Intelligent fault diagnosis and alarm system for high-temperature and high-pressure dyeing machine

Abstract

The invention discloses an intelligent fault diagnosis and alarm system for a high-temperature and high-pressure dyeing machine, and particularly relates to the technical field of intelligent fault diagnosis of industrial equipment, which comprises a data acquisition module, a data preprocessing module, a data processing module, a data processing module, a data processing module and an alarm module, and is characterized in that the data acquisition module is used for acquiring operation data of the high-temperature and high-pressure dyeing machine in real time; the system comprises a data acquisition module used for acquiring data, a clustering analysis module used for identifying similar equipment groups, an analysis module used for carrying out feature deviation analysis and judging whether the current dyeing equipment has potential fault deviation, an adjustment module used for constructing a dynamic threshold value model, and a diagnosis alarm module used for carrying out fault diagnosis alarm operation. According to the method, the actual operation state of the equipment is continuously monitored, dynamic adjustment is carried out in combination with historical data, the appropriate Gaussian process regression kernel function type is automatically selected, the subtle change of the equipment state can be accurately captured, and the problem of false alarm or missing alarm caused by the fact that a threshold value is not suitable for the current equipment state change in a traditional method is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent fault diagnosis of industrial equipment, and more specifically, to an intelligent fault diagnosis and alarm system for a high-temperature and high-pressure dyeing machine. Background Art

[0002] With the continuous improvement of the degree of industrialization, especially in the fields of dyeing and chemical industry, the demand for equipment operation status monitoring and fault diagnosis is increasing. As a key industrial equipment, the operation status of high temperature and high pressure dyeing machine directly affects production efficiency, product quality and safety. However, since the equipment is in an extreme working environment, the temperature, pressure, vibration and other factors fluctuate greatly. The traditional manual inspection and early warning mechanism often has problems of delayed response and low accuracy, and cannot monitor the potential failure of the equipment in real time. Most of the existing diagnostic methods are based on regular maintenance or threshold setting. These methods cannot make full use of real-time data and lack the ability to dynamically adjust to changes in equipment status.

[0003] How to conduct dynamic monitoring and intelligent fault diagnosis based on the real-time operation data of the equipment has become an important issue in modern industrial equipment management. Traditional fault diagnosis methods based on statistical analysis can usually only provide overall status information of the equipment, but ignore local faults or abnormal changes that may occur in the equipment during long-term operation. These changes may not be significant in the early stage, but if not discovered in time, they may lead to the expansion of equipment failures, ultimately affecting production safety and equipment life. Therefore, the present invention proposes an intelligent fault diagnosis and alarm system for high-temperature and high-pressure dyeing machines in order to solve the above problems. Summary of the invention

[0004] To achieve the above object, the present invention provides the following technical solutions: An intelligent fault diagnosis and alarm system for a high-temperature and high-pressure dyeing machine includes a data acquisition module, a data preprocessing module, a cluster analysis module, a deviation analysis module, an adjustment module, and a diagnosis and alarm module; The data acquisition module is used to collect the operation data of the high temperature and high pressure dyeing machine in real time to obtain the original data; The data preprocessing module is used to denoise, standardize and fill missing values ​​of the original data, generate structured time series data, and obtain preprocessed data; The cluster analysis module is used to perform cluster analysis on the pre-processed data of the current dyeing equipment and identify similar equipment groups to other dyeing equipment in normal state; The deviation analysis module extracts the statistical characteristics of the current dyeing equipment based on the cluster analysis results, and compares it with the historical data of similar equipment groups to perform feature deviation analysis to determine whether the current dyeing equipment has potential faults; When there is a potential fault in the current dyeing equipment, the adjustment module switches the preset static threshold alarm mechanism to a dynamic threshold alarm mechanism, and constructs a dynamic threshold model for dynamically adjusting the alarm threshold; The diagnosis alarm module performs fault diagnosis alarm operations based on the static threshold alarm mechanism or dynamic threshold alarm mechanism currently adopted.

[0005] In a preferred embodiment, the data preprocessing module includes a data denoising unit for denoising the original data, a data standardization unit for standardizing the original data, and a data missing value filling unit for filling missing values ​​in the original data.

[0006] In a preferred embodiment, the clustering analysis module uses a K-means clustering algorithm to convert the operating data of the current dyeing device under a preset static threshold alarm mechanism and in a normal state into a first vector, and converts the operating data of other dyeing devices under a preset static threshold alarm mechanism and in a normal state into a second vector, and performs K-means clustering on all the second vectors with the first vector to identify similar device groups between the current dyeing device and other dyeing devices.

[0007] In a preferred embodiment, during feature deviation analysis, the statistical characteristics of the current dyeing equipment are obtained to obtain statistical feature data group one, and the statistical characteristics corresponding to the historical data of a similar equipment group are obtained to obtain statistical feature data group two. Each statistical feature data is preset with a deviation threshold. A range conversion is performed based on the statistical feature data group two and the deviation threshold corresponding to each statistical feature data to obtain a statistical feature range data group, and the statistical feature data group one is compared with the statistical feature range data group.

[0008] In a preferred embodiment, judging whether there is a potential fault in the current dyeing equipment refers to: If, in the comparison result between statistical feature data group one and statistical feature range data group, the statistical feature data in statistical feature data group one exceeds the total number of statistical features of the corresponding statistical feature range in the statistical feature range data group, if the total number of statistical features is not zero, it is determined that the current dyeing equipment has a potential fault; if the total number of statistical features is zero, it is determined that the current dyeing equipment does not have a potential fault.

[0009] In a preferred embodiment, the dynamic threshold model is constructed by a Gaussian process regression algorithm. The constructed dynamic threshold model predicts the normal operating parameters of the current dyeing equipment and provides a confidence interval for each predicted value of the normal operating parameter, indicating the credibility range of the predicted value.

