Intelligent abnormality identification method for food heavy metal detection data

Abstract

The application relates to the technical field of data detection, in particular to an intelligent abnormality identification method for food heavy metal detection data; overall fluctuation change coefficients and local fluctuation change coefficients are obtained according to fluctuation characteristics of absorbance in detection curves; trend coefficients are obtained according to change trend characteristics of the absorbance in the detection curves; detection reference curves are obtained according to curve difference characteristics of all detection curves and preset standard curves; difference coefficients are obtained according to difference characteristics of the absorbance of the detection reference curves and the detection curves and difference characteristics of the trend coefficients; fluctuation consistency degrees are obtained according to difference characteristics of the overall fluctuation change coefficients and the local fluctuation change coefficients of the detection curves; and abnormality indexes are obtained according to the difference coefficients and the fluctuation consistency degrees. Normal detection curves are obtained according to distribution characteristics of the abnormality indexes of all detection curves; and the content of noble metals in grains is obtained according to the normal detection curves, so that the accuracy of noble metal detection in food is improved.

Description

Technical Field

[0001] This invention relates to the field of data detection technology, and specifically to an intelligent anomaly identification method for food heavy metal detection data. Background Technology

[0002] Because some foods may contain high levels of heavy metals, and excessive intake of heavy metals can damage multiple organs in the human body—for example, cadmium poisoning can severely damage the kidneys and bones, leading to kidney failure, osteoporosis, and other diseases—it is necessary to test the heavy metal content in food to ensure food safety and public health. The core quantitative method for detecting cadmium content in food using existing GFAAS graphite furnace atomic absorption spectrometry is the standard curve method, which achieves quantification by plotting a linear relationship between signal intensity and cadmium concentration. However, in practical applications, this method may experience signal drift due to matrix effects, interference factors, instrument fluctuations, and operational errors. At high concentrations, the instrument may exhibit nonlinear responses due to signal saturation or enhanced matrix effects, causing the standard curve to bend and resulting in deviations in the calculated cadmium content, ultimately affecting the accuracy of cadmium content detection in food. Summary of the Invention

[0003] To address the aforementioned technical problems, the present invention aims to provide an intelligent anomaly identification method for food heavy metal detection data. The specific technical solution adopted is as follows:

[0004] The detection curves of heavy metals in multiple grain samples were obtained using the GFAAS method.

[0005] The overall fluctuation coefficient and local fluctuation coefficient of the detection curve are obtained based on the fluctuation characteristics of absorbance at different concentrations in the detection curve; the trend coefficient of the detection curve is obtained based on the trend characteristics of absorbance at different concentrations in the detection curve.

[0006] A detection reference curve is obtained based on the curve difference characteristics between all detection curves and the preset standard curve; the difference coefficient of the detection curve is obtained based on the difference characteristics of absorbance and trend coefficient between the detection reference curve and the detection curve; the fluctuation consistency is obtained based on the difference characteristics of the overall fluctuation change coefficient and the local fluctuation change coefficient of the detection curve.

[0007] The abnormality index of the detection curve is obtained based on the difference coefficient and the fluctuation consistency; the normal detection curve is obtained based on the distribution characteristics of the abnormality indices of all detection curves; and the precious metal content in the grain is obtained based on the normal detection curve.

[0008] Furthermore, the step of obtaining the overall fluctuation coefficient and local fluctuation coefficient of the detection curve based on the fluctuation characteristics of absorbance at different concentrations in the detection curve includes:

[0009] In the formula, A represents the overall fluctuation coefficient of the detection curve, and N represents the number of concentration values ​​in the detection curve. Indicates the first The slope of the line connecting the current concentration value and the previous concentration value. The slope of the line connecting the nth concentration value and the previous concentration value is represented; the detection curve is divided into different concentration range intervals on average, and the overall fluctuation coefficient of each concentration range interval is calculated as the different local fluctuation coefficients of the detection curve.

