Quality detection method for soymilk processing process
By acquiring monitoring data during the soymilk production process, the overall data fluctuation, anomaly, noise interference, anomaly confidence, and correlation of changes were determined. This solved the problem that existing technologies could not identify abnormal states during soymilk processing, enabling early and accurate quality detection and improved production efficiency.
Patent Information
- Application Number
- CN202511809717.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-03
- Publication Date
- 2026-02-13
AI Technical Summary
Existing technologies cannot effectively identify abnormal states in the production process of soy milk processing. As a result, when an alarm is triggered, a quality problem has already occurred, and it is impossible to distinguish whether it is a sensor malfunction or a real process abnormality, leading to waste of raw materials and a decrease in production efficiency.
By acquiring monitoring data during the soymilk production process, we can determine the overall data fluctuation, anomaly, noise interference, anomaly confidence level, and correlation of changes. By combining the comprehensive confidence level of the data, we can detect the degree of anomaly in soymilk, eliminate the influence of noise and sensor malfunctions, and achieve accurate quality testing.
It enables early and accurate quality detection during the soymilk processing, eliminates the impact of noise and sensor malfunctions on detection, and improves production efficiency and quality control.
Smart Images
Figure CN121526433A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of soybean milk production monitoring and analysis technology, and in particular to a quality detection method for soybean milk processing. Background Technology
[0002] Soy milk is a widely consumed plant-based beverage. To ensure its quality, strict quality control and testing are required during production to guarantee that its taste, nutritional components, and safety meet standards. With the development of intelligent technology, sensors can acquire and analyze production data in real time during soy milk processing. This allows for the effective identification of abnormalities in the soy milk's condition and timely alerts, thereby reducing scrap rates, improving production efficiency, and ensuring soy milk quality.
[0003] Traditional methods for monitoring soymilk processing data only focus on whether the instantaneous value of a single sensor exceeds the limit, ignoring the dynamic changes in processing parameters and failing to effectively link the synergistic relationships between data from various dimensions. As a result, by the time an alarm is triggered, a quality problem has already occurred, and it is impossible to distinguish whether it is a sensor malfunction or a real process anomaly. Consequently, early and accurate warnings cannot be achieved, leading to the waste of a large amount of raw materials due to substandard quality and a decline in production efficiency. Summary of the Invention
[0004] In view of this, the present invention provides a quality detection method for the soy milk processing process, in order to solve the technical problems of current methods for monitoring soy milk during production that cannot detect abnormal states of soy milk early and have low monitoring accuracy.
[0005] The present invention provides a quality inspection method for a soy milk processing process, comprising: The monitoring data during the soy milk production process is acquired at set intervals, and the overall data fluctuation at the current moment is determined based on the instantaneous change of each monitoring data and the degree to which it exceeds the preset normal range. Based on the degree of change and fluctuation of each monitoring data under the preset time period before the current moment, and the difference between each monitoring data and the monitoring values of the same period in the historical processing cycle, the data anomaly degree at the current moment is determined; based on the fluctuation duration and fluctuation amount of each monitoring data at the current moment, the noise interference degree at the current moment is determined; based on the data anomaly degree and noise interference degree at the current moment, the anomaly confidence degree at the current moment is determined. The change relationship between any two monitoring data under normal conditions is recorded as the preset relationship. The number of times the change relationship between any two monitoring data at the current time is the same as the corresponding preset relationship is recorded as the number of normal relationships. The similarity of any monitoring data under a long time window and a short time window is calculated with the current time as the starting point of the backtracking. The correlation of various monitoring data at the current time is determined based on the difference between any two change similarities and the number of normal relationships. The overall confidence level at the current moment is determined by the anomaly confidence level and the correlation with the change at the current moment. Then, the degree of anomaly of the soy milk at the current moment is determined by combining the overall fluctuation of the data at the current moment. The quality inspection of soy milk processing is completed based on the degree of anomaly of soy milk at each moment.
