Seasoning physical and chemical index detection method based on calibration model
Technical Field
The invention relates to a chemometric detection method, in particular to a seasoning physicochemical index detection method based on a calibration model.
Background
In the quality monitoring of the seasoning, indexes such as total acid, amino acid nitrogen, sodium chloride, total solid matters and the like need to be detected, and in the current detection process, the detection process is complex, the detection is interfered by the state of a sample, the detection time is long, and the cost of test reagents and the labor cost are high.
For quantitative analysis establishment method and rapid detection method analysis, the related technologies at present mainly have the following aspects:
patent numbers: CN201811412543.9 "a method for establishing quantitative analysis model of soy sauce" discloses a method for establishing quantitative analysis model, comprising: collecting a spectrogram of a sample to be detected; determining the type of a sample to be detected by using the classification model; extracting all correction samples with the same type as the sample to be detected in the original correction set to form a temporary criterion set; calculating the distance between each correction sample and the sample to be detected in the temporary criterion set; judging whether the number of the correction samples with the distance less than the threshold distance is less than E, if so, stopping the calculation, and if not, sorting the correction samples with the distance less than the threshold distance according to the descending distance, selecting the first E spectrums closest to the sample to be detected, and forming a temporary correction set; and constructing a quantitative analysis model by utilizing the temporary correction set, wherein the quantitative analysis model is used for predicting the quantitative detection value of the sample to be detected. The method can judge the type of the sample to be detected through the spectral characteristics of the sample to be detected, and construct a high-specificity quantitative analysis model in real time, so that the prediction precision of quantitative analysis is improved, the model maintenance frequency and difficulty are reduced, and further, when the sample composition of a corrected sample set is enough to cover the daily fluctuation of materials, the method can avoid the model maintenance work and ensure the accuracy and continuity of the detection work. The method judges the type of the sample to be detected according to the spectral characteristics of the sample to be detected, and constructs a high-specificity quantitative analysis model in real time. However, the method is only suitable for soy sauce products, and can not detect homogeneous semi-solid products such as seasoning products.
Patent No. CN200910136253.0 & lt & gt quick determination method for preservative content in seasoning & gt discloses a quick determination method for preservative content in seasoning, comprising the following steps: a) weighing a seasoning sample; b) pretreating a sample to obtain a sample solution; c) sequentially adding an internal standard substance solution and an extracting agent into the sample solution; d) shaking up, and taking supernatant; e) carrying out chromatographic analysis on the supernatant to obtain chromatographic analysis data; f) and comparing the chromatographic analysis data with preset preservative standard data to obtain the preservative content in the seasoning. The method for rapidly measuring the content of the preservative in the seasoning can accurately and rapidly measure the content of the preservative in the seasoning, and can measure the content of various preservatives in the seasoning at one time, so that the efficiency is high, and the result is accurate. The method is mainly applied to detection of the condiment preservative, and other physical and chemical indexes of the condiment, such as indexes of total acid, amino acid nitrogen and salt, cannot be detected.
Disclosure of Invention
The invention aims to overcome the defects in the prior art and provides a seasoning physicochemical index detection method based on a calibration model.
In order to achieve the purpose, the invention adopts the technical scheme that: a seasoning physicochemical index detection method based on a calibration model comprises the following steps:
(1) establishing a calibration model:
taking A sample quantities as model basic data, selecting B samples from the A samples, and obtaining physicochemical index chemical detection values and near infrared spectrum curves of the B samples; analyzing the physicochemical index chemical detection values and the near infrared spectrum curves of the B samples by adopting a partial least square method, determining an optimal spectrum band, and establishing a calibration model between the near infrared spectrum characteristic information and the physicochemical index chemical detection values; b is less than or equal to A, and A is an integer more than or equal to 200
(2) And (3) verifying the calibration model:
cross-validation SECV and 1-VR; the SECV is a standard deviation between a near infrared predicted value and a chemical analysis value obtained when cross validation is carried out in a calibration modeling process; the 1-VR is a correlation coefficient between a near infrared predicted value and a chemical analysis value;
(3) and inputting the near infrared spectrum curve of the sample to be detected into the calibration model to obtain a physical and chemical index detection result of the sample to be detected.
On the near infrared spectrum, according to the difference of chemical components contained in different kinds of substances, the frequency multiplication and frequency combination vibration frequencies of hydrogen-containing radicals are different, so that the peak position, the peak number and the peak intensity of the near infrared spectrum are different, and different spectral sections are cut from a scanned spectral curve by adopting a partial least square method to determine the optimal spectral section.
