Electronic nose system and application thereof in detection of different mildew degrees of cereals

Abstract

The invention belongs to the field of rapid detection of grain mildew, and particularly relates to an electronic nose system and application thereof to detection of different mildew degrees of grains. The electronic nose system is composed of a data storage module, a wireless communication module, a conditioning circuit board, a single-chip microcomputer, an air inlet pipe, a sensor array, a sensor air chamber, a connecting pipe, an air outlet pipe and an air pump. Wherein a guide plate is arranged in the sensor air chamber to divide the air chamber into different parts, and sensors are arranged at the different parts; therefore, the flow path of the grain volatile gas in the gas chamber is prolonged, and the sensor is ensured to be fully contacted with the gas. Meanwhile, an electronic nose system and a machine learning algorithm are used for grain mildewing detection in the granary, the mildewing degree can be accurately determined, quantitative prediction is carried out, and the method has the advantages of being high in precision and rapid in detection and is suitable for on-site rapid qualitative detection of grains in the granary; and the method has the characteristics of low cost, simplicity, convenience and high efficiency, and has a remarkable application value in the field of grain mildewing.

Description

Technical Field

[0001] The invention belongs to the field of rapid detection of grain mildew, and specifically relates to an electronic nose system and its use in detecting grain mildew degrees of different degrees. Background Art

[0002] Grain mildew is a phenomenon of grain quality decline caused by mold contamination, which can bring many hazards. Moldy grain poses a serious threat to human health. If toxins produced by molds, such as aflatoxin and ochratoxin, enter the human body, they may cause acute or chronic poisoning, manifested as symptoms such as nausea, vomiting, diarrhea, and even life-threatening in severe cases; moldy grain reduces its nutritional value. During the growth and reproduction of mold in grain, it consumes nutrients, especially protein and vitamins, which greatly reduces the nutritional content of grain and affects people's health; moldy grain affects the processing and storage of grain. It is also easy to cause secondary pollution in grain depots. The processed products may continue to contain mold and mycotoxins, posing a threat to food safety; moldy grain will cause economic losses. A large amount of moldy grain cannot be used for human consumption or processing, resulting in waste and loss of grain resources, and also increasing the risk of grain in trade. Therefore, it is necessary to accurately detect grain mildew and take timely measures.

[0003] Determining whether wheat is moldy is of great significance for food safety of humans and animals. The degree of moldiness of wheat is determined by levels including odor, color, total colony count, toxins, moisture content, fatty acids and oxidation. These parameters include basic information about wheat quality and are the main basis for wheat mold detection, utilization and quality evaluation, which helps to determine whether wheat is moldy.

[0004] At present, there are two main aspects of wheat mold detection technology. On the one hand, there are laboratory detection technologies such as plate colony counting method, gas chromatography-mass spectrometry and enzyme activity detection. On the other hand, it is the monitoring of large granaries, such as relying on human sensory evaluation and temperature monitoring of grain to make judgments. Although traditional methods are relatively mature and reliable, with good stability and sensitivity, they consume a lot of manpower and material resources in actual application. They generally have the disadvantages of high cost, large volume, complex pretreatment, long analysis time, expensive instruments and equipment, and hysteresis, and cannot realize real-time detection applications. In addition, these methods can only analyze one volatile compound or physical and chemical property in the gas being tested in each test, and cannot describe the odor characteristics of wheat. Therefore, traditional detection methods are difficult to meet the current needs for rapid detection of wheat mold.