[0010] Furthermore, the step of obtaining the trend coefficient of the detection curve based on the trend characteristics of absorbance changes at different concentrations in the detection curve includes:

[0011] The average slope corresponding to different concentration values ​​in the detection curve is calculated to obtain the trend coefficient of the detection curve.

[0012] Furthermore, the step of obtaining the detection reference curve based on the curve difference characteristics between all detection curves and the preset standard curve includes:

[0013] Calculate the dynamic time-normalized distance between the detection curve and the preset standard curve and round it to obtain the difference distance; construct a histogram of the difference distance, calculate the average value of the detection curve corresponding to the maximum value in the histogram, and obtain the detection reference curve.

[0014] Further, the step of obtaining the difference coefficient of the detection curve based on the difference characteristics of absorbance and the difference characteristics of trend coefficient between the detection reference curve and the detection curve includes:

[0015] In the formula, R represents the difference coefficient of the detection curve, and N represents the number of concentration values ​​in the detection curve. This represents the absorbance corresponding to the nth concentration value in the detection curve. The absorbance corresponding to the nth concentration value in the detection reference curve is represented by L, the trend coefficient of the detection curve is represented by T, and the trend coefficient of the detection reference curve is represented by T.

[0016] Furthermore, the step of obtaining the fluctuation consistency degree based on the difference characteristics between the overall fluctuation change coefficient and the local fluctuation change coefficient of the detection curve includes:

[0017] In the formula, E represents the consistency of fluctuation in the detection curve, and M represents the number of local fluctuation variation coefficients in the detection curve. Let A represent the local fluctuation coefficient of the m-th local fluctuation, and let A represent the overall fluctuation coefficient of the detection curve.

[0018] Further, the step of obtaining the anomaly index of the detection curve based on the difference coefficient and the fluctuation consistency includes:

[0019] In the formula, W represents the anomaly index of the detection curve, R represents the difference coefficient of the detection curve, and E represents the fluctuation consistency of the detection curve.

[0020] Furthermore, the step of obtaining the normal detection curve based on the distribution characteristics of the abnormality index of all detection curves includes:

[0021] Density clustering is performed on all detection curves based on their abnormality indices to obtain different clusters; the maximum value of the abnormality index in the cluster with the largest number of occurrences is taken as the abnormality threshold; and the detection curves with abnormality indices lower than the abnormality threshold are taken as the normal detection curves.

[0022] The present invention has the following beneficial effects:

[0023] In this invention, obtaining the overall fluctuation coefficient and local fluctuation coefficient of the detection curve can reflect the overall and local fluctuation characteristics of the detection curve, thereby enabling the analysis of the possibility of anomalies in the detection curve. Obtaining the trend coefficient of the detection curve can reflect the changing trend of the detection curve, thereby enabling the analysis of the possibility of anomalies in the detection curve. Since the occurrence of signal drift is accidental, obtaining a detection reference curve allows for further comparison with the detection curve, improving the accuracy of anomaly analysis. Obtaining the difference coefficient reflects the data difference characteristics between the detection curve and the detection reference curve, thereby assessing whether the detection curve is abnormal. Obtaining the fluctuation consistency degree allows for the assessment of whether the detection curve is abnormal based on its fluctuation characteristics. Finally, obtaining the anomaly index accurately analyzes the degree of anomaly in the detection curve, thereby screening out normal detection curves; and obtaining the precious metal content in grain based on normal detection curves improves the accuracy of precious metal content detection. Attached Figure Description

[0024] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, 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.

[0025] Figure 1 The flowchart illustrates an intelligent anomaly identification method for food heavy metal detection data, as provided in one embodiment of the present invention. Detailed Implementation

[0026] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an intelligent anomaly identification method for food heavy metal detection data proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0027] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0028] The following description, in conjunction with the accompanying drawings, details the specific scheme of the intelligent anomaly identification method for food heavy metal detection data provided by the present invention.

