjust the contribution of the cycle time index at different historical time points to the current moment when calculating the cycle time index. Generally, data closer to the current moment will be given a higher weight so that it can more sensitively reflect the recent operating status and fluctuations of the loom and quickly identify possible anomalies or problems. Cycle time data at more distant moments will be given a lower weight to reduce its impact on the current evaluation and avoid outdated fluctuations interfering with real-time production analysis. By adjusting w j The size and change mode (such as using attenuation weights) of the cycle time index can optimize the response sensitivity of the cycle time index, so that it can reflect the changes in the production status of the loom in a timely manner while maintaining stability, helping production managers to make more accurate judgments and decisions.

[0119] Based on the weighted cycle time index I cycle , the final loom operation cycle time index of the loom is obtained, and the expression is as follows:

[0120]

[0121] , where CTDI is the loom cycle time index and θ is the adjustment coefficient used to scale the index according to actual production needs.

[0122] Within the detection window, the loom cycle time index, generated by analyzing and processing the changes in the loom's cycle time during production, indicates that a larger value indicates potential, progressive anomalies in the loom's operation. While these anomalies may not immediately impact production efficiency, their accumulation over time could lead to a decline in fabric quality or equipment failure. A smaller and more stable loom cycle time index indicates that the equipment is operating normally, with no significant cycle time fluctuations and all loom parameters remaining within the predetermined range, indicating no anomalies.

[0123] The machine learning model is not limited here. Any machine learning model that can comprehensively analyze the loom tension variation index TVI and the loom operation cycle time index CTDI to generate an abnormality identification coefficient ADC is acceptable. In order to implement the technical solution of the present invention, the present invention provides a specific implementation method:

[0124] The formula for generating the abnormal identification coefficient ADC is as follows:

[0125]

[0126] , where k1 and k2 are the preset proportional coefficients of the loom tension variation index TVI and the loom operation cycle time index CTDI, respectively, and both k1 and k2 are greater than 0.

[0127] The preset scaling factors (k1 and k2) are two fixed constants set in the machine learning model to comprehensively analyze the loom tension variation index (TVI) and the loom cycle time index (CTDI). They are used to adjust the impact of these two indices on the final calculated anomaly identification coefficient (ADC). Specifically, k1 and k2 determine the importance or weighting of the loom tension variation index (TVI) and the loom cycle time index (CTDI) in the calculation of the anomaly identification coefficient (ADC). These preset scaling factors can be used to adjust the contribution of each index to the overall fluctuation coefficient based on the specific operating characteristics and technical requirements of the loom. The values ​​of k1 and k2 are typically set based on historical data, experimental results, or industry standards to ensure that they effectively reflect the actual operating conditions of the loom and help optimize the loom's control and fault detection processes. In this formula, k1 and k2 are both greater than 0, indicating that their contribution to their respective indices is positive, that is, both have a positive impact on the calculation of the anomaly identification coefficient (ADC).

[0128] It can be seen from the calculation expression of the abnormality recognition coefficient that, within the detection window, the larger the loom tension change index expression value generated after analyzing and processing the fluctuation amplitude of the loom tension, and the larger the loom operation cycle time index expression value generated after analyzing and processing the cycle time change of the loom in the production process, the larger the abnormality recognition coefficient expression value generated after analyzing and processing the extracted key features within the detection window, indicating that the probability of minor abnormalities occurring during the operation of the equipment is greater, and vice versa, the probability of minor abnormalities occurring during the operation of the equipment is smaller.

[0129] Based on the evaluation results of the machine learning model, the abnormal operating status of the equipment is classified into two categories: recognizable abnormalities and unrecognizable subtle abnormalities;

[0130] The abnormality identification coefficient generated after analyzing and processing the extracted key features is compared with the pre-set abnormality identification coefficient reference threshold to divide the equipment operation status. The division steps are as follows:

[0131] If the abnormality recognition coefficient is greater than or equal to the preset abnormality recognition coefficient reference threshold, the operating state of the equipment is classified as an unrecognizable minor abnormality;

[0132] If the abnormality identification coefficient is less than a preset abnormality identification coefficient reference threshold, the operating state of the equipment is classified as an identifiable abnormality;

[0133] Identifiable anomalies refer to those obvious and significant abnormal changes, which usually manifest as large fluctuations or mutations. These abnormal changes can be effectively identified and handled under the standard monitoring frequency; unidentifiable subtle anomalies refer to those tiny abnormal changes. These anomalies will not immediately affect the overall operating efficiency of the equipment, but will gradually accumulate over time and have a potential impact on product quality.