And verifying the accuracy of the calibration model by adding unqualified samples with different index types. The accuracy of the calibration model can be roughly estimated by SECV, the lower the SECV value, the higher the accuracy of the calibration model. The 1-VR refers to a correlation coefficient between a near infrared predicted value and a chemical analysis value obtained when cross validation is carried out in a calibration modeling process, and the value is required to be greater than 0.6 and is better when being closer to 1.
The detection method adopts a least square method (PLS) to establish a near infrared model of the physical and chemical indexes of the seasoning, and establishes a correlation model between the characteristic spectrum of the substance and the component to be detected through a correlation model metrology method, so that the rapid quantitative detection of the unknown sample components by using the near infrared spectrum information of the substance is realized. In the process of establishing the model, the accuracy of the calibration model is evaluated through SECV (standard error of cross validation) by adopting the standard deviation between the near infrared predicted value and the chemical analysis value obtained in the cross validation, and the lower the SECV value is, the higher the accuracy of the calibration model is. The detection method can quickly detect the physical and chemical indexes of the seasoning, such as total acid, amino acid nitrogen, salt indexes, total solid matters and the like, improves the detection efficiency by more than 60 percent compared with the national standard method of similar products, and meets the requirement of actual production monitoring.
Preferably, in the step (2), the model SECV value is less than or equal to 0.3, and the model 1-VR value is more than or equal to 0.8, and the calibration model is confirmed.
Preferably, in the step (1), the physicochemical index chemical detection values comprise total acid, amino acid nitrogen, volatile basic nitrogen, salt index and total solid.
Preferably, the test method of the total acid is GB/T12456-; the testing method of the amino acid nitrogen is GB5009.235-2016 (determination of amino acid nitrogen in food safety national standard food); the testing method of the volatile basic nitrogen is GB 5009.228-2016 (national food safety standard) measurement of volatile basic nitrogen in food; the testing method of the salt index is GB 5009.42-2016 (national standard for food safety) determination of salt index; the method for testing the total solid is GB 5009.3-2016 (national food safety standard) determination of moisture in food.
Preferably, step (2a) is further included between step (2) and step (3): and (3) adopting an independent sample set which does not participate in the model calibration process to externally verify the calibration model, if the external verification is passed, indicating that the calibration model is usable, continuing to perform the step (3), otherwise, indicating that the calibration model is unusable, and repeating the step (1) and the step (2) to adjust the parameters and then verifying until the external verification is passed.
Preferably, the external verification pass is: the tolerance between the index value obtained by adopting calibration model inspection and the index value detected by a chemical method is within 1 time of standard deviation; and/or the index value obtained by adopting calibration model test and the index value detected by chemical method do not have significant difference by adopting pairing T test. In the T test, the confidence coefficient is more than or equal to 95 percent.
Preferably, in the step (1), the index range of the chemical detection value of the physical and chemical index is 10% of the execution standard upper and lower limit tolerance of the physical and chemical index. If the chemical detection value of the physicochemical index is in the index range, the sample is available, and if the chemical detection value is not in the index range, the sample is rejected.
Preferably, the sample to be detected is semisolid sauce; more preferably, the sample to be detected is hoisin sauce.
The invention has the beneficial effects that: the invention provides a seasoning physicochemical index detection method based on a calibration model. The detection method adopts a least square method (PLS) to establish a near infrared model of the physical and chemical indexes of the seasoning, and establishes a correlation model between the characteristic spectrum of the substance and the component to be detected through a correlation model metrology method, so that the rapid quantitative detection of the unknown sample components by using the near infrared spectrum information of the substance is realized. According to the calibration model of the infrared characteristics, the physical and chemical indexes of the seasoning, including total acid, amino acid nitrogen and the like, can be quickly detected, compared with the national standard method of similar products, the detection efficiency is improved by more than 60%, and the requirement of actual production monitoring is met.
Detailed Description
To better illustrate the objects, aspects and advantages of the present invention, the present invention will be further described with reference to specific examples.