[0029] Please see Figure 1 The diagram illustrates a flowchart of an intelligent anomaly identification method for food heavy metal detection data according to an embodiment of the present invention. The method includes the following steps:

[0030] Step S1: Use the GFAAS method to obtain detection curves of heavy metals in multiple grain samples.

[0031] In this embodiment of the invention, the implementation scenario is to detect the content of precious metals in grains and improve the detection accuracy. The grain is corn, and the precious metal is cadmium. The implementer can determine the grain and precious metal target according to the implementation scenario. First, the detection curves of heavy metals in multiple grain samples are obtained using the GFAAS method. In this embodiment of the invention, the rapid digestion-standard addition method based on GFAAS is used to determine cadmium in grains. Corn or other grain varieties are used as the detection target. The pretreatment conditions and steps of rapid digestion are optimized, the optimal ratio of sample to nitric acid is determined, and the accuracy of the standard addition method under different grain matrices is explored. In this method, 0.3 grams of sample are weighed and placed in a digestion vessel, and 8.0 ml of nitric acid is accurately added. The national standard method is to add 5 ml of nitric acid and 2 ml of hydrogen peroxide. In this embodiment of the invention, the preset standard curve is the detection curve generated according to the national standard method. The horizontal axis of the detection curve represents different cadmium concentrations, and the difference between the absorbance value of the corresponding concentration and the absorbance value of the sample is used as the vertical axis to obtain the detection curve for each grain sample. It should be noted that the standard curve method is prior art, and the specific steps are not described in detail.

[0032] Step S2: Obtain the overall fluctuation coefficient and local fluctuation coefficient of the detection curve based on the fluctuation characteristics of absorbance at different concentrations in the detection curve; obtain the trend coefficient of the detection curve based on the trend characteristics of absorbance at different concentrations in the detection curve.

[0033] Instantaneous changes in parameters such as plasma stability, nebulizer efficiency, and detector gain can cause signal drift. For example, plasma may undergo acid-base reactions or hydrolysis under acidic conditions, while it may be more stable under alkaline conditions. High-viscosity or particulate samples may cause tubing blockage, affecting nebulization efficiency. Electronic circuit noise may affect the stability of the detection signal, leading to instantaneous gain changes. Furthermore, at high concentrations, the instrument may exhibit a nonlinear response due to signal saturation or enhanced matrix effects, causing the standard curve to bend. Therefore, the standard addition method is used to place the standard and sample in the same matrix environment, thereby reducing the influence of matrix effects. However, the signal drift problem still exists, so further analysis of the detection curve is needed to improve the accuracy of cadmium content detection in grains.

[0034] Due to the instantaneous characteristics and instability of signal drift, the fluctuation and trend characteristics of the detection curve generated in each experiment can be analyzed. The detection curve expresses the linear relationship between the response signal and the concentration of the target substance, therefore, the curve presents an approximately straight line. First, based on the fluctuation characteristics of absorbance at different concentrations in the detection curve, the overall fluctuation coefficient and local fluctuation coefficient of the detection curve are obtained. Preferably, in this embodiment of the invention, the steps of obtaining the overall fluctuation coefficient and local fluctuation coefficient of the detection curve include:

[0035]

[0036] In the formula, A represents the overall fluctuation coefficient of the detection curve, and N represents the number of concentration values ​​in the detection curve. Indicates the first The slope of the line connecting the current concentration value and the previous concentration value. The slope of the line connecting the nth concentration value and the previous concentration value is represented. The detection curve is divided into different concentration range intervals on average, and the overall fluctuation coefficient of each concentration range interval is calculated as the local fluctuation coefficient of the detection curve. That is, the calculation method of the local fluctuation coefficient is the same as that of the overall fluctuation coefficient, only the selection of the concentration range is different. In this embodiment of the invention, the detection curve is divided into 10 different concentration range intervals on average. The implementer can determine the interval according to the implementation scenario. This reflects the difference in slope between two adjacent concentration values. The smaller the slope difference, the more similar the changes are and the weaker the fluctuation characteristics. The larger the overall fluctuation coefficient, the more obvious the fluctuation characteristics of the detection curve.