[0134] For identifiable anomalies, continue to collect and analyze data at the preset monitoring frequency. By maintaining the existing frequency, we ensure that significant equipment failures can be discovered and resolved in a timely manner to prevent the problem from escalating.

[0135] For identifiable anomalies, data collection and analysis will continue at the pre-set monitoring frequency to ensure that significant equipment failures are promptly identified and effectively resolved. By maintaining the current frequency, the system continuously monitors equipment operating status, capturing significant abnormal fluctuations and quickly identifying faults. This allows for early resolution of problems, preventing them from expanding or worsening, impacting production efficiency and fabric quality, minimizing equipment downtime, and reducing repair costs. This not only helps ensure production stability but also maximizes process safety and product quality.

[0136] For subtle anomalies that are difficult to identify, the monitoring frequency is dynamically increased based on the evaluation results of the machine learning model to capture more production data points, thereby enhancing sensitivity to subtle anomalies and promptly identifying their potential impact;

[0137] The specific steps to dynamically increase monitoring frequency based on the evaluation results of the machine learning model, capture more production data points, and thus enhance sensitivity to subtle anomalies and promptly identify their potential impact are as follows:

[0138] The generated abnormality identification coefficient ADC is compared with the preset abnormality identification coefficient reference threshold to obtain the degree of deviation. The calculation expression is as follows:

[0139]

[0140] , where ΔADC is the abnormal deviation value, ADC ref is the reference threshold of the abnormal identification coefficient;

[0141] Based on the abnormal deviation value, if a minor abnormality is detected, the monitoring frequency will be dynamically adjusted. In order to enhance the sensitivity to minor abnormalities, the monitoring frequency will be adjusted according to the deviation degree of the current abnormal recognition coefficient. The calculation expression is as follows:

[0142]

[0143] , where DFR new is the adjusted detection frequency, DFR base is the preset monitoring frequency, τ is the frequency adjustment coefficient;

[0144] The frequency adjustment factor, τ, controls the sensitivity and amplitude of monitoring frequency changes, especially when minor anomalies occur. It achieves dynamic adjustments by adjusting the system's response to the anomaly deviation, ΔADC. When minor anomalies occur in equipment operation, the frequency adjustment factor determines whether the system needs to increase the monitoring frequency to capture more subtle changes. A larger τ value makes the system more sensitive to minor anomalies and rapidly increases the monitoring frequency to improve its ability to identify potential problems. A smaller τ value results in a slower system response and smaller adjustments, potentially missing minor but cumulative anomalies. Therefore, the frequency adjustment factor plays a crucial role in ensuring production stability and quality consistency, enabling the system to flexibly respond to changing production conditions without imposing excessive computational burdens.

[0145] After dynamically adjusting the monitoring frequency, more data points are collected at a higher frequency. Each collected data point will be further used to analyze the operating status of the equipment. At this time, the updated anomaly recognition coefficient is calculated based on each new data point. The calculation expression is as follows:

[0146]

[0147] , where X y is the yth collected data point, indicating the yth production data point collected at the new frequency, Y is the total number of data points collected at the adjusted monitoring frequency, and X mean is the mean of the current data set, taking into account all newly collected data points, σ X is the standard deviation of the data set, ADC new is the updated anomaly identification coefficient;

[0148] By calculating the deviation between the newly collected data points and the mean, normalizing it with the standard deviation, and updating the anomaly identification coefficient, we can accurately evaluate minor anomalies in equipment operation and provide a basis for further adjusting the monitoring frequency.

