Carbon black quality analysis system based on production data processing

Abstract

The invention relates to the technical field of carbon black quality analysis, in particular to a carbon black quality analysis system based on production data processing, which comprises a data collection module, a feature processing module and an abnormal quality analysis module, wherein the historical production data comprises normal carbon black production data and abnormal carbon black production data; the feature processing module generates new abnormal data by calculating the feature vector difference between the abnormal data points and combining with a random proportion until the number of the abnormal data points is consistent with the number of the normal data points; the abnormal quality analysis module generates an abnormal deviation index by standardizing the feature vector difference of the abnormal data points, and then determines a safety deviation threshold value according to the feature vector difference calculated by normal carbon black production data; by comparing the deviation index with the threshold value, the abnormal quality index of each abnormal data point is generated, and the process improves the recognition and classification capability of the system on abnormal conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon black mass analysis, and in particular to a carbon black mass analysis system based on production data processing. Background Art

[0002] Conventional carbon black quality analysis systems face several significant technical issues, especially when dealing with the imbalance between abnormal and normal production data.

[0003] First, since normal production data accounts for the vast majority, traditional systems are often limited by the lack of sample size when faced with rare abnormal data in actual working conditions, making it difficult to fully capture the true distribution and variation characteristics of abnormal data. This makes it possible for traditional models to perform poorly in identifying abnormal situations, and it is easy to misjudge some potential risks as normal, and fail to identify and respond to abnormal situations in a timely manner.

[0004] In addition, traditional systems may rely mainly on fixed rules or thresholds for anomaly detection. This static method based on experience lacks the necessary flexibility and adaptability when facing dynamic production environments and changing anomaly types. Due to the inability to effectively identify the complex relationships and feature differences between abnormal data, the analysis accuracy and prediction capabilities of traditional systems are often constrained.

[0005] Moreover, due to the lack of quantitative analysis of the degree of deviation of data features, traditional methods are difficult to provide detailed abnormal quality analysis, which limits the production monitoring system's detailed management and rapid decision-making of abnormal events. In addition, traditional systems have difficulty in dealing with the noise and deviation that may be introduced by abnormal data, which further reduces the accuracy and reliability of anomaly detection. In summary, these technical problems not only restrict the detection capability and flexibility of traditional carbon black quality analysis systems, but also affect their ability to maintain high efficiency and continuity in a dynamic production environment. Summary of the invention

[0006] The object of the present invention is to provide a carbon black quality analysis system based on production data processing to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: a carbon black quality analysis system based on production data processing, comprising a data collection module, a feature processing module and an abnormal quality analysis module, wherein: The data collection module collects historical production data for producing carbon black, wherein the historical production data includes normal carbon black production data and abnormal carbon black production data; The feature processing module calculates the feature vector difference between each abnormal data point in the abnormal carbon black production data and other abnormal data points; generates new abnormal data points by random proportion based on the feature vector difference; and continuously calculates the feature vector difference and generates the random proportion for each abnormal data point in the abnormal carbon black production data until the number of generated new abnormal data points is consistent with the number of normal data points in the normal carbon black production data; The abnormal quality analysis module performs standardization processing on the characteristic vector difference between each abnormal data point in the abnormal carbon black production data and other abnormal data points, and uses the processing result as the abnormal deviation index; calculates the characteristic vector difference between each normal data point in the normal carbon black production data and other normal data points based on the normal carbon black production data, and determines the safety deviation threshold based on the characteristic vector difference; determines the abnormal quality index of each abnormal data point in the abnormal carbon black production data based on the size relationship between the abnormal deviation index and the safety deviation threshold.

[0008] As a further improvement of the technical solution, the process of generating new abnormal data points by the feature processing module specifically includes: Select an abnormal data point in the abnormal carbon black production data as a current abnormal data point, extract relevant features from the current abnormal data point, and generate a current feature vector to represent the attributes of the current abnormal data point; According to the generated current feature vector, the feature vector difference between it and other abnormal data points is calculated as the abnormal feature vector difference set; the abnormal feature vector difference set is sorted in order from small to large, and the first one after sorting is randomly selected. One eigenvector difference among the abnormal data points is used as the current eigenvector difference, where Determined by cross validation, it means that in the sorted abnormal feature vector difference set, the front The number of outlier data points; Generate a random ratio, and generate new abnormal data points based on the random ratio and the current vector difference. The generation formula is as follows: ,in Indicates the generation of new abnormal data points; represents a current abnormal data point in the abnormal carbon black production data; represents the random proportion; Represents the current feature vector difference between the current abnormal data point and other abnormal data points; For each abnormal data point in the abnormal carbon black production data, the characteristic vector difference is calculated and new abnormal data points are generated continuously until the number of new abnormal data points generated is consistent with the number of normal data points in the normal carbon black production data.