[0010] In a preferred embodiment, when the diagnostic alarm module performs fault diagnosis alarm operation based on the dynamic threshold alarm mechanism, if the actual operation data of the current dyeing equipment exceeds the confidence interval predicted by the dynamic threshold model, an alarm is issued to remind that the current dyeing equipment has a fault.

[0011] In a preferred embodiment, when constructing the Gaussian process regression algorithm, an RBF kernel or a Matern kernel function is used.

[0012] In a preferred embodiment, before constructing the Gaussian process regression algorithm, the individual smoothness index corresponding to each fault parameter used to measure the rate of change is calculated, and the individual regularity index used to evaluate the regularity of each fault parameter is calculated at the same time. Then, the comprehensive smoothness index and the comprehensive regularity index are calculated separately by weighting the fault parameter variance. Fuzzy reasoning is used, and the comprehensive smoothness index and the comprehensive regularity index are used as input data together. The kernel function type used when constructing the Gaussian process regression algorithm is used as output data, and the kernel function type used when constructing the Gaussian process regression algorithm is inferred.

[0013] Technical effects and advantages of the present invention: The present invention can perform intelligent monitoring based on the real-time operating data (such as temperature, pressure, vibration, etc.) of the high-temperature and high-pressure dyeing machine, and analyze the changes in the equipment status in real time. Through the Gaussian process regression (GPR) model, the present invention can dynamically predict the normal working range of the equipment and adjust the early warning threshold in real time. Once the operating data of the equipment exceeds the confidence interval predicted by the model, the system can issue an alarm in time to remind the operator to check and maintain, thereby effectively reducing production stagnation and equipment damage caused by equipment failure. This dynamic threshold adjustment method has higher flexibility and accuracy compared to the traditional static threshold alarm method. By continuously monitoring the actual operating status of the equipment and dynamically adjusting it in combination with historical data, the present invention can accurately capture subtle changes in the equipment status, avoiding the false alarm or missed alarm problem caused by the threshold not adapting to the current equipment status changes in the traditional method.

[0014] The present invention adopts the Gaussian process regression (GPR) algorithm, uses the historical data of the equipment to model the operating status of the equipment, and provides a confidence interval for each predicted value. In this way, the system can not only give the predicted value of the equipment in the current state, but also quantify the uncertainty of the prediction, thereby providing a more reliable basis for equipment failure warning. When dealing with the state of complex systems such as high-temperature and high-pressure equipment, the GPR algorithm can better cope with the nonlinear changes and high volatility of data, and accurately predict the normal operating range of the equipment. Even before the equipment fails, the system can identify potential failure risks through the predicted confidence interval, issue a warning in advance, and avoid sudden failures of the equipment.

[0015] Through the fuzzy reasoning mechanism, the present invention can automatically select the appropriate Gaussian process regression (GPR) kernel function type (such as RBF kernel or Matern kernel) according to the comprehensive smoothness index and comprehensive regularity index of the equipment. The fuzzy reasoning system is based on the fuzzification results of the input comprehensive smoothness index and regularity index, and performs reasoning through preset rules to automatically select the kernel function type that best suits the current equipment operation status, thereby ensuring that the GPR model can provide the best prediction accuracy under different working conditions. This automatic selection of kernel functions improves the system's adaptive ability, enabling it to cope with complex situations of different equipment, different working conditions and different fault types, and further improves the accuracy of diagnosis and the flexibility of the system.

[0016] The present invention can not only achieve early warning of equipment, but also early intervention of potential failures by comprehensively analyzing the equipment's operating status and historical data. Through real-time monitoring and analysis of key parameters such as temperature, pressure, and vibration, the system can identify abnormal patterns and risk points before equipment failure occurs, and provide early warning prompts. Operators can inspect and repair equipment before problems occur, reducing downtime and maintenance costs and improving production efficiency. Unlike traditional fault diagnosis methods, the present invention has high early warning accuracy and sensitivity, and can detect problems in time at the early stages of equipment failure, reducing serious equipment failures and production interruptions caused by ignoring early warnings.

[0017] Through continuous intelligent monitoring and dynamic prediction, the present invention can significantly improve the operational reliability and safety of the equipment. The intelligent fault diagnosis and early warning system ensures that the equipment can promptly detect potential problems and issue early warnings during operation, thus preventing the equipment from being in an abnormal state for a long time. Through regular maintenance and timely intervention, the service life of the equipment is extended and the risk of interruption in the production process is greatly reduced. In addition, the system can effectively reduce the need for human intervention and reduce the probability of equipment failure due to human misjudgment, thereby improving the overall safety of the equipment and the continuity of production. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings; Figure 1 The schematic diagram of the intelligent fault diagnosis and alarm system of a high-temperature and high-pressure dyeing machine in the present invention.

[0019] Figure 2 It is a schematic diagram of the data preprocessing module in the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Reference Figure 1-2 The following examples are obtained:

[0022] Embodiment 1:

[0023] With the advancement of industrialization, especially in the fields of textiles and chemicals, high temperature and high pressure dyeing machines have become one of the key equipment in the production process. These dyeing equipment undertake important dyeing tasks in production and require them to maintain stable operation under high temperature and high pressure conditions. However, due to the harsh working environment of the equipment, factors such as mechanical failure, equipment aging, and sensor failure may cause equipment abnormalities. Once these problems occur, they are often difficult to be discovered in time, resulting in production stagnation, product quality decline, and even serious equipment damage, causing huge economic losses.

[0024] Traditional fault diagnosis methods mostly rely on manual inspections, regular maintenance, and empirical judgment, which have great limitations. Manual inspections are easily affected by the operator's experience and cannot monitor the operating status of the equipment in real time and comprehensively. Regular maintenance may not be able to accurately predict the time when equipment failures will occur, and the maintenance cost is high and the cycle is long. Therefore, how to achieve intelligent, real-time monitoring and fault diagnosis of dyeing machine equipment has become a technical problem that needs to be solved urgently in the textile, chemical and other industries.