[0037] Furthermore, the trend coefficient of the detection curve can be obtained based on the trend characteristics of absorbance changes at different concentrations in the detection curve; preferably, in this embodiment of the invention, the step of obtaining the trend coefficient includes: calculating the average slope corresponding to different concentration values ​​in the detection curve to obtain the trend coefficient of the detection curve. The larger the trend coefficient, the more obvious the trend of change of the detection curve.

[0038] Step S3: Obtain the detection reference curve based on the curve difference characteristics between all detection curves and the preset standard curve; obtain the difference coefficient of the detection curve based on the difference characteristics of absorbance and trend coefficient between the detection reference curve and the detection curve; obtain the fluctuation consistency based on the difference characteristics of the overall fluctuation change coefficient and the local fluctuation change coefficient of the detection curve.

[0039] Since signal drift is random, reference values ​​can be obtained through statistical analysis. To improve the accuracy of the reference results, a preset standard curve is introduced. Under the same sample type, the similarity between the preset standard curve and each detection curve is calculated. Therefore, a detection reference curve is obtained based on the curve difference characteristics between all detection curves and the preset standard curve. Preferably, in this embodiment of the invention, the step of obtaining the detection reference curve includes: calculating the dynamic time warping distance between the detection curve and the preset standard curve and rounding it to obtain the difference distance. It should be noted that the dynamic time warping distance is obtained through existing dynamic time warping algorithms, and the specific steps will not be elaborated here. The larger the dynamic time warping distance between two curves, the less similar the two curves are. A histogram of difference distances is constructed, with the horizontal axis representing the numerical value of the difference distance and the vertical axis representing the number of each difference distance. Since the probability of most detection curves being abnormal in normal experiments is relatively small, the difference distances of most detection curves are relatively similar. Therefore, the difference distance with the largest number in the histogram corresponds to the more normal detection curve. Then, the average value of the detection curves corresponding to the maximum value in the histogram is calculated to obtain the detection reference curve. This detection reference curve is constructed by the average value of the vertical axis corresponding to each concentration of the selected detection curve. Therefore, the difference coefficient of the detection curve can be obtained based on the difference characteristics of absorbance and trend coefficient between the detection reference curve and the detection curve; preferably, in this embodiment of the invention, the step of obtaining the difference coefficient includes:

[0040]

[0041] In the formula, R represents the difference coefficient of the detection curve, and N represents the number of concentration values ​​in the detection curve. This represents the absorbance corresponding to the nth concentration value in the detection curve. The absorbance corresponding to the nth concentration value in the detection reference curve is represented by L, the trend coefficient of the detection curve is represented by T, and the trend coefficient of the detection reference curve is represented by T. The larger the value, the greater the difference between the detection curve and the detection reference curve; The larger the coefficient of variation, the greater the difference between the detection curve and the reference curve. Therefore, the larger the coefficient of variation of the detection curve, the greater the difference between the detection curve and the reference curve, and the more likely the detection curve is to be abnormal.

[0042] Furthermore, if the detection curve is not subject to any instantaneous interference at any given time, and there are no recording errors, then the local fluctuation characteristics of the detection curve are consistent with the overall fluctuation characteristics. Therefore, the fluctuation consistency degree is obtained based on the difference between the overall fluctuation coefficient and the local fluctuation coefficient of the detection curve. Preferably, in this embodiment of the invention, the step of obtaining the fluctuation consistency degree includes:

[0043]

[0044] In the formula, E represents the uniformity of fluctuation in the detection curve, and M represents the number of local fluctuation variation coefficients in the detection curve. Let represent the local fluctuation coefficient of the m-th local variation, and A represent the overall fluctuation coefficient of the detection curve. The smaller the difference between local and overall fluctuations, the more normal and reliable the local data is, and the lower the probability of data anomalies. Simultaneously, the smaller the overall fluctuation coefficient of the detection curve, the smaller the fluctuation characteristics of the detection curve, and the more normal the detection curve is. Therefore, the greater the fluctuation consistency of the detection curve, the weaker and more consistent its fluctuation characteristics are, and the more its variation characteristics conform to the normal characteristics of a detection curve. Conversely, the smaller the fluctuation consistency, the more obvious the fluctuation characteristics of the detection curve and the more inconsistent the fluctuation characteristics across different concentration ranges, and the more likely the detection curve is to be abnormal.