[0149] After collecting new data, we can determine whether the monitoring frequency needs to be further adjusted by calculating the potential impact of anomalies on production in order to promptly identify and address these potential problems. The calculation expression is as follows:

[0150]

[0151] , where Impact potential It is the potential impact, which means the potential impact of minor abnormalities on product quality or production process. y is the anomaly recognition coefficient of the y-th data point, δ y is the weight coefficient of the y-th data point, which is used to indicate the importance of the y-th data point to the final impact, 1+μ×(ΔADC) 2 is the nonlinear adjustment term of the impact, is the adjustment factor used to smooth the impact calculation and ensure that the sensitivity to anomalies is not over-amplified, and μ is the adjustment coefficient used to control the nonlinear relationship of the impact;

[0152] If Impact potential Greater than the preset impact threshold Impact threshold , the feedback mechanism is triggered to adjust the operating status of the equipment and further adjust the monitoring frequency to ensure that minor anomalies can be identified and eliminated in a timely manner to prevent them from causing long-term impact on production.

[0153] In the textile production process, minor anomalies often don't immediately affect equipment efficiency, but they can gradually accumulate over time, affecting fabric quality and even leading to defective products. Existing monitoring systems typically use a fixed frequency for data acquisition, making this approach effective for handling significant faults or anomalies. However, they are less sensitive to minor operational anomalies and can be easily overlooked. This is particularly true when there are slight fluctuations in loom parameters such as tension, speed, and temperature. Although these minor fluctuations won't immediately cause noticeable production failures, they can gradually accumulate, ultimately leading to fluctuations in product quality, such as uneven fabric thickness, texture deviations, or surface blemishes.

[0154] By dynamically increasing the monitoring frequency based on the evaluation results of the machine learning model, the system can automatically adjust the collection frequency according to the deviation between real-time data and historical data. When the machine learning model assesses that the equipment has a subtle abnormal trend, the monitoring frequency will be dynamically adjusted according to the change in the abnormality identification coefficient, increasing the density and frequency of data collection. This adjustment can capture more subtle production data points and enhance the ability to identify subtle anomalies. Through continuous monitoring and timely collection of more data, the system can detect potential quality problems in advance and quickly correct them, thereby preventing minor anomalies from accumulating into serious failures or large-scale defective products. The core role of this step is to greatly improve the system's sensitivity to subtle fluctuations through intelligent data collection and frequency adjustment mechanisms, so that minor anomalies in loom operation can be discovered in time and effective measures can be taken before their impact expands, ensuring production efficiency and product quality consistency.

[0155] The present invention can significantly improve the ability to capture minor anomalies by introducing a mechanism for dynamically adjusting the monitoring frequency. Traditional fixed-frequency monitoring methods often ignore smaller, low-frequency abnormal changes, causing these minor fluctuations to accumulate into quality problems without being identified in time. By combining the intelligent evaluation of the machine learning model and extracting key features such as the loom tension change index and the cycle time change index, the system can keenly capture minor abnormal fluctuations. For example, the tension unevenness of the fabric may not be discovered in the early stages, but as production progresses, these minor fluctuations gradually affect the uneven thickness, texture deviation or surface defects of the fabric. If they can be discovered and adjusted in time at the early stage, the accumulation of quality problems can be effectively avoided, thereby improving the consistency and pass rate of products and avoiding the risk of defective products, rework or scrap.

[0156] The present invention dynamically adjusts the monitoring frequency, and the system can proactively increase the monitoring frequency when it identifies unrecognizable subtle anomalies. This can enhance sensitivity to potential problems while ensuring that resources are not over-consumed. Compared with the traditional method of continuous monitoring at a fixed frequency, this flexible adjustment method can be optimized according to the actual needs of the production environment. For significant faults, the system continues to use the traditional frequency for monitoring to ensure that problems are discovered and resolved in a timely manner; for minor anomalies, the system obtains more data points by increasing the monitoring frequency, thereby accurately capturing these potential impacts and preventing the waste of computing resources caused by overly intensive monitoring data. This method can ensure that the operating status of the equipment is effectively monitored while avoiding unnecessary monitoring burdens and waste of computing resources, thereby improving the efficiency and economy of production monitoring.