[0005] The electronic nose is a new type of detection equipment that imitates biological olfaction and is composed of various gas sensors. The sensor array responds to a variety of volatile substances. At present, electronic nose technology based on gas sensor arrays has been widely used in medical diagnosis, environmental testing, tobacco identification, food and other fields; however, its detection accuracy and sensitivity need to be improved. On the other hand, there are currently few reports on the use of electronic noses in the detection of wheat mold, and the volatile gas components of wheat in the early stage of mold are relatively low, which also limits the detection accuracy and sensitivity of the electronic nose. Summary of the invention

[0006] In order to solve the shortcomings of the existing grain mildew detection methods, such as high detection cost, large experimental instruments, complex sample pretreatment, long analysis time and inability to monitor in real time, the present invention provides a method for detecting wheat with different degrees of mildew during grain storage in a granary using an electronic nose; based on the relationship between the voltage response signal of the electronic nose and the detection results of the national standard method, an electronic nose system for detecting wheat with different degrees of mildew is invented. By optimizing the design of the sensor air chamber and collecting the wheat detection and sensor voltage response signals, the construction of the recognition model is completed, and the real-time and accurate prediction of the odor concentration and the degree of mildew is further realized. According to the prediction results, an early warning is actively issued and the mildewed wheat is controlled, which greatly avoids the harm of mildewed wheat to the health of humans and animals. Provide technical support for remote real-time monitoring of the harmful activities of stored grain mold and improve the early prediction theory of grain storage safety.

[0007] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0008] The present invention first provides an electronic nose system: the electronic nose system is composed of a data storage module, a wireless communication module, a conditioning circuit board, a single chip microcomputer, an air inlet pipe, a sensor array, a sensor air chamber, a connecting pipe, an air outlet pipe and an air pump;

[0009] Among them, the wireless communication module and the data storage module are located below the conditioning circuit board, a single chip microcomputer is provided on the surface of the conditioning circuit board, the conditioning circuit board is located below the sensor air chamber, one end of the sensor air chamber is connected to the air inlet pipe, and the other end is connected to the air pump through a connecting pipe; at the same time, one end of the air pump is also connected to the air outlet pipe;

[0010] A sensor array is provided inside the sensor air chamber, and the sensor array is electrically connected to the data storage module so that the detected data can be stored through the data storage module;

[0011] In addition, the data storage module is electrically connected to the wireless communication module so as to upload the data in the data storage module to the database of the terminal by using the wireless communication module;

[0012] The interior of the sensor air chamber is provided with Y guide plates, where Y is a positive integer. The guide plates are independent of each other and do not contact each other. One end of the guide plate is connected to the side wall of the air chamber, and there is a gap between the other end and the side wall of the air chamber to facilitate gas flow and extend the flow path of the volatile gas of wheat in the air chamber, thereby ensuring sufficient contact between the sensor and the gas.

[0013] The guide plate divides the interior of the sensor air chamber into N independent parts, and n sensor arrays are correspondingly arranged in the N independent parts; N and n are both positive integers;

[0014] The sensor array is a gas sensor; the sensor array, the data storage module and the wireless communication module are all electrically connected to the single chip microcomputer.

[0015] Preferably, the terminal includes a mobile phone, a computer, or a tablet computer.

[0016] Preferably, the data storage module includes a data storage card;

[0017] Preferably, the sensor air chamber is in the shape of a hexagonal prism with an open bottom, and has a built-in sensor array; the two ends of the sensor air chamber are an air inlet pipe and a connecting pipe respectively; the bottom surface of the sensor air chamber is snapped with the conditioning circuit board and contains the sensor array.

[0018] Preferably, a fixing structure is provided at each connection of the six sides of the hexagonal prism of the sensor air chamber; the fixing structure is fixedly connected to the conditioning circuit board to seal the sensor air chamber; so that air can only be taken in and out through the air inlet pipe and the connecting pipe.

[0019] Preferably, the fixing structure is a cylindrical structure, and the cylindrical structure has built-in screws fixedly connected to the conditioning circuit board;

[0020] Preferably, the guide plate is a T-shaped structure, a cross structure or a staggered cross structure; there are 4 guide plates, which divide the inner part of the sensor chamber into 8 independent parts, that is, N=8, corresponding to 8 sensor arrays, namely TGS2611, TGS2600, MQ6, TGS2620, TGS2610, TGS2602, MQ3, and MQ138.