[0025] At present, with the rapid development of sensor technology, Internet of Things, data analysis and artificial intelligence (AI) technology, the automation and intelligence of equipment fault diagnosis has gradually become possible. In particular, data-driven predictive maintenance methods have gradually emerged. By analyzing the historical operation data of equipment, possible equipment failures can be predicted in advance, thereby improving the operating efficiency and safety of the equipment and reducing downtime and maintenance costs.

[0026] However, although there are some equipment diagnosis systems based on data analysis on the market, most systems still rely on fault alarm methods based on preset thresholds, and cannot effectively consider the dynamic changes of equipment operating parameters and the similarities between equipment groups for the operating status of complex high-temperature and high-pressure dyeing machines. Existing methods are usually unable to flexibly adjust and accurately warn the actual operating status of different equipment in a complex and dynamic environment. Therefore, there is an urgent need for a new, intelligent fault diagnosis and alarm system that can provide accurate fault diagnosis and dynamic warning mechanisms based on the real-time and historical data of the equipment, and can automatically adjust and optimize according to the working status of the equipment.

[0027] It is against this background that the present invention is proposed. Through intelligent data analysis methods, combined with advanced technologies such as Gaussian process regression (GPR), it aims to provide an intelligent fault diagnosis and alarm system for high-temperature and high-pressure dyeing machines. Its main significance is reflected in the following aspects: Real-time fault diagnosis and early warning: The present invention collects the operation data (such as temperature, pressure, vibration, etc.) of the dyeing machine in real time and uses data analysis technology to achieve real-time monitoring of the equipment status. Through the dynamic adjustment of the Gaussian process regression (GPR) model, it can timely identify the abnormal status of the equipment and issue a fault alarm according to the actual operation situation, greatly improving the fault diagnosis accuracy and response speed of the equipment.

[0028] Dynamic adjustment of warning threshold: The present invention can switch to a dynamic threshold mechanism instead of using the traditional fixed threshold setting. Through cluster analysis of device groups, device groups similar to the target device are identified, and the warning threshold is dynamically adjusted based on the historical data of these similar devices. This makes the warning system more flexible and can adapt to changes in different devices and different operating conditions, improving the accuracy of warnings.

[0029] Statistical deviation analysis and intelligent adjustment mechanism: The present invention can identify whether the equipment is in normal working condition by analyzing the statistical characteristic deviation of the equipment data. Through the automatic adjustment mechanism, the operating state of the equipment is adjusted according to the deviation between the real-time monitoring data and the preset target, thereby realizing intelligent equipment maintenance and fault prediction. This not only reduces manual intervention, but also effectively prolongs the service life of the equipment and reduces the equipment failure rate.

[0030] Improve production efficiency and safety: Since the system can detect potential failure risks in a timely manner and issue early warnings, it can reduce production interruptions and maintenance downtime caused by equipment failures. In addition, the system can provide operators with detailed fault diagnosis reports, which improves operators' understanding of equipment conditions and ensures production safety.

[0031] Reduce manual intervention and operating costs: This invention reduces the need for manual inspections and the probability of human error by realizing intelligent equipment monitoring and fault diagnosis. At the same time, the data-driven dynamic adjustment and early warning mechanism not only improves the accuracy of fault warnings, but also reduces frequent manual intervention, further reducing operating costs.

[0032] Through the implementation of the present invention, the operating reliability and intelligence level of the high-temperature and high-pressure dyeing machine can be effectively improved, and the equipment management in the textile, chemical and other industries can be promoted to develop in the direction of automation and intelligence, thereby providing strong technical support for the improvement of production efficiency, safety assurance and cost savings in related industries.

[0033] The present invention proposes an intelligent fault diagnosis and alarm system for a high-temperature and high-pressure dyeing machine, comprising a data acquisition module, a data preprocessing module, a cluster analysis module, a deviation analysis module, an adjustment module, and a diagnosis and alarm module; the data acquisition module, the data preprocessing module, the cluster analysis module, the deviation analysis module, the adjustment module, and the diagnosis and alarm module are communicatively connected.

[0034] The data acquisition module is used to collect the operating data of the high-temperature and high-pressure dyeing machine in real time to obtain the original data; the data acquisition module is used to collect the operating data of the high-temperature and high-pressure dyeing machine in real time. The operating data includes key parameters such as temperature, pressure, and vibration. The data acquisition module is connected to the dyeing machine through sensors to obtain various operating indicators of the equipment and generate original data.

[0035] The data preprocessing module is used to denoise, standardize and fill in missing values ​​on the original data, generate structured time series data, and obtain preprocessed data; the data preprocessing module ensures the quality of the data through signal denoising, normalization, and missing data filling, eliminates the influence of external noise or abnormal data, and facilitates subsequent analysis.

[0036] The clustering analysis module is used to perform cluster analysis on the pre-processed data of the current dyeing equipment and identify the equipment groups that are similar to other dyeing equipment under normal conditions; clustering algorithms such as K-means or DBSCAN are used to group the equipment operation data to find the most similar equipment groups under normal working conditions and form a benchmark data set to provide a reference for subsequent analysis.

[0037] Based on the cluster analysis results, the deviation analysis module extracts the statistical characteristics of the current dyeing equipment, compares them with the historical data of similar equipment groups, performs feature deviation analysis, and determines whether the current dyeing equipment has potential faults; by calculating the mean, standard deviation, deviation and other statistical characteristics of the equipment parameters, feature deviation analysis is performed to determine whether the operating status of the current dyeing equipment is abnormal, and then identify potential fault risks.

[0038] When there is a potential fault in the current dyeing equipment, the adjustment module switches the preset static threshold alarm mechanism to a dynamic threshold alarm mechanism, and builds a dynamic threshold model for dynamically adjusting the alarm threshold; by adaptively adjusting the early warning threshold of the equipment, it ensures that the system can respond to the actual operation of the equipment in real time.