[0006] Furthermore, determining the overall data fluctuation at the current moment includes: The absolute value of the difference between the current monitoring value and the previous monitoring value of any monitoring data is taken as the instantaneous change of the monitoring data at the current time. The absolute value of the amount by which the monitoring value of any monitoring data exceeds the preset normal range at the current time is calculated. The sum of the absolute value of the obtained excess amount and 1 is taken as the degree to which the monitoring data at the current time exceeds the preset normal range. The sum of the products of the instantaneous changes of all monitoring data at the current moment and the degree to which they exceed the preset normal range is used to obtain the overall data fluctuation at the current moment.
[0007] Furthermore, determining the data anomaly level at the current moment includes: Calculate the slope of the linear fitting function of each monitoring data under the preset time period before the current time, as well as the standard deviation and peak number of each monitoring data under the preset time period before the current time, and calculate the mean of the absolute values of the differences between the monitoring value of any monitoring data at the current time and the monitoring values at the same moment in the preset number of historical processing cycles. The slope, standard deviation, number of peaks, and mean of the absolute values of the differences of the linear fitting function corresponding to any monitoring data at the current time are calculated as the data outlier of the monitoring data at the current time. The sum of the data outliers of all monitoring data at the current time is used as the data outlier degree at the current time.
[0008] Furthermore, determining the noise interference level at the current moment includes: The peak point closest to the current time in any monitoring data is determined as the target peak point. The data segment between two adjacent valley points of the target peak point is taken as the fluctuation segment corresponding to any monitoring data at the current time. The duration of the fluctuation segment is taken as the fluctuation duration corresponding to any monitoring data at the current time. Calculate the average of the two valley points used when determining the fluctuation segment corresponding to any monitoring data at the current time, and calculate the difference between the target peak point used when determining the fluctuation segment corresponding to any monitoring data at the current time and the average of the two valley points corresponding to any monitoring data at the current time as the fluctuation difference corresponding to any monitoring data at the current time; Calculate the product of the fluctuation difference corresponding to any monitoring data at the current time and the reciprocal of the fluctuation duration corresponding to any monitoring data at the current time, and use it as the noise interference value of any monitoring data at the current time. Use the sum of the noise interference values of all monitoring data at the current time as the noise interference degree at the current time.
[0009] Furthermore, determining the anomaly confidence level at the current moment includes: The anomaly confidence level at the current moment is the ratio of the data anomaly level to the noise interference level at the current moment.
[0010] Furthermore, determining the correlation between changes in various monitoring data at the current moment includes: The absolute value of the difference in the similarity of changes between any two monitoring data at the current time is recorded as the change difference value. The sum of all change difference values and the preset minimum value is calculated and the reciprocal is recorded as the first correlation value. The sum of the number of normal relationships and the preset minimum value is recorded as the second correlation value. The product of the first correlation value and the second correlation value is used as the change correlation of various monitoring data at the current time.
[0011] Furthermore, determining the overall confidence level at the current moment includes: The overall confidence level at the current moment is calculated by multiplying the anomaly confidence level at the current moment by the correlation between the changes in various monitoring data at the current moment.
[0012] Furthermore, determining the degree of abnormality of the soy milk at the current moment includes: The normalized value of the product of the overall data fluctuation at the current moment and the overall confidence level at the current moment is used as the degree of anomaly of the soy milk at the current moment.
[0013] Furthermore, the step of performing soy milk processing quality testing based on the degree of soy milk abnormality at each time point includes: Determine whether the degree of abnormality of soy milk at any given time exceeds the preset first abnormality threshold. If so, determine that the soy milk quality is abnormal and issue an early warning. Determine whether the duration of the soy milk's abnormality exceeding the preset second abnormality threshold reaches the preset abnormality time threshold. If so, determine that the soy milk is of abnormal quality and issue an early warning. The preset first abnormality threshold is greater than the preset second abnormality threshold.