[0021] The present invention also provides an electronic nose system for detecting different degrees of mildew in grains, comprising the following steps:

[0022] (1) Sample processing:

[0023] Firstly, the grain sample is soaked in a sampling bottle containing sterile water, and allowed to stand to balance the moisture to obtain a sample with balanced moisture; then the grain sample with balanced moisture is transferred to a constant temperature and humidity chamber for storage;

[0024] Samples are collected at different storage times (t); the collected samples are recorded as mold degree samples, and are correspondingly marked as mold 0, mold 1, mold 2, ... mold t, where t is a positive integer;

[0025] The grain samples include wheat, rice and oats; the temperature of the static equilibrium moisture is 4°C and the time is 36-48h; the temperature of the constant temperature and humidity chamber is 35°C and the humidity is 95%;

[0026] The sampling at different storage times (t) refers to sampling on the 0th day, the 2nd day, the 4th day, the 6th day and the 8th day; wherein the sample on the 0th day refers to a sample with balanced moisture;

[0027] (2) Performing sample analysis based on the electronic nose system includes the following steps:

[0028] S1. During the headspace process, place the moldy sample in a sampling bottle and let it stand for a period of time for future use;

[0029] S2, after standing and volatilizing, the air inlet pipe is connected to the sampling bottle, and the headspace gas is transported to the sensor air chamber through the air inlet pipe by an air pump, so that the headspace gas reacts with the sensor array in the sensor air chamber, and a voltage value signal (i.e., odor sample data) is obtained, and the data storage module and the wireless communication module are used to transmit and store the data, thereby completing the information collection of the moldy degree sample;

[0030] When the information collection of a mildew degree sample is completed, before the next sample information collection is carried out, the sensor air chamber is purged by connecting the air inlet pipe with clean air to ensure that the response curve of the sensor array returns to the baseline state;

[0031] S3. Determination of total colony count: the total colony count of the sample corresponding to the degree of mold; according to the total colony count, the grain samples can be divided into three stages: no mold [<2.7log(CFU / g)], beginning to mold [2.7~4log(CFU / g)] and severe mold [>4log(CFU / g)];

[0032] The total bacterial count was determined according to GB4789.2-2016 standard. The operation method was as follows: 25 g of the moldy sample was weighed into a beaker, and 225 mL of phosphate buffer or saline was added; the grain sample was shaken for 60 seconds using an oscillator to make a 10 -1 The bacterial suspension was diluted to a certain degree, and then spread on agar medium after gradient dilution, and cultured at 36℃ for 28±2h for statistics;

[0033] S4, data preprocessing: performing data preprocessing on the voltage value signal (i.e., odor sample data) obtained in step S2, i.e., extracting features from the voltage value signal obtained by the sensor array, and selecting the voltage stability value as the gas sample feature data;

[0034] Among them, the voltage value signal of 120s is selected as the characteristic data of the gas sample; (the voltage value signal rises with the increase of the gas concentration in the gas chamber, and the curve gradually slows down and tends to be stable after 120s; therefore, the voltage stability value of 120s is selected as the characteristic value processing);

[0035] S5. Build a predictive regression model:

[0036] S5.1. The characteristics of samples with different degrees of mildew are different. The collected voltage stability values ​​are used as characteristic values, and the characteristic values ​​are used as input data to apply to the parameters of the LDA model, and the LDA is reduced in dimension to obtain a visualized three-dimensional LDA graph.

[0037] S5.2, using the gas sample characteristic data obtained in S4 as input data of SVM, training the SVM model for qualitative identification, and classifying grains with different degrees of mold;

[0038] S5.3. Quantitative prediction of the total number of bacterial colonies in grains using the BPNN model includes the following steps:

[0039] S5.3.1. Set an H×N element electronic nose data input matrix, where the columns of the matrix include the voltage stability values ​​of N sensors, and H is the steady-state voltage value of H samples; at the same time, the total number of colonies of the samples corresponding to the degree of mold is stored in an H×1 matrix;

[0040] S5.3.2. Use BPNN method to build regression model:

[0041] A linear relationship is formed between the voltage stability value and the total number of colonies of the corresponding moldy samples, and a global calibration curve of the voltage response signal of the electronic nose of grain is obtained. A prediction regression model is constructed based on the BPNN method; S6, actual warehouse warning:

[0042] The electronic nose system was placed at different locations in the granary, and the voltage stability values ​​of samples at different locations were measured; then the voltage stability values ​​were input into the constructed prediction regression model to predict the total colony count of the grain sample. The moldy condition of the grain was judged based on the total colony count. If the total colony count was <2.7log(CFU / g), it meant that the grain was not moldy; if the total colony count was in the range of 2.7 to 4log(CFU / g), it meant that the grain had started to mold; if the total colony count was >4log(CFU / g), it meant that the grain was seriously moldy.

[0043] If the test result is less than 2.7log(CFU / g), store or use it normally; if the test result is within the range of starting mold, treat the grain with food-grade mildew inhibitor to control and inhibit mold growth. If the test result is severe mold, clean the storage facilities to prevent mold from contaminating new batches of grains again, and determine the cause of the mold to avoid severe mold in the future.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] 1. The present invention has the characteristics of miniaturization, integration and wirelessness. The gas sensor is integrated into a conditioning circuit board, reducing the volume of the air chamber. The air chamber is also designed to add a guide plate structure. The guide plates are independent of each other and do not touch each other. One end of the guide plate is connected to the side wall of the air chamber, and there is a gap between the other end and the side wall of the air chamber, which is convenient for gas flow and prolongs the flow path of volatile gas from grains in the air chamber, thereby ensuring that the sensor is in full contact with the gas. While increasing the flow path of the gas, the time for gas washing is reduced. In addition, the guide plate can accelerate the flow of gas, and when cleaning the sensor air chamber, the sensor array can quickly return to the baseline level, effectively saving the sampling process time.

[0046] 2. The method for detecting mildew in grain in granaries using an electronic nose system and a machine learning algorithm proposed in the present invention can control both the collection process and data transmission by the same chip, and can remotely control the collection process and upload the received data to the database of the terminal. The degree of mildew can be accurately determined and quantitatively predicted, and it has the advantages of high accuracy and fast detection, and is suitable for on-site rapid qualitative analysis of grains in granaries; and it also has the characteristics of low cost, simplicity and high efficiency.

[0047] 3. The present invention detects the grains stored in the granary and issues an early warning based on the predicted total bacterial count of the grains, thereby avoiding the mildew of the grains to a great extent. This not only helps to protect the health of consumers and food safety, but also can improve the quality of grains, reduce waste and improve production efficiency, which is of great significance to the entire food industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is the overall framework diagram of the present invention;

[0049] Figure 2 It is a schematic diagram of the structure of the electronic nose system; the figure numerals are: 1-data storage module, 2-wireless communication module, 3-conditioning circuit board, 4-single chip microcomputer, 5-inlet pipe, 6-sensor array, 7-sensor air chamber, 8-connecting pipe, 9-outlet pipe, 10-air pump.

[0050] Figure 3It is the sensor air chamber; the figure numbers are: 5-inlet pipe, 7-2-cylinder one, 7-3-cylinder two, 7-4-bottom surface, 7-5-cylinder three, 8-connecting pipe, 7-6-cylinder four, 7-7-guide plate one, 7-8-guide plate two, 7-9-cylinder five, 7-1-cylinder six.

[0051] Figure 4 It is a sensor array; the figure numerals are: 11 - sensor one, 12 - sensor two, 13 - sensor three, 14 - sensor four, 15 - sensor five, 16 - sensor six, 17 - sensor seven, 18 - sensor eight.

[0052] Figure 5 It is a data processing flow chart;

[0053] Figure 6 is the change of total colony count with the number of days of mildew;

[0054] Figure 7 To visualize the PCA graph;

[0055] Figure 8 To visualize the LDA graph;

[0056] Fig. 9 Test the SVM training set;

[0057] Fig.10 Test on the SVM test set;

[0058] Fig.11 Test for PLSR training set;

[0059] Fig.12 Tested on the PLSR test set;

[0060] Fig.13 Test the BPNN training set;

[0061] Fig.14 Tested on the BPNN test set. Specific implementation methods

[0063] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention; obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The connections mentioned in the present invention include wired and wireless electrical connection methods, among which wired connections include wire connections, USB connections and the like; wireless includes Bluetooth connections and the like.