[0039] The diagnostic alarm module performs fault diagnosis alarm operations based on the static threshold alarm mechanism or dynamic threshold alarm mechanism currently used. After the adjustment module adjusts the threshold, the real-time operating status of the equipment is monitored. When the equipment operating parameters exceed the adjusted warning threshold, the system immediately triggers an alarm to remind the operator that there is a potential fault in the equipment and to take preventive measures in time to avoid further expansion of the fault. If the static threshold alarm mechanism is currently used, the real-time operating status parameters of the equipment are compared with the corresponding static thresholds respectively. When the comparison results meet the alarm conditions, an alarm is issued.

[0040] The data preprocessing module includes a data denoising unit for denoising the raw data, a data standardization unit for standardizing the raw data, and a data missing value filling unit for filling missing values ​​in the raw data. The main task of the data denoising unit is to remove noise in the acquisition process to ensure that the data reflects the true state of the device. Noise usually comes from sensor failure, environmental interference or other non-device related factors, which may seriously affect the accuracy of the analysis. Existing technology can use wavelet transform: commonly used for signal denoising, by decomposing the signal into components of different frequency bands, and then restoring the signal by removing high-frequency noise components. Filter method: For example, using a Kalman filter or a median filter, random noise can be effectively removed. Empirical mode decomposition (EMD): an adaptive denoising method that can extract effective components from multiple signals and eliminate noise.

[0041] The data standardization unit is used to convert various sensor data to the same scale to avoid deviations in analysis results due to dimensional differences in different parameters. Standardized data is conducive to the processing and calculation of subsequent algorithms (such as cluster analysis, machine learning models, etc.). Existing technologies can use Z-score standardization: standardize the data so that the mean of the data is 0 and the standard deviation is 1; Min-Max standardization: normalize the data to a fixed interval (usually [0,1]).

[0042] The missing value filling unit is used to process the missing parts in the data to ensure that the entire data set is complete for subsequent analysis. Missing data may be caused by a variety of reasons, such as sensor failure, network problems, etc. The processing of missing values ​​has an important impact on the final analysis results. Existing technologies can be used: Mean / median filling: This is a simple and common method of filling missing values ​​by filling the missing data with the mean or median of the column (or row). KNN filling: Filling missing values ​​based on the K nearest neighbor algorithm. Filling is performed by finding neighbor data similar to the missing data. This method is suitable when there is a certain correlation between the data. Regression filling: Predicting missing values ​​by establishing a regression model, usually using data from other columns as input features to predict missing data.

[0043] The clustering analysis module uses the K-means clustering algorithm to convert the operating data of the current dyeing equipment under the preset static threshold alarm mechanism and normal state into a first vector, and converts the operating data of other dyeing equipment under the preset static threshold alarm mechanism and normal state into a second vector, and performs K-means clustering on all the second vectors with the first vector to identify similar equipment groups between the current dyeing equipment and other dyeing equipment.

[0044] The main goal of the cluster analysis module is to cluster the operating data of the current dyeing equipment with the operating data of other equipment through the K-means clustering algorithm, identify the equipment group that is most similar to the current equipment status, and thus provide basic data for subsequent fault detection and early warning. Through cluster analysis, the similarities between different equipment under normal working conditions can be found, thus providing a reference for fault diagnosis. The specific steps are: Data preparation: First, the system uses the data acquisition module to obtain the operating data of the current dyeing equipment in normal working state in real time. These data include key parameters such as temperature, pressure, vibration, etc. In order to perform cluster analysis, all these operating data need to be converted into vectors. The operating data of each device (such as temperature, pressure, vibration and other parameters) will be regarded as a high-dimensional vector, in which each dimension represents a characteristic parameter of a device.

[0045] Constructing the first vector: The operating data of the current dyeing equipment under the preset static threshold alarm mechanism and normal state is converted into the first vector. This vector contains the values ​​of various operating indicators (such as temperature, pressure, etc.) of the equipment under normal conditions within a given time period.

[0046] Constructing the second vector: The operating data of other dyeing equipment under the preset static threshold alarm mechanism and normal state are converted into the second vector. These vectors are also composed of the normal working state data of each equipment, representing the characteristic parameters of the equipment.

[0047] K-means clustering algorithm: Initialization: First randomly select K center points (i.e. cluster centers), which represent the initial "cluster clusters". Assign data points: Assign the second vectors of all devices to the cluster where the center point closest to them is located, that is, assign them to different groups according to the operating data characteristics of the equipment. Update cluster centers: Calculate the feature mean of the equipment in each cluster, and update the cluster centers until the cluster centers converge, that is, they no longer change. Result: After clustering is completed, the system will be able to identify a group of equipment similar to the current dyeing equipment, and the parameter characteristics of the equipment in this group under normal working conditions have a high similarity.

[0048] The K-means algorithm divides the data set into K clusters, each of which contains data points with similar characteristics. The core idea is to perform clustering by minimizing the distance between each data point and the center point (mean) of the cluster to which it belongs. The goal of clustering is to minimize the total distance (usually the Euclidean distance) of the samples within each cluster, that is, to make the similarity of the data points within the cluster as high as possible and the difference between the data points between clusters as large as possible. By performing K-means clustering on the operating data of the current dyeing equipment and the data of other equipment, the system can identify groups of equipment under similar working conditions as the current equipment. This provides basic data for subsequent deviation analysis. If the operating data of the current equipment has a significant deviation in a certain group, it can indicate that the equipment may be in an abnormal state.

[0049] The identified similar device groups can analyze their working rules and patterns based on the historical data of the devices in the group. By comparing the actual operating data of the current device with the historical data of similar device groups, the system can provide early warning of potential equipment failure risks. Through cluster analysis, the system can distinguish the working characteristics of different devices and avoid using a unified static threshold to judge abnormalities of all devices. The "normal" state of each device is defined in its own similar group, which not only improves the accuracy of the warning, but also avoids overly broad or overly strict threshold settings.