[0157] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. A data processing method, characterized in that: The following steps are involved: First, various parameter information during the operation of the textile equipment is collected according to the preset monitoring frequency; Collected textile equipment operating parameters are aggregated in real time to form a data set, and key features reflecting potential subtle anomalies are extracted from the data set; Within the detection window, the extracted key features are analyzed and processed in detail, and the analyzed and processed key features are input into a pre-trained machine learning model, which then performs an intelligent assessment of the abnormal operating status of the equipment. Based on the evaluation results of the machine learning model, the abnormal operating status of the equipment is classified into two categories: normal operating status and minor abnormal operating status; For normal operation, continue to collect and analyze data at the preset monitoring frequency. By maintaining the existing frequency, ensure that significant equipment faults can be discovered and resolved in a timely manner to prevent the problem from escalating. For minor abnormal operating conditions, the monitoring frequency is dynamically increased based on the evaluation results of the machine learning model to capture more production data points, thereby enhancing sensitivity to minor anomalies and promptly identifying their potential impact; Extract key features reflecting potential low-frequency anomalies from the data set. The extracted features include the fluctuation amplitude of the loom tension and the change of the loom cycle time during the production process. Within the detection window, the extracted fluctuation amplitude of the loom tension and the change of the loom cycle time during the production process are analyzed and processed to generate the loom tension variation index and loom operation cycle time index. The loom tension variation index reflects the tension unevenness of the fabric during the weaving process by quantifying the amplitude and frequency of the loom tension fluctuation during the production process; the loom operation cycle time index quantifies the slight changes in the cycle time during the loom production process. The specific steps for analyzing and processing the fluctuation amplitude of loom tension and generating the loom tension variation index are as follows: Within the detection window, the intensity of the loom tension fluctuation is first quantified by calculating the tension variation amplitude factor. The tension variation amplitude factor measures the change in the tension signal at adjacent time points and is analyzed in combination with the maximum deviation of the fluctuation to effectively reflect the amplitude of the tension fluctuation. The calculation expression is as follows: , where is the tension variation factor, It's time The loom tension signal value at the moment, It is time points , It's the previous time point The loom tension signal value at the moment, is the total number of time points, is the weight factor; Next, we extract nonlinear dynamic features from the tension fluctuations. We use the tension fluctuation index to quantify the nonlinear change trend of the tension signal. Through Fourier transform, we decompose the tension signal into different frequency components to capture the periodic fluctuation pattern. The calculation expression is as follows: , where is the tension fluctuation index, is the tension signal component after Fourier transform, that is, the tension signal Frequency in the frequency domain The Fourier transform result of is the adjustment index, is the maximum frequency in the frequency domain, It is the high frequency component attenuation factor, which is used to attenuate the influence of high frequency components. is the adjustment parameter of the attenuation factor; The complexity of tension fluctuation is calculated based on the dynamic model. The tension non-stationarity coefficient is used to measure the variation pattern of loom tension. The tension non-stationarity coefficient is used to capture the time dependence of tension fluctuation and reflect the fluctuation difference of loom tension in different time intervals. The calculation expression is as follows: , where is the tension nonstationarity coefficient, is the rate of change of tension, that is, the rate of change of loom tension over time, Represents the nonlinear index, which is a regulation index. is the natural base, is the attenuation factor, is the total time length of the detection window; Comprehensive tension change amplitude factor , tension fluctuation index and the tension nonstationarity coefficient Generate the loom tension change index, the calculation expression is as follows: , where is the loom tension variation index.

2. A data processing method according to claim 1, characterized in that: After obtaining the loom tension change index and loom operation cycle time index generated after analyzing and processing the key features, the loom tension change index and the loom operation cycle time index are input into a pre-learned machine learning model, and an abnormality recognition coefficient is generated by the machine learning model. The abnormality recognition coefficient is used to perform an intelligent evaluation of the equipment operation status.