[0014] The advantages of this invention compared to the prior art are: This invention, after determining the overall fluctuation of the data at the current moment, does not compare it with a preset threshold as in traditional monitoring methods. Instead, it determines the anomaly level of the data at the current moment based on the trend and fluctuation of the data at that moment, as well as the deviation of the data at the current moment from the data of the same period in the historical soymilk processing cycle. Simultaneously, it determines the noise interference level at the current moment based on the similarity between the data near the current moment and the noise data. The anomaly confidence level at the current moment is obtained by combining the data anomaly level and the noise interference level, thus eliminating noise interference and accurately characterizing the degree of anomaly. Subsequently, the conformity between the changing relationships of various monitoring data at the current moment and known changing relationships is examined to obtain the correlation degree of the changes in various monitoring data at the current moment. After correcting the anomaly confidence level, a comprehensive confidence level is obtained. The correlation degree eliminates monitoring errors caused by sensor malfunctions. Finally, the specific and quantifiable degree of soymilk anomaly is determined by the comprehensive confidence level and the overall data fluctuation, completing the quality inspection of the soymilk processing process. This invention can eliminate the influence errors caused by noise and sensor malfunctions on the detection of soymilk processing status, and achieve quantitative characterization of the soymilk status. Accurate detection is achieved based on this quantitative characterization, providing a foundation for early detection of anomalies in the soymilk processing status. Attached Figure Description
[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0016] Figure 1 This is a flowchart illustrating a quality inspection method for soybean milk processing provided in Embodiment 1 of the present invention. Detailed Implementation
[0017] The overall concept of this invention is as follows: Unlike traditional quality inspection methods that compare real-time monitoring values with preset thresholds, this invention first determines the overall fluctuation of the monitored data at the current moment. Then, based on the degree of change and fluctuation of each data point, as well as the degree of conformity between each data point and the change pattern of noise data, the anomaly confidence level of the data is determined. Next, by judging the difference between the change relationships between different types of monitoring data and the change relationships under normal soymilk processing conditions in real time, the correlation degree of various data changes in real time is obtained. This characterizes the probability that the real-time changes in the data are due to changes in the state of the soymilk rather than sensor malfunction. The anomaly confidence level of each data point is then corrected using the obtained correlation degree to obtain a comprehensive confidence level that eliminates noise interference and sensor malfunction interference, providing a more accurate representation of the anomaly in the state of the soymilk. Finally, the degree of soymilk quality anomaly is determined by the comprehensive confidence level and the overall fluctuation of the data, thus completing the soymilk quality inspection during the soymilk processing process.
[0018] To further illustrate the technical solution of the present invention, specific embodiments are described below.
[0019] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of the invention include a particular feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. Furthermore, a particular feature, structure, or characteristic in one or more embodiments may be combined in any suitable form, and the terms "comprising," "including," "having," and variations thereof mean "including, but not limited to," unless otherwise specifically emphasized.
[0020] It should be understood that the sequence number of each step in the following embodiments does not imply 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 invention.
[0021] Method Implementation Examples: See Figure 1 This is a flowchart illustrating a quality inspection method for soy milk processing provided in Embodiment 1 of the present invention. Figure 1 As shown, the detection method may include the following steps: S101, acquire monitoring data during the soy milk production process at set intervals, and determine the overall data fluctuation at the current moment based on the instantaneous change of each monitoring data and the degree to which it exceeds the preset normal range.
[0022] In the process of soy milk processing, after grinding, soy milk needs to go through multiple processes to become the finished product. During these processes, the temperature, pH value, and viscosity of soy milk are key quality control parameters, which directly affect product safety, nutrient retention, and sensory quality. Therefore, this embodiment uses the above parameters as examples to conduct real-time quality detection of the soy milk processing process.
[0023] The aforementioned monitoring data of the soymilk liquid are collected at each stage of soymilk processing. Specifically, temperature, pH value, and viscosity data are acquired in real-time at a fixed frequency or set interval during the soymilk processing. In this embodiment, the collection frequency is set to 1 second / time, which can be adjusted by the operator according to actual needs; a higher frequency yields more accurate monitoring results. Since subsequent data analysis requires combining monitoring data from historical processing cycles, this embodiment preferably stores one month's worth of monitoring data from the soymilk processing process as historical data.
[0024] Once the data collected by each sensor is acquired in real time, the next step of analysis can be carried out.