Example 1
The embodiment is a method for detecting total acid in hoisin sauce, which comprises the following steps:
1. establishing a calibration model:
(1) in the embodiment, 548 data sample quantities are extracted as model basic data;
(2) 210 samples are selected from the sample amount 548 for chemical value detection and analysis, the detection method of the total acid is GB/T12456-;
(3) through the analysis of chemical detection values of 210 samples, the partial least square method is adopted for analyzing a spectral curve, the optimal spectral band is determined, a near infrared model is established, and the modeled spectral interval is as follows: 5743cm-1-6961cm-1;
2. And (3) verifying the calibration model:
(1) cross validation the accuracy of the calibration model is evaluated by SECV; the model SECV value is required to be less than or equal to 0.3. Through calculating the total acid model SECV value, the total acid SECV value is 0.011 and meets the requirement by referring to the GB/T6379.6-2009/ISO 5725-6:1994 measurement method and the actual application of the accuracy value of the part 6 of the accuracy (accuracy and precision) of the result;
(2) the cross validation model 1-VR, 1-VR refers to a correlation coefficient between a near infrared predicted value and a chemical analysis value obtained when cross validation is carried out in a calibration modeling process, and the model 1-VR value is required to be more than or equal to 0.8 by calculating according to 'GB/T6379.6-2009/ISO 5725-6:1994 measuring method and actual application requirements of part 6 accurate value of accuracy (accuracy and precision) of results'. The total acid 1-VR value is 0.82, which meets the requirement;
3. tolerance control
(1) The tolerance between the index value obtained by adopting calibration model inspection and the index value detected by a chemical method is within 1 time of standard deviation, the 1 time of standard deviation of the total acid model is 0.011g/100g, which is less than the method requirement that the absolute difference value of two parallel measurements is less than or equal to 0.03g/100 g;
(2) the accuracy (T) of the verification method is verified by adopting pairing T test (the confidence coefficient is more than or equal to 95 percent)Critical value, 95%=1.96,TComputing=0.68),TComputingAre all less than corresponding TCritical value, 95%The method shows that the accuracy of the corresponding index prediction of the near infrared method is similar to that of the principle detection method, and the two methods have no systematic error;
4. and inputting the near infrared spectrum curve of the sample to be detected into the calibration model to obtain a physical and chemical index detection result of the sample to be detected.
Example 2
The embodiment is a method for detecting amino acid nitrogen in hoisin sauce, which comprises the following steps:
1. establishing a calibration model
(1) In the embodiment, 480 data samples are extracted as model basic data.
(2) Selecting 180 samples from 480 samples, and carrying out chemical value detection and analysis, wherein the detection method of amino acid nitrogen is GB5009.235-2016 (determination of amino acid nitrogen in food safety national standard food), the sample index range is set in combination with 10% of the tolerance of the upper and lower limit values of the execution standard, the upper and lower limit values of the execution standard of the amino acid nitrogen (0.35g/100g, 0.40g/100g), and the sample index range is set to (0.32g/100g, 0.44g/100 g);
(3) the chemical detection values of 180 samples are analyzed, and the spectral curve is analyzed by adopting partial least square method to determineEstablishing a near infrared model in an optimal spectrum section, wherein the modeled spectrum interval is as follows: 5287cm-1-8955cm-1;
2. And (3) verifying the calibration model:
(1) cross validation the accuracy of the calibration model is evaluated by SECV; the model SECV value is required to be less than or equal to 0.3. By calculating the SECV value of the total acid model, the SECV value of amino acid nitrogen is calculated to be 0.010 and meets the requirement by referring to the practical application of the accurate value of the part 6 of GB/T6379.6-2009/ISO 5725-6:1994 measuring method and the accuracy (accuracy and precision) of the result;
(2) the cross validation model 1-VR, 1-VR refers to a correlation coefficient between a near infrared predicted value and a chemical analysis value obtained when cross validation is carried out in a calibration modeling process, and the model 1-VR value is required to be more than or equal to 0.8 by calculating according to 'GB/T6379.6-2009/ISO 5725-6:1994 measuring method and actual application requirements of part 6 accurate value of accuracy (accuracy and precision) of results'. Amino acid nitrogen 1-VR value is 0.98, which meets the requirement;
3. tolerance control
(1) The tolerance between the index value obtained by adopting calibration model inspection and the index value detected by a chemical method is within 1 time of standard deviation, and the 1 time of standard deviation of the amino acid nitrogen model is 0.010g/100g which is less than the requirement of 2 percent of tolerance of a national standard method;
(2) the accuracy (T) of the method is verified by using paired T test (generally, confidence coefficient is more than or equal to 95 percent)Critical value, 95%=1.96,TComputing=0.86),TComputingAre all less than corresponding TCritical value, 95%The method shows that the accuracy of the corresponding index prediction of the near infrared method is similar to that of the principle detection method, and the two methods have no systematic error;
4. and inputting the near infrared spectrum curve of the sample to be detected into the calibration model to obtain a physical and chemical index detection result of the sample to be detected.
Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention and not for limiting the protection scope of the present invention, and although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions can be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.