[0050] During feature deviation analysis, the statistical characteristics of the current dyeing equipment are obtained to obtain statistical feature data group one, and the statistical characteristics corresponding to the historical data of a similar equipment group are obtained to obtain statistical feature data group two. Each statistical feature data is preset with a deviation threshold. A range conversion is performed based on the statistical feature data group two and the deviation threshold corresponding to each statistical feature data to obtain a statistical feature range data group, and the statistical feature data group one is compared with the statistical feature range data group.

[0051] The main goal of the feature deviation analysis module is to compare the statistical characteristics of the current dyeing equipment with the historical data of similar equipment groups to determine whether the current equipment has potential failures. By comparing the degree of deviation of each statistical feature, the system can identify possible equipment anomalies or failures and trigger corresponding warnings. Specifically, the system determines whether the current equipment is operating abnormally by comparing the calculated statistical characteristics with the preset deviation threshold.

[0052] Obtain the statistical characteristics of the current dyeing equipment (statistical characteristic data group 1): Statistical characteristics are characteristic indicators used to describe the equipment operation data, such as mean, standard deviation, skewness, kurtosis, etc. First, the cleaned and standardized equipment data provided by the data preprocessing module is statistically calculated to obtain the statistical characteristics of the current equipment. For example, for the temperature, pressure and vibration signals of the current equipment, common statistical characteristics include: Temperature: mean, standard deviation, skewness. Pressure: mean, standard deviation, kurtosis. Vibration: mean, standard deviation, frequency distribution.

[0053] Obtaining the statistical characteristics of historical data of similar equipment groups (statistical characteristic data group 2): Through the historical data of similar equipment groups identified by the cluster analysis module, the system extracts the statistical characteristics of these equipment groups in normal state to obtain a set of statistical characteristic data representing similar equipment groups. For example, the statistical characteristics of temperature, pressure, and vibration of similar equipment groups will also be extracted to obtain a set of corresponding data (including mean, standard deviation, skewness, etc.).

[0054] Set the deviation threshold: For each statistical feature, the system presets a deviation threshold, which is set based on the working range of the device under normal working conditions. The deviation threshold is used to determine whether the statistical characteristics of the device during operation exceed the normal fluctuation range. For example: Temperature: The temperature fluctuation range of normal equipment is 30℃±5℃, and the temperature deviation threshold is set to 5℃. When the temperature of the device exceeds 30℃±5℃, it is considered that the device may have a fault. Pressure: The normal fluctuation range of pressure is 0.5MPa±0.1MPa, and the deviation threshold is set to 0.1MPa. When the actual pressure exceeds this range, it indicates that the device is abnormal. Vibration: The normal fluctuation range of the vibration signal is 1.2mm / s±0.3mm / s, and the deviation threshold is set to 0.3mm / s. When the vibration exceeds this range, it indicates that there may be a problem with the operating status of the device.

[0055] Range conversion: Combine the statistical characteristics of the historical data group with the preset deviation threshold to obtain a "range conversion" data group within the normal fluctuation range. For example: Temperature: According to the statistical characteristics of the historical group and the deviation threshold, the normal fluctuation range of temperature is converted to [25℃, 35℃]. If the temperature of the current device exceeds this range, it indicates deviation. Pressure: The standard deviation of the pressure of the historical data group is 0.05MPa, and the set deviation threshold is 0.1MPa. The normal fluctuation range is [0.4MPa, 0.6MPa]. Exceeding this range is a deviation. Vibration: The normal fluctuation range of vibration is [0.9mm / s, 1.5mm / s]. If the vibration of the current device exceeds this range, it means that its operating status is abnormal.

[0056] Judging whether there is a potential fault in the current dyeing equipment means: If, in the comparison result between statistical feature data group one and statistical feature range data group, the statistical feature data of statistical feature data group one exceeds the total number of statistical features of the corresponding statistical feature range in the statistical feature range data group, if the total number of statistical features is not zero, it is determined that the current dyeing equipment has a potential fault; if the total number of statistical features is zero, it is determined that the current dyeing equipment does not have a potential fault.

[0057] By comparing each characteristic value of the current device (statistical characteristic data group 1) with the historical data of a group of similar devices (statistical characteristic range data group), the system determines whether the characteristic value of the current device exceeds the allowable range. Example: Temperature: Assume that the mean temperature of the current dyeing device is 80°C and the standard deviation is 1°C. The mean temperature of the group of similar devices in the historical data is 78°C and the standard deviation is 0.8°C. The preset deviation threshold is ±2°C. Through range conversion, the mean temperature (80°C) and standard deviation (1°C) of the current device are both within the allowable range (the mean temperature of the historical device is ±2°C, that is, between 76°C and 80°C), so the temperature characteristic of the device does not exceed the deviation threshold. Pressure: Assume that the pressure of the current dyeing device is 10bar, the mean pressure of the historical device group is 9.8bar, the standard deviation is 0.3bar, and the preset deviation threshold is ±0.5bar. The pressure of the current device (10bar) exceeds the pressure range of the historical device group (9.3bar to 10.3bar). Therefore, the pressure characteristic value exceeds the allowable deviation threshold.

[0058] Potential fault judgment: In the comparison results, if some statistical feature data of the current device exceeds the corresponding normal range (that is, the data in the statistical feature range data group), the system will count the total number of these features that exceed the range. If the total number of statistical features that exceed the threshold is not zero, it means that some features of the current device exceed the preset normal range, and there may be potential faults. The system will determine that the current device has potential faults. If the total number of statistical features that exceed the threshold is zero, it means that all features of the current device are within the normal range, and the system determines that the current device has no potential faults.