[0025] In the process of soymilk processing, although the requirements for various monitoring data of soymilk liquid are different at different stages (grinding, separation, boiling, sterilization, homogenization, blending, etc.), that is, the fluctuation range of various data of soymilk liquid under normal quality conditions is different at different stages, the monitoring data of each individual stage generally remain stable and the relevant data will not change significantly. Therefore, when combining various data to obtain the monitoring data fluctuation at each moment of the soymilk processing process, it is also necessary to combine the optimal data range of each stage. Furthermore, this embodiment further optimizes the method by combining the local data change characteristics to determine the data fluctuation at each moment.
[0026] Furthermore, since the sensors collect different types of data and have different measurement ranges, it is necessary to determine the optimal fluctuation range corresponding to the normal range of the data monitored by each sensor at the current stage of soymilk processing, and then combine this with the instantaneous change in the data at the current moment to obtain the overall fluctuation of the data at each moment: , in, This represents the overall fluctuation of the data at time i during the soy milk processing process. This represents the instantaneous rate of change of the monitoring data collected by the k-th sensor at time i. This instantaneous rate of change is the difference between the current sensor value and the previous sensor value. A larger instantaneous rate of change indicates a greater local variation in the monitoring data collected by the sensor at that time, and a higher probability of an anomaly. This represents the monitoring value of the monitoring data collected by the k-th sensor at time i. This represents the normal range of sensor monitoring data corresponding to the k-th sensor at time i. This represents the absolute value of the amount by which the monitored value collected by the k-th sensor at time i exceeds the normal range, where The value is or Specifically, when Greater than At that time, this part was ,when Less than At that time, this part was , and when Falling within the normal range When the value is 0, the higher the degree to which the sensor data deviates from the normal range at that moment, the greater the degree of anomaly at that moment. Adding a constant of 1 to the base is to highlight the case beyond the given condition.
[0027] S102, determine the data anomaly degree at the current moment based on the degree of change and fluctuation of each monitoring data under the preset duration before the current moment, as well as the difference between each monitoring data and the monitoring value of the same period in the historical processing cycle; determine the noise interference degree at the current moment based on the fluctuation duration and fluctuation amount of each monitoring data at the current moment; determine the anomaly confidence degree at the current moment based on the data anomaly degree and noise interference degree at the current moment.
[0028] Because the production volume of soymilk in a single batch is very large, even if an anomaly occurs during processing, the corresponding changes in monitoring data will not be sudden but will occur over a certain period of time. Therefore, the local trend changes in the data from each sensor at the current moment can be used as a direction to characterize the degree of data anomaly at that moment.
[0029] Secondly, considering that ingredients are added slowly and evenly during soymilk processing, rather than all at once or intermittently, the data will not show significant fluctuations under normal processing conditions. However, when the acid pump or heating system fails, the detected data will exhibit oscillations and more noticeable fluctuations. Furthermore, the processing steps or stages of different batches of soymilk are generally consistent at the same time or period. Therefore, if there is an anomaly in the current soymilk processing status, the current monitored value will deviate from the monitored values at the same period in historical processing cycles. Thus, the local fluctuation characteristics and local deviations from historical periodicity of the current sensor data can also be used as a way to characterize the degree of data anomaly at the current moment.
[0030] Based on the above representation directions, the data anomaly degree at the current moment is constructed as follows: , in, denoted by , represents the data anomaly degree of the monitoring data at time i during the soymilk processing process; denoted by , represents the slope of the linear fitting function of the monitoring data of the k-th sensor within a preset time period before time i at time i. A larger slope indicates a stronger trend towards an abnormal development in the soymilk processing state at the corresponding time, and a greater likelihood of an abnormal trend. In this embodiment, the preset time period is preferably 1 minute. This indicates the number of peak points within a preset time period. More peak points mean more frequent fluctuations in the data within a short period, indicating higher volatility and anomalies. This represents the standard deviation of the monitoring data collected by the k-th sensor within a preset time period. A larger standard deviation indicates greater fluctuation in the data within the preset time period and a higher degree of abnormal fluctuation. This represents the monitoring value of the monitoring data collected by the k-th sensor at time i. This represents the monitoring value of the k-th sensor at time i during the l-th historical processing cycle. This represents the total number of historical processing cycles. Each historical processing cycle contains only one monitoring data value that coincides with the i-th time point. This indicates that the more times the monitoring data collected by the k-th sensor fluctuates within a preset time period at the i-th moment, and the greater the fluctuation amplitude, the higher the abnormality of the fluctuation at that moment. The greater the difference between the monitored value obtained by the k-th sensor at time i and the data at the same time in the historical processing cycle, the higher the deviation of the data at that time and the higher the abnormality of the fluctuation.