[0045] Step S4: Obtain the abnormality index of the detection curve based on the difference coefficient and the consistency of fluctuation; obtain the normal detection curve based on the distribution characteristics of the abnormality index of all detection curves; obtain the precious metal content in the grain based on the normal detection curve.

[0046] After obtaining the difference coefficient and fluctuation consistency of each detection curve, a larger difference coefficient indicates a higher probability of anomaly; conversely, a smaller fluctuation consistency indicates a higher probability of anomaly. Therefore, an anomaly index is obtained based on the difference coefficient and fluctuation consistency. Preferably, in this embodiment of the invention, the step of obtaining the anomaly index includes:

[0047]

[0048] In the formula, W represents the anomaly index of the detection curve, R represents the difference coefficient of the detection curve, and E represents the consistency of fluctuation of the detection curve. A larger anomaly index indicates a more likely anomaly in the detection curve, leading to less accurate calculation of the cadmium concentration in the corresponding sample. Furthermore, a normal detection curve can be obtained based on the distribution characteristics of the anomaly indices of all detection curves.

[0049] Preferably, in this embodiment of the invention, the step of obtaining normal detection curves includes: performing density clustering based on the anomaly index of all detection curves to obtain different clusters; using the maximum value of the anomaly index in the cluster with the largest number of curves as the anomaly threshold; since the proportion of detection curves without anomalies is the largest during normal experiments, the detection curve corresponding to the cluster with the largest number of curves is less likely to be abnormal; at the same time, the larger the anomaly index, the more likely the detection curve is to be abnormal, so the maximum value of the anomaly index in the largest cluster is used as the anomaly threshold. Then, detection curves with an anomaly index lower than the anomaly threshold are considered normal detection curves. This anomaly threshold can accurately eliminate abnormal detection curves, thereby obtaining accurate normal detection curves for cadmium content detection. Finally, the precious metal content in grain is obtained based on the normal detection curves. It should be noted that the step of obtaining the cadmium content of a sample through detection curves is prior art, and the specific steps will not be elaborated further. The average cadmium content obtained from all normal detection curves can accurately reflect the cadmium concentration in corn, thereby improving the accuracy of precious metal detection in grain.

[0050] In summary, this invention provides an intelligent anomaly identification method for food heavy metal detection data. It obtains the overall fluctuation coefficient and local fluctuation coefficient based on the absorbance fluctuation characteristics in the detection curve; obtains the trend coefficient based on the absorbance change trend characteristics in the detection curve; obtains a detection reference curve based on the curve difference characteristics between all detection curves and a preset standard curve; obtains a difference coefficient based on the difference characteristics of absorbance and trend coefficient between the detection reference curve and the detection curve; obtains the fluctuation consistency degree based on the difference characteristics of the overall fluctuation coefficient and local fluctuation coefficient of the detection curve; and obtains an anomaly index based on the difference coefficient and fluctuation consistency degree. This invention obtains a normal detection curve based on the distribution characteristics of the anomaly index of all detection curves; and obtains the precious metal content in grains based on the normal detection curve, thus improving the accuracy of precious metal detection in food.