3. A data processing method according to claim 2, characterized in that: The abnormality identification coefficient generated after analyzing and processing the extracted key features is compared with the pre-set abnormality identification coefficient reference threshold to divide the equipment operation status. The division steps are as follows: If the abnormality recognition coefficient is greater than or equal to the preset abnormality recognition coefficient reference threshold, the operating state of the equipment is classified as an unrecognizable minor abnormality; If the abnormality recognition coefficient is less than a preset abnormality recognition coefficient reference threshold, the operating state of the equipment is classified as a recognizable abnormality.

4. A data processing method according to claim 3, characterized in that: The specific steps to dynamically increase monitoring frequency based on the evaluation results of the machine learning model, capture more production data points, and thus enhance sensitivity to subtle anomalies and promptly identify their potential impact are as follows: The generated anomaly identification coefficient Compare with the preset abnormal identification coefficient reference threshold to obtain its deviation degree. The calculation expression is as follows: , where is the abnormal deviation value, is the reference threshold of the abnormal identification coefficient; Based on the abnormal deviation value, if a minor abnormality is detected, the monitoring frequency will be dynamically adjusted. In order to enhance the sensitivity to minor abnormalities, the monitoring frequency will be adjusted according to the deviation degree of the current abnormal recognition coefficient. The calculation expression is as follows: , where is the adjusted detection frequency, is the preset monitoring frequency, is the frequency adjustment coefficient; After dynamically adjusting the monitoring frequency, more data points are collected at a higher frequency. Each collected data point will be further used to analyze the operating status of the equipment. At this time, the updated anomaly recognition coefficient is calculated based on each new data point. The calculation expression is as follows: , where It is The collected data points represent the first data points collected at the new frequency. Production data points, is the total number of data points collected at the adjusted monitoring frequency, is the mean of the current collected data set, is the standard deviation of the data set, is the updated anomaly identification coefficient; After collecting new data, we can determine whether the monitoring frequency needs to be further adjusted by calculating the potential impact of anomalies on production in order to promptly identify and address these potential problems. The calculation expression is as follows: , where It is the potential impact, It is The anomaly recognition coefficient of the data point is It is The weight coefficient of the data point is used to represent the The importance of each data point to the final impact, is the nonlinear adjustment term of the influence, is the adjustment coefficient, which is used to control the nonlinear relationship of the influence; like Greater than the preset impact threshold The feedback mechanism is triggered to adjust the operating status of the equipment and further adjust the monitoring frequency to ensure that minor anomalies can be identified and eliminated in a timely manner to prevent them from causing long-term impact on production.

5. A data processing method according to claim 1, characterized in that: In the detection window, the specific steps for analyzing and processing the cycle time changes of the loom during the production process and generating the loom operation cycle time index are as follows: In the detection window, first define the cycle time of the loom, that is, the time required for the loom to complete one cycle of production. The calculation expression is as follows: , where It's at the time The cycle time of the moment, It's at the time The total production time of the moment, It's at the time The number of units produced by the loom at any given moment; Through cycle time Calculate the variation rate of cycle time. The calculation expression is as follows: , where is the cycle time variation rate, It's the previous time point The cycle time of the moment; After analyzing the variation rate of the cycle time, the loom operation cycle time index is generated based on the cycle time variation rate. The calculation expression is as follows: , where is the loom cycle time index, is the amplification factor, is the natural base, is the attenuation coefficient; To prevent the cycle time index from fluctuating sharply in the short term and affecting the entire production process, we then perform a time window weighted average on the cycle time index to retain trend changes over a longer period of time and suppress occasional abnormal fluctuations in the short term. The expression is as follows: , where is the weighted period time index, Refers to the previous time point The loom cycle time index at time t, is the total number of data points, It refers to the data points, representing the time from The offset step of the moment to the past time point, is the weight coefficient of each loom operation cycle time index; Based on the weighted cycle time index The final loom cycle time index of the loom is obtained, and the expression is as follows: , where is the loom cycle time index, is the adjustment factor used to scale the index according to actual production needs.

Images