[0031] After obtaining the data anomaly level at the current moment, considering that the sensor may be affected by noise or external interference, which may cause pulses in the data, it is necessary to adjust the data anomaly level based on the magnitude of the local pulse characteristics of the current data to prevent external interference from affecting the accuracy of the anomaly detection results.
[0032] Therefore, the noise interference level at the current moment is first determined based on the duration and magnitude of fluctuations in each type of monitoring data: , in, This represents the noise interference level of the monitoring data at time i during the soymilk processing. This represents the duration of the fluctuation segment corresponding to the monitoring data collected by the k-th sensor at time i, or the duration of the fluctuation corresponding to the monitoring data collected by the k-th sensor at time i. The fluctuation segment is determined as follows: taking time i as the cutoff time, the peak point closest to time i is identified in the monitoring data collected by the i-th sensor as the target peak point. Then, the data segment between two adjacent valley points of the target peak point is taken as the fluctuation segment corresponding to the monitoring data collected by the k-th sensor at time i. This represents the target peak point in the fluctuation segment corresponding to the monitoring data collected by the k-th sensor at time i. This represents the average of the two valley points in the fluctuation segment corresponding to the monitoring data collected by the k-th sensor at time i. This means that the shorter the duration and the greater the fluctuation of the fluctuation segment corresponding to the monitoring data collected by the k-th sensor at the i-th time, the greater the possibility that the data fluctuation at that time is a pulse fluctuation caused by external interference.
[0033] Then, by combining the noise level and the data anomaly level, the anomaly confidence level at the current moment is determined: , in, This represents the anomaly confidence level at time i during the soymilk processing.
[0034] S103, record the change relationship between any two monitoring data under normal conditions as the preset relationship, calculate the number of times the change relationship between any two monitoring data at the current time is the same as the corresponding preset relationship and record it as the normal relationship number, calculate the change similarity of any monitoring data under the long time window and the short time window with the current time as the backtracking starting point, and determine the change correlation of various monitoring data at the current time based on the difference between any two change similarities and the normal relationship number.
[0035] When a sensor malfunctions, it does not necessarily mean that the state of the soymilk is abnormal during the processing. This is because a single sensor may malfunction or cause errors in the data, resulting in abnormal data collection. Therefore, after obtaining the abnormal confidence level of the sensor data through the above steps, considering that abnormal soymilk conditions usually cause changes in various monitoring data, and that this change relationship is actually the same as that in the normal soymilk processing process, this embodiment preferably also combines the change correlation between the data of each sensor to correct the obtained abnormal confidence level.
[0036] Specifically, when soymilk experiences an abnormal state during processing and the sensors are not interfered with, the overall fluctuations in the resulting data will show consistent relationships between various data points over a longer period. In other words, the correlation characteristics of short-term window data should be consistent with those of long-term window data. A rapid change in one data point due to an abnormal soymilk state will cause other monitoring data to change rapidly in the same short period, and the relative magnitude of these changes should be similar. Secondly, there are correlation rules between different data points; that is, there are certain relationships between the changes in various data points during soymilk processing. For example, protein charge decreases, molecular aggregation occurs, polysaccharide-protein binding is enhanced, pH decreases and viscosity increases, and heat treatment promotes protein dissociation and release. The Maillard reaction produces acidic substances by increasing temperature and decreasing pH, while protein denaturation and aggregation occur by increasing temperature and increasing viscosity. If the current data fluctuation does not conform to the above relationships, it indicates that the data fluctuation is likely caused by sensor malfunction.
