CN109711603B - A method for rapidly predicting the number of rice infected by Aspergillus fungi based on electronic nose - Google Patents

Abstract

The invention discloses a method for rapidly predicting the number of rice infected by Aspergillus fungi based on an electronic nose. After the rice is sterilized by ultraviolet rays, it is inoculated with a certain amount of Aspergillus fungi. The electronic nose was used to detect the headspace gas of the inoculated rice samples at different storage times; at the same time, the traditional plate counting method was used to detect the number of colonies on the rice samples; the electronic nose sensor array was optimized according to the principal component analysis, and the stable value method was used Feature extraction is performed on the optimized sensor response signal. Finally, the partial least squares regression algorithm is used to establish a prediction model based on the characteristic value of the electronic nose signal and the number of colonies, and the regression model with a large correlation coefficient and a small root mean square error is selected as the final prediction model of the number of colonies, so as to obtain the predicted number of colonies . The invention has no damage to rice samples, is simple to operate, has good prediction effect, and has high practical application value.

Description

A method for rapidly predicting the number of rice infected by Aspergillus fungi based on electronic nose

technical field

The invention belongs to the field of microbial detection and relates to a method for quickly predicting the number of rice infected by Aspergillus fungi based on an electronic nose.

Background technique

Rice is one of the most important food varieties in the world. About 50% of the world's population uses rice as a staple food, and more than 2 billion people in Asia use rice and its products as the main source of calorie intake. Over the years, my country's rice output has ranked first in the world, accounting for about 30% of the world's total rice output, accounting for about 1/3 of the domestic total grain output. And with the improvement of people's living standards and the increase of population, its consumption is also in a gradual upward trend. However, grains are rich in nutrients, and are very susceptible to fungal infection and deterioration under suitable moisture and temperature conditions. It is reported that the loss of agricultural products and industrial raw materials caused by grain mildew or contamination by mycotoxins in the world reaches tens of billions of dollars every year. What's more serious is that if humans mistakenly eat food seriously contaminated by fungi, they will be poisoned or induce some diseases, even cancer. Therefore, the detection of fungal contamination and the amount of fungal contamination in food is of great significance to ensure food safety and reduce the outbreak of foodborne diseases. As a non-destructive and rapid detection method, the electronic nose has broad application prospects in the detection of rice contaminated by fungi.

Contents of the invention

Aiming at the problems of complicated time-consuming, low efficiency, and high cost of the current fungal detection methods, the present invention provides a method for rapidly predicting the degree of rice infection by Aspergillus fungi based on an electronic nose, which can accurately and quickly identify contaminated rice degree without damaging the rice sample.

A method for quickly predicting the degree of rice infection by Aspergillus fungi based on electronic nose, its specific steps are as follows:

(1) Sterilize the rice samples, inoculate and store them with Aspergillus fungi, and take out 7 groups of samples from the rice samples stored for 0 days to 6 days according to the daily interval (repeat N times for each group of samples, N>10), and store them at room temperature. Bottom seal, the sealed volume is not less than 500mL (according to the ratio of rice to container volume 1g:25mL). The sample is left to stand for 30-60 minutes to saturate the headspace gas in the sealed container, thereby obtaining the headspace gas. The headspace gas in the sealed container is sucked into the sensor array channel of the electronic nose through the built-in pump of the electronic nose, and the response signal of the sensor is detected and recorded, so as to obtain the response curve of the sensor to the rice samples with different storage times;

(2) Rinse, dilute, and count colonies after 5-7 days of plate culture for the detection samples of different storage times after the electronic nose detection in step (1), so as to obtain the number of colonies in samples of different storage times;

(3) Extract the signal when the sensor detects stability in step (1) as the eigenvalue, use the extracted eigenvalue as the independent variable, and use the number of colonies of the samples detected in the step (2) at different storage times as the dependent variable, and pass the partial minimum The quadratic regression established a quantitative prediction model for the response signal of the electronic nose sensor and the number of colonies of samples stored for different periods of time after fungal inoculation, and selected the regression model with a large correlation coefficient and a small root mean square error as the final prediction model for the number of colonies;

The partial least squares regression model is: Y=a ₁ ×X ₁ +a ₂ ×X ₂ +...+a ₁₀ ×X ₁₀ +b, where Y is the number of colonies, a ₁ , a ₂ ... a ₁₀ and b are both constants;

(4) According to step (1) to detect the rice sample with unknown number of Aspergillus fungi, obtain the response curve of the electronic nose, use the method of step (3) to extract the eigenvalue, and substitute the eigenvalue into the quantitative prediction model of step 3, predict The number of colonies infected by fungi in rice samples is unknown, so as to achieve the purpose of effectively predicting the number of rice infected by Aspergillus fungi only by using the electronic nose.