[0059] For example: suppose that the statistical characteristic data group 1 of the current dyeing equipment is: temperature: mean 80°C, standard deviation 1°C; pressure: 10bar; vibration: 1.2mm / s; and the similar equipment group is: temperature: mean 78°C, standard deviation 0.8°C, deviation threshold is ±2°C; pressure: mean 9.8bar, standard deviation 0.3bar, deviation threshold is ±0.5bar; vibration: mean 1mm / s, standard deviation 0.2mm / s, deviation threshold is ±0.5mm / s.

[0060] Judgment process: Comparison of temperature: The current average temperature is 80°C, which is in line with the historical equipment range (76°C to 80°C) and does not exceed the deviation threshold. Comparison of pressure: The current pressure is 10 bar, which exceeds the historical equipment range (9.3 bar to 10.3 bar) and exceeds the deviation threshold. Comparison of vibration: The current vibration is 1.2 mm / s, which exceeds the historical equipment range (0.8 mm / s to 1.2 mm / s) and also exceeds the deviation threshold.

[0061] Conclusion: The current device has two statistical characteristics (pressure and vibration) that exceed the preset range. Therefore, the total number of characteristic deviations is 2. The system determines that the current device has a potential fault.

[0062] The dynamic threshold model is constructed by using the Gaussian process regression algorithm. The constructed dynamic threshold model predicts the normal operating parameters of the current dyeing equipment and provides a confidence interval for each predicted value of the normal operating parameter, indicating the credibility range of the predicted value.

[0063] The dynamic threshold model is constructed using the Gaussian process regression (GPR) algorithm to predict the normal operating parameters of the current dyeing equipment and provide a confidence interval for each predicted value. The model can dynamically adjust the warning threshold of the equipment based on historical data to ensure real-time and accurate monitoring of the equipment status. When the actual operating data of the equipment exceeds the confidence interval predicted by the model, the system will trigger an alarm to remind the operator of the potential risk of failure. The steps for building a Gaussian process regression (GPR) model: Taking temperature as an example, a detailed description of how to use the Gaussian process regression algorithm to build a dynamic threshold model: Data preparation: First, collect historical temperature data. This data can come from the normal operating status of different devices, covering different working environments, loads, etc. Assuming that enough historical data has been collected, this data will be used as the training set for the Gaussian process regression model.

[0064] Select kernel function: The key to the Gaussian process regression model is to select a suitable kernel function, which defines the similarity between data points. The kernel function is set to RBF kernel (radial basis function kernel) or Matern kernel. The choice of kernel function depends on the smoothness and volatility of the data. If the temperature data changes smoothly, you can choose RBF kernel, which is more effective for smooth time series data without obvious jumps. If the temperature changes are more complex and contain irregular fluctuations, you can consider using Matern kernel, which can better adapt to irregular data.

[0065] Training model: Use historical temperature data to train the Gaussian process regression model. The training process includes: Input: Historical temperature data and relevant features such as timestamps.

[0066] Output: predicted value of temperature (i.e. normal temperature value under the current state).

[0067] By maximizing the log-likelihood function and optimizing the hyperparameters of the kernel function, the model can accurately fit the historical data.

[0068] A Gaussian process is a random process indexed by input points, and the joint distribution of any finite number of points is a multivariate normal distribution. Specifically, given an input , assuming the output The joint distribution of can be expressed as: ;in, represents a multivariate normal distribution, is the covariance matrix, which describes the similarity between data points. Suppose there is a set of training data, where is the input data, Represents the input data value corresponding to the nth data point, is the output data, i.e. the historical operation data of the equipment, such as temperature, pressure, etc.) Represents the output data value corresponding to the nth data point.

[0069] For Gaussian process regression, the goal is to give some input points (usually the current input data to be predicted, such as the current operating status of the device), predicting its corresponding output , and calculate the corresponding confidence interval. The specific regression formula is as follows: ; Is the training data The covariance matrix with itself That is, the covariance of the historical equipment operation data, which represents the similarity between every two data points in the training data. This matrix describes the correlation of the historical equipment operation data. Is the current data to be predicted With historical data The covariance matrix between : . is the covariance matrix The inverse matrix of is used to adjust the results predicted based on the training data. T represents the transpose of the matrix. is the variance of the current point to be predicted, indicating the autocorrelation of the point. Represents the prediction of the current device state based on the training data.

[0070] The confidence interval of the predicted value can be calculated by using the variance of the Gaussian process regression. Given the result of the regression, the variance of the output value is It can be calculated by the following formula: ; This variance represents the uncertainty of the model for the current prediction, that is, the confidence interval of the predicted value is: ; is the predicted value, is the standard deviation, and the confidence interval is generally 95%, that is, a multiple of 1.96 is used. Represents the covariance between the current input point and itself, which is used to calculate the variance of the prediction (i.e. the confidence interval of the prediction).

[0071] Model prediction: After training is completed, the model is used to predict the current temperature. Given a new time point or device status, the GPR model will output the predicted temperature value at that time point. At the same time, the GPR model provides a confidence interval for each predicted temperature value. This confidence interval represents the credibility range of the predicted value, usually expressed by the predicted mean and variance. For example, if the predicted temperature is 80°C and the confidence interval is [78°C, 82°C], it means that the prediction of 80°C has high credibility and the probability that the actual temperature is within this range is high.

[0072] Dynamically adjust thresholds: Determination of dynamic thresholds: Dynamic warning thresholds can be set through the confidence interval provided by the GPR model. For example, the normal range of the current device temperature is predicted by historical data. If the device temperature exceeds the confidence interval predicted by the model (such as the actual temperature exceeds 82°C or is lower than 78°C), it means that the device may be faulty or abnormal, and the system will trigger an alarm. As time goes by and the equipment operation data is continuously updated, the model will dynamically adjust the prediction results and thresholds based on the new data, thereby achieving continuous and accurate fault diagnosis and early warning.

[0073] Example description: Assume that the historical temperature data of a dyeing device shows that under normal working conditions, the temperature changes smoothly and has periodic fluctuations. After training the Gaussian process regression model, the model calculates the following: The predicted value of the current device temperature is 80°C. The predicted confidence interval is [78°C, 82°C]. If the device temperature exceeds 82°C or is lower than 78°C in subsequent operation, it means that the current temperature is out of the normal range, and the system will trigger an alarm through fault diagnosis to remind the operator to pay attention to possible fault risks.