[0051] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0052] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for intelligent anomaly identification in food heavy metal detection data, characterized in that, The method includes the following steps: The detection curves of heavy metals in multiple grain samples were obtained using the GFAAS method. The overall fluctuation coefficient and local fluctuation coefficient of the detection curve are obtained based on the fluctuation characteristics of absorbance at different concentrations in the detection curve; the trend coefficient of the detection curve is obtained based on the trend characteristics of absorbance at different concentrations in the detection curve. A detection reference curve is obtained based on the curve difference characteristics between all detection curves and the preset standard curve; the difference coefficient of the detection curve is obtained based on the difference characteristics of absorbance and trend coefficient between the detection reference curve and the detection curve; the fluctuation consistency is obtained based on the difference characteristics of the overall fluctuation change coefficient and the local fluctuation change coefficient of the detection curve. The abnormality index of the detection curve is obtained based on the difference coefficient and the fluctuation consistency; the normal detection curve is obtained based on the distribution characteristics of the abnormality indices of all detection curves; and the heavy metal content in the grain is obtained based on the normal detection curve. The step of obtaining the detection reference curve based on the curve difference characteristics between all detection curves and the preset standard curve includes: Calculate the dynamic time-normalized distance between the detection curve and the preset standard curve and round it to obtain the difference distance; construct a histogram of the difference distance, calculate the average value of the detection curve corresponding to the maximum value in the histogram, and obtain the detection reference curve; The step of obtaining the anomaly index of the detection curve based on the difference coefficient and the fluctuation consistency includes: In the formula, W represents the anomaly index of the detection curve, R represents the difference coefficient of the detection curve, and E represents the fluctuation consistency of the detection curve. The step of obtaining the normal detection curve based on the distribution characteristics of the abnormal index of all detection curves includes: Density clustering is performed on all detection curves based on their abnormality indices to obtain different clusters; the maximum value of the abnormality index in the cluster with the largest number of occurrences is taken as the abnormality threshold; and the detection curves with abnormality indices lower than the abnormality threshold are taken as the normal detection curves.

2. The intelligent anomaly identification method for food heavy metal detection data according to claim 1, characterized in that, The step of obtaining the overall fluctuation coefficient and local fluctuation coefficient of the detection curve based on the fluctuation characteristics of absorbance at different concentrations in the detection curve includes: In the formula, A represents the overall fluctuation coefficient of the detection curve, and N represents the number of concentration values ​​in the detection curve. Indicates the first The slope of the line connecting the current concentration value and the previous concentration value. The slope of the line connecting the nth concentration value and the previous concentration value is represented; the detection curve is divided into different concentration range intervals on average, and the overall fluctuation coefficient of each concentration range interval is calculated as the different local fluctuation coefficients of the detection curve.

3. The intelligent anomaly identification method for food heavy metal detection data according to claim 2, characterized in that, The step of obtaining the trend coefficient of the detection curve based on the trend characteristics of absorbance changes at different concentrations in the detection curve includes: The average slope corresponding to different concentration values ​​in the detection curve is calculated to obtain the trend coefficient of the detection curve.

4. The intelligent anomaly identification method for food heavy metal detection data according to claim 1, characterized in that, The step of obtaining the difference coefficient of the detection curve based on the difference characteristics of absorbance and the difference characteristics of trend coefficient between the detection reference curve and the detection curve includes: In the formula, R represents the difference coefficient of the detection curve, and N represents the number of concentration values ​​in the detection curve. This represents the absorbance corresponding to the nth concentration value in the detection curve. The absorbance corresponding to the nth concentration value in the detection reference curve is represented by L, the trend coefficient of the detection curve is represented by T, and the trend coefficient of the detection reference curve is represented by T.

5. The intelligent anomaly identification method for food heavy metal detection data according to claim 1, characterized in that, The step of obtaining the fluctuation consistency based on the difference between the overall fluctuation coefficient and the local fluctuation coefficient of the detection curve includes: In the formula, E represents the consistency of fluctuation in the detection curve, and M represents the number of local fluctuation variation coefficients in the detection curve. Let A represent the local fluctuation coefficient of the m-th local fluctuation, and let A represent the overall fluctuation coefficient of the detection curve.

Images