Further, the calculation formulas of correlation coefficient and root mean square error in the described step (3) are as follows:

为预测模型建立过程中所有样本菌落数真实值的平均值；Y_i为预测模型建立过程中第i个样本菌落数的预测值；/>

为预测模型过程中所有样本菌落数预测值的平均值。r is the correlation coefficient; RMSE is the root mean square error; N is the number of samples whose degree of rice infection by Aspergillus fungus is known in the process of building the prediction model; X _i is the number of colonies in the ith sample in the process of building the prediction model the actual value of

is the average value of the true value of the colony number of all samples in the process of establishing the prediction model; Y _i is the predicted value of the colony number of the i-th sample in the process of establishing the prediction model; />

It is the average value of the predicted values of the colony number of all samples in the process of predicting the model.

The beneficial effects of the present invention are as follows: use the electronic nose to predict the infection degree of rice by Aspergillus fungus, take the stable value of the sensor response curve as the characteristic value, and use the Logistic equation according to the X-axis barycentric coordinates of the two-dimensional score map in the principal component analysis. Fungal growth curve simulation, and partial least squares regression algorithm to establish a prediction model based on the electronic nose signal characteristic value and the number of colonies, so as to obtain the predicted number of colonies. This method realizes the quantitative prediction of the number of rice infected by Aspergillus fungus directly by using the electronic nose, and has the characteristics of fast and non-destructive, which provides a new method for the prediction of the degree of rice and even agricultural products.

Description of drawings

Fig. 1 is the sensor response signal of the electronic nose detecting the degree of rice being infected by Aspergillus japonica;

Fig. 2 is the regression model curve between Aspergillus japonica infection degree and actual value;

Fig. 3 is the sensor response signal that electronic nose detects the degree of infection of rice by Aspergillus fumigatus;

Figure 4 is the regression model curve between the infection degree of Aspergillus fumigatus and the actual value.

Detailed ways

The present invention will be further described below in conjunction with accompanying drawing and example.

The present invention adopts a method for quickly predicting the number of rice infected by Aspergillus fungi based on electronic nose, and establishes an effective prediction model based on electronic nose data. The specific steps are as follows:

(1) After sterilizing the commercially available rice under a 110mW s/cm ² ultraviolet lamp for 30-60min, inoculate it with a certain concentration of Aspergillus fungi, and store it at 28±1°C and 85% relative humidity. Take out 7 groups of samples from rice samples stored for 0 days to 6 days at daily intervals (each group of samples is repeated N times, N>10), and seal them at room temperature with a sealed volume of not less than 500mL (according to the ratio of rice weight to container volume) 1g:25mL). The sample was left to stand for 30-60 minutes to saturate the headspace gas in the sealed container to obtain headspace gas; before each electronic nose test, clean the electronic nose system with dry and clean air, and set the cleaning flow rate to 400mL/ min-600mL/min, the cleaning time is 60-80 seconds; after the cleaning is completed, the headspace gas in the sealed container is sucked into the sensor array channel of the electronic nose through the built-in pump of the electronic nose, and the sensor of the electronic nose reacts with the sample gas to generate a sensor signal; the sensor signal is the ratio of the conductivity G of the sensor when it is in contact with the sample gas to the conductivity G0 of the sensor when it passes through the calibration gas, that is, G/G0; the gas flow rate of the built-in pump is 200-300mL/min, and the detection time is 70 -90 seconds; detect and record the response signal of the sensor array, thereby obtaining the response curve of the sensor array to the rice samples of different storage times;

(2) Rinse, dilute, and count colonies after 5-7 days of plate culture for the detection samples of different storage times after the electronic nose detection in step (1), so as to obtain the number of colonies in samples of different storage times;

(3) Extract the signal when the sensor detects stability in step (1) as the eigenvalue, use the extracted eigenvalue as the independent variable, and use the number of colonies of the samples detected in the step (2) at different storage times as the dependent variable, and pass the partial minimum The quadratic regression established a quantitative prediction model for the response signal of the electronic nose sensor and the number of colonies of samples stored for different periods of time after fungal inoculation, and selected the regression model with a large correlation coefficient and a small root mean square error as the final prediction model for the number of colonies;