[0064] The grain sample used in this embodiment is wheat, the wheat variety used is Wanmai (harvested in 2023), and the initial moisture content of the wheat is 12.4%.

[0065] Embodiment 1:

[0066] The present invention first provides an electronic nose system: Figure 2 As shown: the electronic nose system consists of a data storage card 1, a wireless communication module 2, a conditioning circuit board 3, a single-chip microcomputer 4, an air inlet pipe 5, a sensor array 6, a sensor air chamber 7, a connecting pipe 8, an air outlet pipe 9 and an air pump 10; wherein the wireless communication module 2 and the data storage card 1 are located below the conditioning circuit board 3, a single-chip microcomputer 4 is provided on the conditioning circuit board 3, the conditioning circuit board 3 is located below the sensor air chamber 7, one end of the sensor air chamber 7 is connected to the air inlet pipe 5, and the other end is connected to the air pump 10 through the connecting pipe 8; at the same time, one end of the air pump 10 is also connected to the air outlet pipe 9; a sensor array 6 is provided inside the sensor air chamber 7, and the sensor array 6, the data storage module 1 and the wireless communication module 2 are all electrically connected to the single-chip microcomputer; the data detected by the sensor array 6 is stored through the data storage card 1; in addition, the wireless communication module 2 uploads the data to the terminal database;

[0067] The air pump 10 pumps the gas in the sample into the sensor chamber 7, and the voltage response signal generated by the sensor array 6 is transmitted to the database through the single chip computer 4 and saved in the data storage card 1. After the sample collection is completed, the air intake pipe 5 is connected to clean air for air washing to restore the baseline state of the sensor.

[0068] Design of sensor air chamber 7: Since the volatile gas components of wheat in the early stage of mildew are relatively low and the detection accuracy and sensitivity of the electronic nose are limited, this study optimized the design of sensor air chamber 7.

[0069] The sensor air chamber 7 is in the shape of a hexagonal prism with an open bottom surface. Figure 3 As shown, the two ends of the sensor air chamber 7 are respectively the air inlet pipe 5 and the connecting pipe 8. The bottom surface 7-4 is buckled with the conditioning circuit board 3 to include the sensor array 6. A fixing structure is provided at each connection of the six sides of the hexagonal prism of the sensor air chamber; the fixing structure is a cylindrical structure; the six cylindrical structures around the sensor air chamber 7, cylinder 1 7-2, cylinder 2 7-3, cylinder 7-5 3, cylinder 4 7-6, cylinder 5 7-9 and cylinder 6 7-1, are tightened and reinforced with the conditioning circuit board 3 by screws.

[0070] An opening for the conditioning circuit board 3 is provided on the bottom surface of the sensor air chamber 7; the bottom surface of the sensor air chamber 7 is consistent in size with the conditioning circuit board 3, and can completely seal the air chamber, and is fixedly connected to the air chamber through a fixed structure, forming a completely closed environment except for the air inlet and outlet; the interior of the sensor air chamber 7 is provided with 4 guide plates, and the guide plates are independent of each other and do not contact each other, one end of the guide plate is connected to the side wall of the air chamber, and the other end has a gap with the side wall of the air chamber, which is convenient for gas flow and prolongs the flow path of wheat volatile gas in the air chamber, thereby ensuring full contact between the sensor and the gas. In addition, the guide plate can accelerate the flow of gas, and when cleaning the sensor air chamber, the sensor array can quickly return to the baseline level, effectively saving the sampling process time.