[0009] As a further improvement of the technical solution, the method for calculating the feature vector difference includes direct difference summation, Euclidean distance and cosine similarity.

[0010] As a further improvement of the technical solution, the abnormal quality analysis module includes an abnormal deviation index calculation unit, and the process of calculating the abnormal deviation index by the abnormal deviation index calculation unit specifically includes: According to the current feature vector difference in the abnormal feature vector difference set generated by the feature vector difference between the current abnormal data point and other abnormal data points, the current feature vector difference is standardized. The standardized processing formula is as follows: ,in represents the abnormal deviation index, represents the mean of the feature vector of abnormal data points in the abnormal carbon black production data, Represents the variance of the feature vector of abnormal data points in the abnormal carbon black production data.

[0011] As a further improvement of the technical solution, the abnormal quality analysis module includes an abnormal quality index analysis unit, and the process of determining the abnormal quality index by the abnormal quality index analysis unit specifically includes: According to normal carbon black production data, the characteristic vector difference between each normal data point and other normal data points in the normal carbon black production data is calculated, and the calculated result is used as a normal characteristic vector difference set; the normal characteristic vector difference set is sorted from small to large; Using the quantile method, the eigenvector difference in the sorted normal eigenvector difference set is selected as the deviation limit, and it is standardized to obtain the safe deviation threshold ; The abnormal deviation index calculated for each abnormal data point in the abnormal carbon black production data is compared with the safety deviation threshold, where: if , then the abnormal quality index corresponding to the abnormal data point is high, and the difference between the abnormal deviation index and the safety deviation threshold is used as the abnormal quality index of the abnormal data point; if , then the abnormal quality index corresponding to the abnormal data point is low, and the difference between the abnormal deviation index and the safety deviation threshold is used as the abnormal quality index of the abnormal data point.

[0012] Compared with the prior art, the present invention has the following beneficial effects: The carbon black quality analysis system based on production data processing selects an abnormal data point in the abnormal carbon black production data, and the system calculates the feature vector difference between this point and other abnormal data points, and then generates a new abnormal data point based on the difference and a random ratio; this process continues until the number of new abnormal data points generated reaches a balance with the number of normal data points; the advantage of this is that by generating additional data points that simulate real abnormal situations, the system can better capture the complexity and diversity of abnormal data distribution during training; this method not only improves the recognition and classification capabilities of the machine learning model for minority classes (abnormal data points), but also increases the generalization ability of the model through a wide range of simulated abnormal backgrounds, avoiding its sensitivity to specific abnormal reactions; in addition, the use of random ratios to generate new data points helps to reduce the introduction of bias while capturing abnormal features, providing rich and solid basic data for subsequent model optimization and application.

[0013] Furthermore, by normalizing the feature vector difference set, an abnormal deviation index is generated, which effectively quantifies the degree of abnormality of each abnormal data point; specifically, by calculating the feature vector difference between each abnormal data point and other abnormal data points and normalizing them, the abnormal deviation index is used to reflect the degree of dispersion of these data points in the feature space; at the same time, the system calculates the feature vector difference between normal data points based on normal carbon black production data, and determines a deviation threshold used as a safety benchmark through quantile selection; when the abnormal deviation index is greater than this safety deviation threshold, the system determines that the corresponding abnormal data point has a higher abnormal quality index; otherwise, it is considered to have a lower abnormal quality. This quantitative mechanism can accurately evaluate the degree of deviation of each abnormal data point from normal operating conditions, thereby providing strong support for production monitoring and early warning; ultimately, this mechanism ensures normal operation even in the case of an unbalanced data set, so that the model can not only accurately detect and classify abnormal situations, but also help identify potential risks and maintain production stability and product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the overall module of the present invention; Figure 2 It is a schematic diagram of the abnormal quality analysis module unit of the present invention.

[0015] In the figure: 100, data collection module; 200, feature processing module; 300, abnormal quality analysis module; 301, abnormal deviation index calculation unit; 302, abnormal quality index analysis unit. DETAILED DESCRIPTION

[0016] The following will be combined with the accompanying drawings in 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.

[0017] Next, see Figure 1-Figure 2 The present invention provides a technical solution: a carbon black quality analysis system based on production data processing, including a data collection module 100, a feature processing module 200 and an abnormal quality analysis module 300.

[0018] The data collection module 100 collects historical production data for producing carbon black, wherein the historical production data includes normal carbon black production data and abnormal carbon black production data, specifically including: Normal carbon black production data: refers to the production data collected when operating within the ideal or planned parameter range. This data provides baseline information under standard operating conditions. Since this data comes from daily carbon black production, the production process operates under standard conditions most of the time. Therefore, the amount of data included in the normal carbon black production data is huge, including process parameters, equipment status, environmental conditions, etc. within the normal operating range; Abnormal carbon black production data: refers to the data recorded when deviations from normal operating parameters occur during the production process. It may be caused by equipment failure, improper raw material input, changes in environmental factors, etc. Therefore, the amount of data included in the abnormal carbon black production data is relatively small.