The partial least squares regression model is: Y=a ₁ ×X ₁ +a ₂ ×X ₂ +...+a ₁₀ ×X ₁₀ +b, where Y is the number of colonies, a ₁ , a ₂ ... a ₁₀ and b are both constants;

(4) According to step (1) to detect the rice sample with unknown number of Aspergillus fungi, obtain the response curve of the electronic nose, use the method of step (3) to extract the eigenvalue, and substitute the eigenvalue into the quantitative prediction model of step 3, predict The number of colonies infected by fungi in rice samples is unknown, so as to achieve the purpose of effectively predicting the number of rice infected by Aspergillus fungi only by using the electronic nose.

Further, the calculation formulas of correlation coefficient and root mean square error in the described step (3) are as follows:

为预测模型建立过程中所有样本菌落数真实值的平均值；Y_i为预测模型建立过程中第i个样本菌落数的预测值；/>

为预测模型过程中所有样本菌落数预测值的平均值。r is the correlation coefficient; RMSE is the root mean square error; N is the number of samples whose degree of rice infection by Aspergillus fungus is known in the process of building the prediction model; X _i is the number of colonies in the ith sample in the process of building the prediction model the actual value of

is the average value of the true value of the colony number of all samples in the process of establishing the prediction model; Y _i is the predicted value of the colony number of the i-th sample in the process of establishing the prediction model; />

It is the average value of the predicted values of the colony number of all samples in the process of predicting the model.

The invention is suitable for rapid prediction of the number of agricultural products such as rice, wheat and corn infected by different Aspergillus fungi, and is mainly suitable for electronic nose detection and data processing of the results. The following examples facilitate a better understanding of the present invention, but do not limit the present invention.

Example 1

A method for quickly predicting the degree of rice infection by Aspergillus jaundice based on electronic nose, its steps are as follows:

(1) Taking the commercially available Jiangsu Xingjia rice as the experimental object, after sterilizing it under a 110mW s/cm ² ultraviolet lamp for 30-60min, select 7 batches of rice samples and inoculate 0.2mL bright white rice with a concentration of 10 ⁷ CFU/mL respectively. Aspergillus spore suspension and stored at 28±1°C and 85% relative humidity. A batch of rice samples was taken out every 24 hours and placed in a container at room temperature and sealed. A total of 7 groups of samples were taken out (each group of samples was repeated 21 times), and each group was numbered 0d, 1d, 2d, 3d, 4d, 5d, 6d. The volume of the container is 500mL. After the sample is left to stand for 60 minutes, the headspace gas in the sealed container is saturated to obtain the headspace gas; before the start of each electronic nose test, the electronic nose system is cleaned with dry and clean air. The flow rate is 600ml/min, and the cleaning time is 60 seconds; after the cleaning is completed, the headspace gas in the sealed container is sucked into the sensor array channel of the electronic nose through the built-in pump of the electronic nose, and the sensor of the electronic nose reacts with the sample gas to generate a sensor signal; The sensor signal is the ratio of the conductivity G of the sensor when it contacts the sample gas to the conductivity G0 of the sensor when it passes through the calibration gas, that is, G/G0; the gas flow rate of the built-in pump is 200ml/min, and the detection time is 90 seconds; the detection record The sensor array responds to the signal, thereby obtaining the response curve of the sensor array to the rice samples of different storage times;

In this case, the PEN2 electronic nose of the German AIRSENSE company is used as the detection instrument. The electronic nose system consists of 10 metal oxide sensors. The models and corresponding characteristics are shown in Table 1:

Table 1 PEN2 electronic nose sensor array and the response characteristics of each sensor

After obtaining the output result of the electronic nose, perform feature extraction on it, observe the response curve, and find that it tends to be stable after 75 seconds, so the value of the response curve at 75 seconds is used as the characteristic value. Electronic nose response curves for different storage times.