[0037] Therefore, the correlation between various monitoring data at the current moment can be determined based on the long-term and short-term correlation characteristics and whether they conform to the change relationship, as described above. First, the relationship between any two monitoring data under normal conditions is recorded as a preset relationship. The number of times the relationship between any two monitoring data at the current time matches the corresponding preset relationship is recorded as the number of normal relationships. Using the current time as the starting point for backtracking, the similarity of changes in any monitoring data under a long time window and a short time window is calculated, thereby determining the correlation between the changes of various monitoring data at the current time. , Where represents the correlation degree of various monitoring data changes at time i; c is a preset minimum value, which is a constant whose value can be set to prevent the correlation term from being 0. This represents the similarity of changes in one of the monitoring data points under a long time window and a short time window, corresponding to the j-th type of change relationship between any two sets of monitoring data. This represents the similarity between the changes of one of the two monitoring data points in the j-th type of change relationship between any two sets of monitoring data, under a long time window and a short time window. The change similarity is actually calculated as the degree of similarity between the monitoring data sequences under the long time window and the monitoring data sequences under the short time window. This can be accomplished using the DTW algorithm or by calculating the Pearson coefficient. This represents the absolute value of the difference in the similarity between the changes of two monitoring data in the calculation of the change relationship. The larger the absolute value of the difference, the greater the possibility that the current anomaly is caused by external interference and the sensor correlation is weakened, and the lower the confidence level of the anomaly at that point. In this embodiment, the types of monitoring data selected are temperature, viscosity, and pH value. Accordingly, j represents the three sensor relationships, namely temperature-viscosity, temperature-pH value, and viscosity-pH value. This represents the number of times the change relationship between any two monitoring data at time i is the same as the corresponding preset relationship, that is, the number of normal relationships. The larger the value, the more the change between the sensors at the target time conforms to the mutual influence rules of soybean milk processing data, and the higher the confidence level of abnormal soybean milk quality.
[0038] This is still based on the types of monitoring data selected in this implementation. The specific method for determining the value is explained below: If the rate of temperature change at time i is positive, that is, if the rate of viscosity change is greater than 0 when the temperature rises, then Increment the value by 1. If the pH change rate is less than 0 at the same time, then... The value is incremented by 1, where +1 is added for any condition that is satisfied, +2 is added if all conditions are satisfied, otherwise... Adding 0 to the value means that the value remains unchanged; If the rate of temperature change at time i is negative, that is, if the rate of viscosity change is less than 0 when the temperature decreases, then Increment the value by 1. If the pH change rate is greater than 0 at the same time, then... The value is incremented by 1. If any condition is met, the increment is 1; if all conditions are met, the increment is 2; otherwise... Adding 0 to the value means that the value remains unchanged; If the rate of temperature change at time i is 0, similarly, the rate of viscosity change and the rate of pH change can both be 0. If the value is increased by 1, and both are 0, then... Add 2 to the value, otherwise... Adding 0 to the value means that the value remains unchanged.
[0039] S104. Determine the overall confidence level at the current moment based on the anomaly confidence level and the correlation degree of change at the current moment. Then, combine the overall data fluctuation at the current moment to determine the degree of anomaly of the soy milk at the current moment. Complete the soy milk processing quality detection based on the degree of anomaly of the soy milk at each moment.
[0040] Based on the obtained correlation of changes, the aforementioned anomaly confidence scores can be corrected to obtain a more accurate comprehensive confidence score that can exclude noise interference and sensor equipment failure interference, thus providing a more accurate representation of the anomalies in the state of soy milk. , Wherein, represents the overall confidence level at time i during the soymilk processing.
[0041] It's easy to understand that the greater the overall fluctuation of the data at the current moment, and the higher the overall confidence level, the higher the degree of abnormality of the soy milk at that moment, meaning the higher the probability that the soy milk quality is abnormal. , in, This indicates the degree of abnormality in the soymilk at time i during the soymilk processing. Normalization refers to methods such as linear normalization, norm normalization, etc.