[0071] In this embodiment, four guide plates are provided, which divide the interior of the sensor air chamber 7 into eight independent parts, and eight sensors are provided in the eight independent parts, namely, sensor one 11, sensor two 12, sensor three 13, sensor four 14, sensor five 15, sensor six 16, sensor seven 17 and sensor eight 18;

[0072] like Figure 4 As shown, the sensor array 6 contains a total of 8 MOS sensors, and sensors 1 to 18 correspond to TGS2611, TGS2600, MQ6, TGS2620, TGS2610, TGS2602, MQ3 and MQ138 respectively; among them, TGS2611, TGS2600, MQ6, TGS2620, TGS2610, TGS2602 are produced by Figaro Company of Japan. MQ3 and MQ138 are produced by Weisheng Company of China. The gas sensor has cross-sensitivity characteristics, so that the difference information between wheat with different degrees of mold can be obtained.

[0073] An electronic nose system is used for wheat mildew detection in a granary, comprising the following steps:

[0074] (1) Wheat samples were collected using sampling bottles at different treatments in a granary in Zhenjiang City, Jiangsu Province.

[0075] The collected wheat samples were soaked in a sampling bottle containing sterile water, placed in a refrigerator at 4°C to balance the moisture, and the moisture balance time was set to 36 hours to obtain a sample with balanced moisture.

[0076] The wheat samples with balanced moisture are then transferred to a constant temperature and humidity chamber with the temperature set to 35°C and the humidity set to 95%. Samples are collected on the 0th day, the 2nd day, the 4th day, the 6th day and the 8th day, respectively. The wheat samples obtained on different days are recorded as moldy samples, recorded as moldy 0, moldy 2, moldy 4, moldy 6 and moldy 8. Corresponding to the 5 different degrees of mildew of wheat, 30 parallel samples are set for each degree, and the test is repeated twice a day, that is, there are 60 samples for each degree of mildew. Among them, the sample on day 0 refers to the sample with balanced moisture, which has not yet been cultured in a constant temperature and humidity chamber; the sample on day 2 is the sample transferred to the constant temperature and humidity chamber and cultured for 1 day (24h), and so on;

[0077] (2) Using the electronic nose system to perform sample analysis includes the following steps:

[0078] 2.1) The sampling frequency of the single chip microcomputer 4 is set to collect data once per second. Each sampling is divided into two stages: odor sample data collection and sensor array cleaning. The time of the two stages is set to 5 minutes and 3 minutes respectively;

[0079] 2.2) During the headspace process, place the moldy samples in a sampling bottle and let them stand for 60 minutes to evaporate, so that the volatile gases of the wheat can be fully evaporated and enriched in the sampling bottle;

[0080] 2.3) After standing and volatilizing, connect the sampling bottle containing the wheat sample to the air inlet pipe 5, use the air pump 10 to transport the headspace gas to the sensor chamber 7, so that the headspace gas reacts with the sensor array 6 in the sensor chamber, obtain the voltage value signal (i.e., the odor sample data), and transmit and store it through the data storage card 1 and the wireless communication module 2. Before the next measurement, wash the sensor chamber with clean air to ensure that the reaction curve of the sensor array 6 returns to the baseline state;

[0081] (3) Determination of total colony count of samples: Referring to GB4789.2-2016 standard, weigh 25g of moldy sample into a beaker and add 225mL of phosphate buffer or saline. Use an oscillator to shake the wheat sample for 60s to make a 10 -1 The bacterial suspension was diluted to different degrees of mildew. 100 μL of the bacterial suspension of each dilution was taken and cultured in plate counting agar medium. Three parallels were made for each sample and cultured at 36℃ for 28±2h. The total number of colonies changed with the number of days of mildew as shown in the figure. Figure 6 As shown;

[0082] (4) Data preprocessing: The voltage value signal (i.e., odor sample data) obtained in step 2.3 is used for data preprocessing, that is, feature extraction is performed on the voltage value signal obtained by the gas sensor array 6, and the voltage stability value is selected as the gas sample feature data; the voltage value signal increases with the increase of the gas concentration in the gas chamber, and the curve gradually slows down after 120 seconds; therefore, the voltage stability value of 120 seconds is selected as the feature value for processing.