(2) The determination of the total number of colonies of Aspergillus japonica in rice inoculated with different storage times was carried out in accordance with the national food safety standard GB4789.15-2010, and the determination of the number of colonies was repeated three times;

(3) The signal when the sensor detects stability in the extraction step (1) is used as the eigenvalue, and the extracted eigenvalue is used as the independent variable, and the number of colonies of the samples detected in the step (2) with different storage times is used as the dependent variable to carry out partial least binary Multiplicative regression modeling. Its correlation coefficient R ² is 0.894, and its expression is: Y=-32.1256-22.7202×X ₁ +1.9015×X ₂ +68.9305×X ₃ +22.9134×X ₄ -25.5406×X ₅ +3.1017×X ₆ -0.0337× X ₇ -0.1137×X ₈ -24.339×X ₉ +11.0814×X ₁₀ , where Y is the number of colonies, and X ₁ -X ₁₀ are the stable values of the sensors of the electronic nose.

(4) In order to verify the accuracy of the above model, the sensor response value of the prediction set was substituted into the above prediction model to calculate the predicted number of rice infected by Aspergillus lumina, and establish a regression model with the actual number of infections. The results are shown in Figure 2. The model formula is: y=0.846*x+0.415, where y is the predicted value, x is the actual value, the correlation coefficient R ² =0.886, RMSE=0.195 shows that the prediction effect of the model is better.

Example 2

A method for quickly predicting the degree of rice infection by Aspergillus fumigatus based on electronic nose, its steps are as follows:

(1) Taking the commercially available Jiangsu Xingjia rice as the experimental object, after sterilizing it under a 110mW s/cm ² ultraviolet lamp for 30-60min, select 7 batches of rice samples and inoculate 0.2mL of Aspergillus fumigatus at a concentration of 10 ⁷ CFU/mL The spore suspension was stored at 28±1°C and 85% relative humidity. A batch of rice samples was taken out every 24 hours and placed in a container at room temperature and sealed. A total of 7 groups of samples were taken out (each group of samples was repeated 21 times), and each group was numbered 0d, 1d, 2d, 3d, 4d, 5d, 6d. The volume of the container is 500mL. After the sample is left to stand for 60 minutes, the headspace gas in the sealed container is saturated to obtain the headspace gas; before the start of each electronic nose test, the electronic nose system is cleaned with dry and clean air. The flow rate is 600ml/min, and the cleaning time is 60 seconds; after the cleaning is completed, the headspace gas in the sealed container is sucked into the sensor array channel of the electronic nose through the built-in pump of the electronic nose, and the sensor of the electronic nose reacts with the sample gas to generate a sensor signal; The sensor signal is the ratio of the conductivity G of the sensor when it contacts the sample gas to the conductivity G0 of the sensor when it passes through the calibration gas, that is, G/G0; the gas flow rate of the built-in pump is 200ml/min, and the detection time is 90 seconds; the detection record The sensor array responds to the signal, thereby obtaining the response curve of the sensor array to the rice samples of different storage times;

In this case, the PEN2 electronic nose of the German AIRSENSE company is used as the detection instrument. The electronic nose system consists of 10 metal oxide sensors. The models and corresponding characteristics are shown in Table 2:

Table 2 PEN2 electronic nose sensor array and the response characteristics of each sensor

After obtaining the output result of the electronic nose, perform feature extraction on it, observe the response curve, and find that they all tend to be stable after 75 seconds, so the value of the response curve at 75 seconds is used as the characteristic value. Figure 3 shows the difference between rice inoculated with Aspergillus fumigatus Electronic nose response curves for storage time.

(2) The determination of the total number of colonies in rice inoculated samples with different storage times was carried out in accordance with the national food safety standard GB4789.15-2010, and the determination of the number of colonies was repeated three times.

(3) The signal when the sensor detects stability in the extraction step (1) is used as an eigenvalue, and the extracted eigenvalue is used as an independent variable, and the number of sample colonies of different storage times detected in the step (2) is used as a dependent variable for partial minimum Quadratic regression modeling. Its correlation coefficient R ² is 0.938, and its expression is: Y=-98.1099+2.0643×X ₁ -4.376×X ₂ +72.6379×X ₃ -11.3324×X ₄ -45.5001×X ₅ +94.2517×X ₆ +20.0076× X ₇ -97.1027×X ₈ +77.745×X ₉ -7.2422×X ₁₀ , where Y is the number of colonies, and X ₁ -X ₁₀ are the stable values of the sensors of the electronic nose.