[0042] Based on the degree of abnormality of soy milk at each time point, the quality detection of soy milk processing can be completed through analysis. For example, relevant early warning parameters can be set according to the quality requirements of different stages of soy milk processing. When the degree of abnormality of soy milk at a certain time exceeds the preset first abnormality threshold, an early warning will be issued. At the same time, in order to achieve early warning of abnormal situations, a preset second abnormality threshold lower than the preset first abnormality threshold can also be set. When the degree of abnormality of soy milk exceeds the preset second abnormality threshold for a continuous period of time that is greater than the preset abnormality time threshold, an early warning will also be issued.
[0043] Furthermore, to improve the accuracy of the early warning, in a preferred embodiment, an early warning is only issued and relevant staff are reminded to check the processing status when the degree of abnormality of soy milk continues to rise for a continuous period of time greater than a preset abnormal time threshold.
[0044] Furthermore, the quality of the processing of this batch of soymilk can be analyzed by combining historical anomaly data, and improvements can be made to address any deficiencies in the processing, thereby improving the quality of soymilk and reducing anomalies in soymilk production.
[0045] This enables more timely and accurate quality testing of soymilk during the processing.
[0046] This invention, after determining the overall fluctuation of the data at the current moment, does not compare it with a preset threshold as in traditional monitoring methods. Instead, it determines the anomaly of the data at the current moment based on the trend and fluctuation of the data at the current moment, as well as the deviation of the data at the current moment from the data at the same period in the historical soymilk processing cycle. At the same time, it determines the noise interference at the current moment based on the similarity between the data near the current moment and the noise data. After combining the data anomaly and the noise interference, the anomaly confidence level at the current moment is obtained, eliminating noise interference and accurately characterizing the degree of anomaly. Subsequently, the degree of conformity between the change relationship of various monitoring data at the current moment and the known change relationship is examined to obtain the change correlation of various monitoring data at the current moment. After correcting the anomaly confidence level, a comprehensive confidence level is obtained. The change correlation eliminates the monitoring error caused by sensor malfunction. Finally, the specific and quantitative degree of soymilk anomaly is determined by the comprehensive confidence level and the overall data fluctuation, thus completing the quality inspection of the soymilk processing process. This invention can eliminate the errors caused by noise and sensor malfunctions in the detection of soymilk processing status, and achieve quantitative characterization of soymilk status. At the same time, it can provide timely warnings based on the degree of quality anomaly when the soymilk anomaly is not obvious but has been present for a long time. Compared with the traditional threshold comparison method, it can detect anomalies in time and provide early warnings. At the same time, it can also provide immediate warnings when the anomaly is large, just like the traditional threshold comparison method. The combination of the two methods can achieve more efficient soymilk production quality warning, thereby improving the quality and efficiency of soymilk production.
[0047] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A quality inspection method for soy milk processing, characterized in that, The method includes: The monitoring data during the soy milk production process is acquired at set intervals, and the overall data fluctuation at the current moment is determined based on the instantaneous change of each monitoring data and the degree to which it exceeds the preset normal range. Based on the degree of change and fluctuation of each monitoring data under the preset time period before the current moment, and the difference between each monitoring data and the monitoring values of the same period in the historical processing cycle, the data anomaly degree at the current moment is determined; based on the fluctuation duration and fluctuation amount of each monitoring data at the current moment, the noise interference degree at the current moment is determined; based on the data anomaly degree and noise interference degree at the current moment, the anomaly confidence degree at the current moment is determined. The change relationship between any two monitoring data under normal conditions is recorded as the preset relationship. The number of times the change relationship between any two monitoring data at the current time is the same as the corresponding preset relationship is recorded as the number of normal relationships. The similarity of any monitoring data under a long time window and a short time window is calculated with the current time as the starting point of the backtracking. The correlation of various monitoring data at the current time is determined based on the difference between any two change similarities and the number of normal relationships. The overall confidence level at the current moment is determined by the anomaly confidence level and the correlation with the change at the current moment. Then, the degree of anomaly of the soy milk at the current moment is determined by combining the overall fluctuation of the data at the current moment. The quality inspection of soy milk processing is completed based on the degree of anomaly of soy milk at each moment.