[0019] The feature processing module 200 selects an abnormal data point in the abnormal carbon black production data as the current abnormal data point, calculates the feature vector difference between the current abnormal data point and other abnormal data points, and generates a new abnormal data point by random proportion based on the feature vector difference; for each abnormal data point in the abnormal carbon black production data, the feature vector difference is calculated and the new abnormal data point is generated continuously until the number of the generated new abnormal data points is consistent with the number of the normal data points in the normal carbon black production data, specifically including: Select an abnormal data point in the abnormal carbon black production data as a current abnormal data point, extract relevant features from the current abnormal data point, such as temperature, pressure, flow rate, time period, etc., and generate a current feature vector to represent the attributes of the current abnormal data point; According to the generated current feature vector, the feature vector difference between it and other abnormal data points is calculated as the abnormal feature vector difference set. The calculation method includes but is not limited to direct difference sum, Euclidean distance and cosine similarity, etc.; and the abnormal feature vector difference set is sorted in ascending order, and the first one after sorting is randomly selected. One eigenvector difference among the abnormal data points is used as the current eigenvector difference, where Determined by cross validation, it means that in the sorted abnormal feature vector difference set, the front The number of abnormal data points is introduced by The value can control the randomness and debiasing properties of the generated new abnormal data, so as to better simulate the real distribution and structure of abnormal data in the abnormal feature space; Generate a random ratio, and generate new abnormal data points based on the random ratio and the current vector difference. The generation formula is as follows: ,in Indicates the generation of new abnormal data points; represents a current abnormal data point in the abnormal carbon black production data; represents the random proportion; Represents the current feature vector difference between the current abnormal data point and other abnormal data points; For each abnormal data point in the abnormal carbon black production data, the characteristic vector difference is calculated and new abnormal data points are generated continuously until the number of new abnormal data points generated is consistent with the number of normal data points in the normal carbon black production data.

[0020] The feature processing module 200 simulates the real distribution of scaled abnormal data by calculating the feature vector difference between abnormal data points and generating new abnormal data using random proportions; and effectively increases the amount of abnormal data by generating new abnormal data equivalent to the amount of normal data, thereby alleviating sample imbalance and improving the model's learning ability for minority classes (abnormal data).

[0021] The abnormal deviation index calculation unit 301 in the abnormal quality analysis module 300 performs normalization processing on the feature vector difference between the current abnormal data point and other abnormal data points, and uses the processing result as the abnormal deviation index, which specifically includes: According to the current feature vector difference in the abnormal feature vector difference set generated by the feature vector difference between the current abnormal data point and other abnormal data points, the current feature vector difference is standardized. The standardized processing formula is as follows: ,in represents the abnormal deviation index, represents the mean of the feature vector of abnormal data points in the abnormal carbon black production data, Represents the variance of the feature vector of abnormal data points in the abnormal carbon black production data.

[0022] The abnormal quality index analysis unit 302 in the abnormal quality analysis module 300 calculates the characteristic vector difference between each normal data point and other normal data points in the normal carbon black production data according to the normal carbon black production data, and determines the safety deviation threshold according to the characteristic vector difference; determines the abnormal quality index of each abnormal data in the abnormal carbon black production data according to the magnitude relationship between the abnormal deviation index and the safety deviation threshold, specifically including: According to normal carbon black production data, the characteristic vector difference between each normal data point and other normal data points in the normal carbon black production data is calculated, and the calculated result is used as a normal characteristic vector difference set; the normal characteristic vector difference set is sorted from small to large; Using the quantile method, the eigenvector difference in the sorted normal eigenvector difference set is selected as the deviation limit, and it is standardized to obtain the safe deviation threshold ; The abnormal deviation index calculated for each abnormal data point in the abnormal carbon black production data is compared with the safety deviation threshold, where: if , then the abnormal quality index corresponding to the abnormal data point is high, indicating that the feature of the abnormal data point deviates greatly from the normal data point range, and the difference between the abnormal deviation index and the safety deviation threshold is used as the abnormal quality index of the abnormal data point; if , the abnormal quality index corresponding to the abnormal data point is low, indicating that the characteristics of the abnormal data point are still within the normal range or close to the normal data characteristics, and the difference between the abnormal deviation index and the safety deviation threshold is used as the abnormal quality index of the abnormal data point.