(4) In order to verify the accuracy of the above model, the sensor response value of the prediction set was substituted into the above prediction model to calculate the predicted degree of rice infection by Aspergillus fumigatus, and establish a regression model with the actual degree of infection. The results are shown in Figure 4. The model formula is: y=0.86*x+0.284, where y is the predicted value, x is the actual value, the correlation coefficient R ² =0.911, RMSE=0.172, which shows that the prediction effect of the model is better.

Through the detailed introduction of the method for quickly predicting the number of rice infected by Aspergillus fungi based on the electronic nose in the above examples, the established model for predicting the number of rice infected by Aspergillus fungi has a higher predictive performance, further illustrating that the invention discloses The method has high application value and is worthy of widespread promotion.

Claims (1)

1. A method for rapidly predicting the infection quantity of rice by aspergillus fungi based on an electronic nose is characterized by comprising the following steps:

step 1, sterilizing rice samples, inoculating and storing Aspergillus fungi, taking out 7 groups of samples from the rice samples stored for 0-6 days at intervals of each day, repeating the steps of N times, N > 10, sealing at room temperature, keeping the sealed volume at least 500mL, keeping the sample stand for 30-60 minutes according to the volume ratio of rice to the container of 1 g:25: 25mL, enabling the headspace gas in the sealed container to be saturated, thus obtaining headspace gas, sucking the headspace gas in the sealed container into a sensor array channel of an electronic nose through an electronic nose built-in pump, detecting and recording sensor response signals, and thus obtaining response curves of the sensor to the rice samples with different storage times;

step 2, carrying out rinsing, dilution and plate culture on 7 samples in the step 1 for 5-7 days, and counting aspergillus fungus colonies on a plate so as to obtain sample colony numbers with different storage times;

step 3, extracting signals detected by the sensor in the step 1 to be stable as characteristic values, taking the extracted characteristic values as independent variables, taking the colony numbers of the samples detected in the step 2 at different storage times as dependent variables, establishing a quantitative prediction model of response signals of the electronic nose sensor and the colony numbers of the samples stored at different times after fungus inoculation by partial least squares regression, and selecting a regression model with large correlation coefficient and small root mean square error as a final colony number prediction model, wherein the partial least squares regression model is as follows: y=a ₁ ×X ₁ +a ₂ ×X ₂ +……+a ₁₀ ×X ₁₀ +b, wherein Y is the colony count, a ₁ 、a ₂ ...a ₁₀ And b are both constants;

step 4, detecting rice samples with unknown aspergillus fungi quantity according to the step 1, obtaining a response curve of the electronic nose, extracting a characteristic value by using the method of the step 3, substituting the characteristic value into the quantitative prediction model of the step 3, and predicting the colony number of the unknown rice samples infected by fungi, thereby achieving the purpose of effectively predicting the aspergillus fungi infection quantity of the rice by the electronic nose;

in the step 1, the aspergillus fungi are dominant aspergillus fungi in rice mildew, including aspergillus candidus, aspergillus fumigatus, aspergillus clavatus and aspergillus niger;

in step 1, the rice sample is placed at 110mW s/cm ² Sterilizing under ultraviolet lamp for 30-60min, storing at 28+ -1deg.C and 85% relative humidity, cleaning the electronic nose system with dry clean air before each detection, setting cleaning flow rate at 400ml/min-600ml/min, and cleaning for 60-90 seconds; after the cleaning is finished, sucking the headspace gas in the sealed container into a sensor array channel of the electronic nose through an electronic nose built-in pump, and reacting an electronic nose sensor with sample gas to generate a sensor signal; the sensor signal is the ratio of the conductivity G of the sensor when in contact with the sample gas to the conductivity G0 of the sensor when passing through the calibration gas, i.e. G/G0; the flow rate of the gas of the built-in pump is 200-300ml/min, and the detection time is 70-90 seconds; detecting and recording response signals of the sensor array, so as to obtain response curves of the sensor array to rice samples with different storage times;

in step 3, the calculation formula of the correlation coefficient and the root mean square error is as follows:

r is a correlation coefficient; RMSE is root mean square error; n is the number of samples with known infection degree of aspergillus fungi on rice used in the process of establishing a predictive model; x is X _i The actual value of the ith sample colony number in the process of establishing the prediction model is obtained;

establishing an average value of true values of all sample colony numbers in the process of establishing a prediction model; y is Y _i A predicted value of the ith sample colony number in the process of establishing a prediction model; />

Predicting values for colony numbers of all samples in the process of predicting modelAverage value of (2).

Images