[0074] When the diagnostic alarm module performs fault diagnosis alarm operation based on the dynamic threshold alarm mechanism, if the actual operation data of the current dyeing equipment exceeds the confidence interval predicted by the dynamic threshold model, an alarm will be issued to remind the current dyeing equipment of a fault.

[0075] When constructing the Gaussian process regression algorithm, the RBF kernel or the Matern kernel function is used. In Gaussian process regression (GPR), the kernel function (also called the covariance function) determines how the model measures the similarity between different input points. The RBF kernel and the Matern kernel are commonly used kernel functions, and they have different characteristics in dealing with the smoothness and volatility of data. The RBF kernel is smooth, and it assumes that the similarity between data points decays rapidly with the increase of distance. Specifically, for distant points, their similarity decreases rapidly. It means that the correlation between data points decays rapidly as their distance increases, so it is suitable for data that changes smoothly and does not fluctuate significantly. The RBF kernel is mathematically infinitely smooth, which means that it can generate very smooth prediction curves without sudden fluctuations between data points. It is suitable for processing smooth, smooth data without obvious irregular changes, especially when the data changes slowly and there are no obvious mutations.

[0076] Compared with the RBF kernel, the Matern kernel has less smoothness and allows the model to have some fluctuations when predicting. Its smoothness is controlled by a hyperparameter, with common values ​​of 1.5 and 2.5, which determine the smoothness of the kernel function. By adjusting the size of the hyperparameter, the Matern kernel can control the volatility of the model. The Matern kernel is suitable for situations where the data changes irregularly or has jumps, especially in scenarios where the data changes require more volatility, it can better capture such irregular changes. It is suitable for data with irregular fluctuations or local jumps, especially when the data has large changes in certain areas, the Matern kernel can capture these local changes by adjusting the hyperparameters.

[0077] Before constructing the Gaussian process regression algorithm, the monomer smoothness index corresponding to each fault parameter used to measure the rate of change is calculated, and the monomer regularity index used to evaluate the regularity of each fault parameter is calculated at the same time. Then, the comprehensive smoothness index and comprehensive regularity index are calculated respectively by weighting the fault parameter variance. Fuzzy reasoning is used, and the comprehensive smoothness index and comprehensive regularity index are used as input data together. The kernel function type used when constructing the Gaussian process regression algorithm is used as output data, and the kernel function type used when constructing the Gaussian process regression algorithm is inferred.

[0078] A specific example is as follows: The calculation formula for the monomer smoothness index is: ; N represents the total number of samples in the current time window, Indicates fault parameters x such as temperature, pressure, vibration, etc., in the sequence number The corresponding Subsampled data, Indicates fault parameters In the sequence number The corresponding Subsampled data, Indicates fault parameters The corresponding monomer smoothness index; The calculation formula of monomer regularity index is: ; and Respectively represent the fault parameters The corresponding mean and standard deviation are, Indicates fault parameters The corresponding monomer regularity index; The calculation formula of the comprehensive smoothness index is: ; The calculation formula of comprehensive regularity index is: ; represents the category of fault parameters, K represents the total number of categories of fault parameters, Indicates that the fault parameter category is The variance of and Respectively represent the categories The corresponding monomer smoothness index and monomer regularity index are represents the comprehensive smoothness index, Represents the comprehensive regularity index.

[0079] Fuzzy reasoning includes the following steps: Fuzzification: Fuzzification is the process of converting input crisp data into fuzzy values. In this case, we have two input data: comprehensive smoothness index and comprehensive regularity index. These two values ​​are usually numerical, but in the fuzzy inference system, we need to convert them into fuzzy linguistic variables, such as fuzzy levels such as "high", "medium", and "low".

[0080] Comprehensive smoothness index: According to the value, it can be divided into three categories: "high", "medium" and "low". Assuming that the higher the index value, the stronger the signal smoothness, the fuzzy membership can be divided according to the value of the index. For example, when the index value is greater than a certain threshold, the membership is "high", and when it is less than a certain threshold, the membership is "low".

[0081] Comprehensive regularity index: Similarly, the regularity index is divided into three fuzzy levels: "high", "medium" and "low". The higher the regularity, the more stable the data changes, and the lower the regularity, the more unstable the changes.

[0082] Fuzzy rules: Fuzzy rules are used to describe the relationship between input and output. In this case, the rules will combine the comprehensive smoothness index and the comprehensive regularity index to infer the kernel function type of the Gaussian process regression model used for the output (for example, selecting the RBF kernel or the Matern kernel). For example, the rules can be: Rule 1: If the smoothness index is "high" and the regularity index is "high", use the RBF kernel. Rule 2: If the smoothness index is "low" and the regularity index is "low", use the Matern kernel. Rule 3: If the smoothness index is "medium" and the regularity index is "high", other types of kernel functions may be selected. Through these rules, the system is able to infer the kernel function type based on the fuzziness level of the input.

[0083] Reasoning: In the reasoning stage, the fuzzy reasoning system will reason on the input data according to the defined rules and generate output data. By combining fuzzy rules, we input the fuzzified results of the comprehensive smoothness index and the comprehensive regularity index into the reasoning system and output a fuzzy value, indicating whether to choose the RBF kernel or the Matern kernel. In actual operation, this step is processed by methods such as "weighted average method" or "maximum membership method". By calculating the membership of each rule, and calculating the output result based on the membership.

[0084] Defuzzification: Defuzzification is the process of converting fuzzy output values ​​into specific decision results. Ultimately, the inference system outputs a specific Gaussian process regression kernel function type (such as RBF kernel or Matern kernel). The result of defuzzification is the selection of which kernel function to use to build the Gaussian process regression model. For example, if the result of fuzzy reasoning shows that the RBF kernel has the highest degree of membership, then the RBF kernel is ultimately selected as the kernel function for Gaussian process regression; if the Matern kernel has a higher degree of membership, then the Matern kernel is selected.