[0023] The abnormal quality analysis module 300 generates an abnormal deviation index by normalizing the characteristic vector difference of the abnormal data point to quantify the abnormal degree of each abnormal point; and based on the normal data points in the normal carbon black production data, the safety deviation threshold is calculated by quantile calculation to provide a benchmark to judge whether the new data is normal; and the abnormal deviation index is compared with the safety deviation threshold to calculate the abnormal quality index, so that the abnormal degree of each abnormal data point is quantified as an index, and the abnormal degree of each abnormal data can be clarified even in the case of unbalanced samples; this not only improves the prediction ability of the model on unbalanced data sets, but also provides a more reliable basis for subsequent production decisions (such as prediction and maintenance).

[0024] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A carbon black quality analysis system based on production data processing, characterized in that: It comprises a data collection module (100), a feature processing module (200) and an abnormal quality analysis module (300), wherein: The data collection module (100) collects historical production data for producing carbon black, wherein the historical production data includes normal carbon black production data and abnormal carbon black production data; The feature processing module (200) calculates the feature vector difference between each abnormal data point in the abnormal carbon black production data and other abnormal data points; generates new abnormal data points through random proportions based on the feature vector difference; and continuously calculates the feature vector difference and generates random proportions for each abnormal data point in the abnormal carbon black production data until the number of generated new abnormal data points is consistent with the number of normal data points in the normal carbon black production data; The abnormal quality analysis module (300) performs standardization processing on the characteristic vector difference between each abnormal data point in the abnormal carbon black production data and other abnormal data points, and uses the processing result as the abnormal deviation index; calculates the characteristic vector difference between each normal data point in the normal carbon black production data and other normal data points based on the normal carbon black production data, and determines the safety deviation threshold based on the characteristic vector difference; and determines the abnormal quality index of each abnormal data point in the abnormal carbon black production data based on the size relationship between the abnormal deviation index and the safety deviation threshold.

2. The carbon black quality analysis system based on production data processing according to claim 1 is characterized in that: The process of generating new abnormal data points by the feature processing module (200) specifically includes: Select an abnormal data point in the abnormal carbon black production data as a current abnormal data point, extract relevant features from the current abnormal data point, and generate a current feature vector to represent the attributes of the current abnormal data point; According to the generated current feature vector, the feature vector difference between it and other abnormal data points is calculated as the abnormal feature vector difference set; the abnormal feature vector difference set is sorted in order from small to large, and the first one after sorting is randomly selected. One eigenvector difference among the abnormal data points is used as the current eigenvector difference, where Determined by cross validation, it means that in the sorted abnormal feature vector difference set, the front The number of outlier data points; Generate a random ratio, and generate new abnormal data points based on the random ratio and the current vector difference. The generation formula is as follows: ,in Indicates the generation of new abnormal data points; represents a current abnormal data point in the abnormal carbon black production data; represents the random proportion; Represents the current feature vector difference between the current abnormal data point and other abnormal data points; For each abnormal data point in the abnormal carbon black production data, the characteristic vector difference is calculated and new abnormal data points are generated continuously until the number of new abnormal data points generated is consistent with the number of normal data points in the normal carbon black production data.

3. The carbon black quality analysis system based on production data processing according to claim 2 is characterized in that: The calculation method of the feature vector difference includes direct difference summation, Euclidean distance and cosine similarity.

4. The carbon black quality analysis system based on production data processing according to claim 3 is characterized in that: The abnormal quality analysis module (300) comprises an abnormal deviation index calculation unit (301), and the process of calculating the abnormal deviation index by the abnormal deviation index calculation unit (301) specifically comprises: According to the current feature vector difference in the abnormal feature vector difference set generated by the feature vector difference between the current abnormal data point and other abnormal data points, the current feature vector difference is standardized. The standardized processing formula is as follows: ,in represents the abnormal deviation index, represents the mean of the feature vector of abnormal data points in the abnormal carbon black production data, Represents the variance of the feature vector of abnormal data points in the abnormal carbon black production data.

5. The carbon black quality analysis system based on production data processing according to claim 4 is characterized in that: The abnormal quality analysis module (300) comprises an abnormal quality index analysis unit (302), and the process of determining the abnormal quality index by the abnormal quality index analysis unit (302) specifically comprises: According to normal carbon black production data, the characteristic vector difference between each normal data point and other normal data points in the normal carbon black production data is calculated, and the calculated result is used as a normal characteristic vector difference set; the normal characteristic vector difference set is sorted from small to large; Using the quantile method, the eigenvector difference in the sorted normal eigenvector difference set is selected as the deviation limit, and it is standardized to obtain the safe deviation threshold ; The abnormal deviation index calculated for each abnormal data point in the abnormal carbon black production data is compared with the safety deviation threshold, where: if , then the abnormal quality index corresponding to the abnormal data point is high, and the difference between the abnormal deviation index and the safety deviation threshold is used as the abnormal quality index of the abnormal data point; if , then the abnormal quality index corresponding to the abnormal data point is low, and the difference between the abnormal deviation index and the safety deviation threshold is used as the abnormal quality index of the abnormal data point.

Images