2. The quality inspection method for the soybean milk processing process according to claim 1, characterized in that, Determining the overall data fluctuation at the current moment includes: The absolute value of the difference between the current monitoring value and the previous monitoring value of any monitoring data is taken as the instantaneous change of the monitoring data at the current time. The absolute value of the amount by which the monitoring value of any monitoring data exceeds the preset normal range at the current time is calculated. The sum of the absolute value of the obtained excess amount and 1 is taken as the degree to which the monitoring data at the current time exceeds the preset normal range. The sum of the products of the instantaneous changes of all monitoring data at the current moment and the degree to which they exceed the preset normal range is used to obtain the overall data fluctuation at the current moment.
3. The quality inspection method for the soybean milk processing process according to claim 1, characterized in that, Determining the data anomaly level at the current moment includes: Calculate the slope of the linear fitting function of each monitoring data under the preset time period before the current time, as well as the standard deviation and peak number of each monitoring data under the preset time period before the current time, and calculate the mean of the absolute values of the differences between the monitoring value of any monitoring data at the current time and the monitoring values at the same moment in the preset number of historical processing cycles. The slope, standard deviation, number of peaks, and mean of the absolute values of the differences of the linear fitting function corresponding to any monitoring data at the current time are calculated as the data outlier of the monitoring data at the current time. The sum of the data outliers of all monitoring data at the current time is used as the data outlier degree at the current time.
4. The quality inspection method for the soybean milk processing process according to claim 1, characterized in that, Determining the noise interference level at the current moment includes: The peak point closest to the current time in any monitoring data is determined as the target peak point. The data segment between two adjacent valley points of the target peak point is taken as the fluctuation segment corresponding to any monitoring data at the current time. The duration of the fluctuation segment is taken as the fluctuation duration corresponding to any monitoring data at the current time. Calculate the average of the two valley points used when determining the fluctuation segment corresponding to any monitoring data at the current time, and calculate the difference between the target peak point used when determining the fluctuation segment corresponding to any monitoring data at the current time and the average of the two valley points corresponding to any monitoring data at the current time as the fluctuation difference corresponding to any monitoring data at the current time; Calculate the product of the fluctuation difference corresponding to any monitoring data at the current time and the reciprocal of the fluctuation duration corresponding to any monitoring data at the current time, and use it as the noise interference value of any monitoring data at the current time. Use the sum of the noise interference values of all monitoring data at the current time as the noise interference degree at the current time.
5. The quality inspection method for the soybean milk processing process according to claim 1, characterized in that, Determining the anomaly confidence level at the current moment includes: The anomaly confidence level at the current moment is the ratio of the data anomaly level to the noise interference level at the current moment.
6. The quality inspection method for the soybean milk processing process according to claim 1, characterized in that, Determining the correlation between changes in various monitoring data at the current moment includes: The absolute value of the difference in the similarity of changes between any two monitoring data at the current time is recorded as the change difference value. The sum of all change difference values and the preset minimum value is calculated and the reciprocal is recorded as the first correlation value. The sum of the number of normal relationships and the preset minimum value is recorded as the second correlation value. The product of the first correlation value and the second correlation value is used as the change correlation of various monitoring data at the current time.
7. The quality inspection method for the soybean milk processing process according to claim 1, characterized in that, The determination of the overall confidence level at the current moment includes: The overall confidence level at the current moment is calculated by multiplying the anomaly confidence level at the current moment by the correlation between the changes in various monitoring data at the current moment.
8. The quality inspection method for the soybean milk processing process according to claim 1, characterized in that, Determining the degree of abnormality of the soy milk at the current moment includes: The normalized value of the product of the overall data fluctuation at the current moment and the overall confidence level at the current moment is used as the degree of anomaly of the soy milk at the current moment.
9. The quality inspection method for the soybean milk processing process according to claim 1, characterized in that, The process of detecting the quality of soy milk processing based on the degree of abnormality at each time point includes: Determine whether the degree of abnormality of soy milk at any given time exceeds the preset first abnormality threshold. If so, determine that the soy milk quality is abnormal and issue an early warning. Determine whether the duration of the soy milk's abnormality exceeding the preset second abnormality threshold reaches the preset abnormality time threshold. If so, determine that the soy milk is of abnormal quality and issue an early warning. The preset first abnormality threshold is greater than the preset second abnormality threshold.