[0085] In Gaussian process regression (GPR), choosing the right kernel function is key. Two commonly used kernel functions are: RBF kernel (radial basis function kernel): RBF kernel is suitable for smooth data without obvious fluctuations, and can generate smooth prediction results. RBF kernel performs well when data changes slowly and there is no significant fluctuation. Matern kernel: Matern kernel is suitable for situations where data changes are complex or volatile. Matern kernel can control smoothness by adjusting hyperparameters and is suitable for data with irregular changes or large fluctuations. Through fuzzy reasoning, according to the input comprehensive smoothness index and comprehensive regularity index, the most suitable data model (RBF kernel or Matern kernel) can be inferred, thereby improving the accuracy and prediction ability of the regression model.

[0086] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0087] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0088] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0089] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0090] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An intelligent fault diagnosis and alarm system for a high temperature and high pressure dyeing machine, characterized in that: It includes data acquisition module, data preprocessing module, cluster analysis module, deviation analysis module, adjustment module, and diagnosis alarm module; The data acquisition module is used to collect the operation data of the high temperature and high pressure dyeing machine in real time to obtain the original data; The data preprocessing module is used to denoise, standardize and fill missing values ​​of the original data, generate structured time series data, and obtain preprocessed data; The cluster analysis module is used to perform cluster analysis on the pre-processed data of the current dyeing equipment and identify similar equipment groups to other dyeing equipment in normal state; The deviation analysis module extracts the statistical characteristics of the current dyeing equipment based on the cluster analysis results, and compares them with the historical data of similar equipment groups to perform feature deviation analysis to determine whether the current dyeing equipment has potential faults; When there is a potential fault in the current dyeing equipment, the adjustment module switches the preset static threshold alarm mechanism to a dynamic threshold alarm mechanism, and constructs a dynamic threshold model for dynamically adjusting the alarm threshold; The diagnosis alarm module performs fault diagnosis alarm operations based on the static threshold alarm mechanism or dynamic threshold alarm mechanism currently adopted.

2. According to claim 1, the intelligent fault diagnosis and alarm system for a high temperature and high pressure dyeing machine is characterized in that: The data preprocessing module includes a data denoising unit for denoising the original data, a data standardization unit for standardizing the original data, and a data missing value filling unit for filling missing values ​​in the original data.

3. The intelligent fault diagnosis and alarm system for a high temperature and high pressure dyeing machine according to claim 2 is characterized in that: The clustering analysis module uses the K-means clustering algorithm to convert the operating data of the current dyeing equipment under the preset static threshold alarm mechanism and normal state into a first vector, and converts the operating data of other dyeing equipment under the preset static threshold alarm mechanism and normal state into a second vector, and performs K-means clustering on all the second vectors with the first vector to identify similar equipment groups between the current dyeing equipment and other dyeing equipment.

4. The intelligent fault diagnosis and alarm system for a high temperature and high pressure dyeing machine according to claim 3 is characterized in that: During feature deviation analysis, the statistical characteristics of the current dyeing equipment are obtained to obtain statistical feature data group one, and the statistical characteristics corresponding to the historical data of a similar equipment group are obtained to obtain statistical feature data group two. Each statistical feature data is preset with a deviation threshold. A range conversion is performed based on the statistical feature data group two and the deviation threshold corresponding to each statistical feature data to obtain a statistical feature range data group, and the statistical feature data group one is compared with the statistical feature range data group.

5. The intelligent fault diagnosis and alarm system for a high temperature and high pressure dyeing machine according to claim 4 is characterized in that: Judging whether there is a potential fault in the current dyeing equipment means: If, in the comparison result between statistical feature data group one and statistical feature range data group, the statistical feature data in statistical feature data group one exceeds the total number of statistical features of the corresponding statistical feature range in the statistical feature range data group, if the total number of statistical features is not zero, it is determined that the current dyeing equipment has a potential fault; if the total number of statistical features is zero, it is determined that the current dyeing equipment does not have a potential fault.

6. The intelligent fault diagnosis and alarm system for a high temperature and high pressure dyeing machine according to claim 5, characterized in that: The dynamic threshold model is constructed by using the Gaussian process regression algorithm. The constructed dynamic threshold model predicts the normal operating parameters of the current dyeing equipment and provides a confidence interval for each predicted value of the normal operating parameter, indicating the credibility range of the predicted value.

7. The intelligent fault diagnosis and alarm system for a high temperature and high pressure dyeing machine according to claim 6, characterized in that: When the diagnostic alarm module performs fault diagnosis alarm operation based on the dynamic threshold alarm mechanism, if the actual operation data of the current dyeing equipment exceeds the confidence interval predicted by the dynamic threshold model, an alarm will be issued to remind the current dyeing equipment of a fault.

8. The intelligent fault diagnosis and alarm system for a high temperature and high pressure dyeing machine according to claim 7, characterized in that: When constructing the Gaussian process regression algorithm, the RBF kernel or Matern kernel function is used.

9. The intelligent fault diagnosis and alarm system for a high temperature and high pressure dyeing machine according to claim 8, characterized in that: Before constructing the Gaussian process regression algorithm, the monomer smoothness index corresponding to each fault parameter used to measure the rate of change is calculated, and the monomer regularity index used to evaluate the regularity of each fault parameter is calculated at the same time. Then, the comprehensive smoothness index and comprehensive regularity index are calculated respectively by weighting the fault parameter variance. Fuzzy reasoning is used, and the comprehensive smoothness index and comprehensive regularity index are used as input data together. The kernel function type used when constructing the Gaussian process regression algorithm is used as output data, and the kernel function type used when constructing the Gaussian process regression algorithm is inferred.

Images