[0083] (5) Constructing a predictive regression model:

[0084] The voltage stability value collected in step (4) is used as a characteristic value, and principal component analysis (PCA), linear discriminant analysis (LDA), support vector machine (SVM), partial least squares regression (PLSR) and back propagation neural network (BPNN) are used for pattern recognition and data analysis, including the following steps:

[0085] 5.1) It was determined that the characteristics of wheat samples with different degrees of mildew in the granary were different. The voltage stability value was used as input data to apply to the parameters of the PCA and LDA models. Visual PCA and LDA graphs of wheat samples with 5 degrees of mildew were obtained. There were 60 samples in each category. The PCA and LDA graphs were respectively as follows: Figure 7 and Figure 8 As shown in the figure, the results show that LDA has a good classification effect on wheat with different degrees of mold. It proves that the electronic nose system can distinguish the volatile content of different samples. Among them, there is some overlap in the PCA analysis of samples from 2 to 6 days, which may be caused by the basically consistent composition of volatile gases, but the different volatile amounts. However, LDA analysis can make a clear distinction.

[0086] 5.2) Use the characteristic value (voltage stability value) as the input data of the SVM model, train the SVM model for qualitative identification, and classify wheat with different degrees of mildew. Use the radial basis kernel function to establish a classifier model; train the SVM model for qualitative identification and determine the degree of mildew of wheat. Among them, mildew 0, mildew 2, mildew 4, mildew 6 and mildew 8 correspond to sample labels 1, 2, 3, 4, 5;

[0087] The test samples were 5 degrees of mildew, each sample had 60 repetitions, a total of 300 groups, 225 groups as training sets, and 75 groups as test sets. The accuracy of the training set was 97.33% (219 / 225). The accuracy of the test set was 96% (72 / 75). Fig. 9 In the model prediction of the training set, only two of the 2, 3, and 4 categories were incorrectly predicted as other categories. Fig.10 Only 3 of the 5 degrees in the test set were misjudged. This shows that SVM can distinguish different degrees of wheat mold. The SVM analysis and discrimination model is shown in Table 1.

[0088] Table 1 Results of SVM discriminant analysis of wheat with different mildew degrees

[0089]

[0090] 5.3)PLSR is a method that can use multiple predictor variables to model one or more response variables, especially suitable for multicollinearity and high-dimensional data. BPNN is widely used due to its excellent function approximation ability. The quantitative prediction of total wheat colony count using PLSR and BPNN models includes the following steps:

[0091] 5.3.1) Set up an input matrix of 150×8 elements, with 5 rows for each degree of mildew and 30 rows for each degree. The columns include the voltage stability values ​​of the 8 sensors in the electronic nose; the total number of colonies of the samples corresponding to the degree of mildew measured using the national standard method is stored in a 150×1 matrix;

[0092] 5.3.2) Based on the above data, the regression coefficient was calculated and PLSR regression was used to quantitatively link the characteristic value (voltage stability value) with the degree of fungal reproduction on wheat. The data set contains 150 moldy wheat samples (30 samples per day), which are randomly divided into a training set of 100 training samples and a test set of 50 test samples. The RMSE and R of the training set and test set are 2 Value Fig.11 and Fig.12 As shown in Figure 2, it can be seen that the PLSR regression model based on the voltage stability value detected by the electronic nose system has good performance on both the training set and the test set (R 2 >0.85). The predicted colony count of moldy wheat by the PLSR model based on the training set was positively correlated with the actual colony count (R 2 c=0.916,RMSE=0.432). Fig.13 and Fig.14 It shows that the effect of PLSR model is slightly worse than that of BPNN model (R 2 c=0.992,RMSE=0.087). Higher R 2 The lower RMSE and the smaller RMSE indicate that the colony count estimated by the BP neural network is closer to the true value. Similarly, the prediction model of BPNN is more accurate (R 2 p=0.988, RMSE=0.123).

[0093] In summary, the BP neural network can better predict the number of colonies in wheat with different degrees of mildew. It proves that the electronic nose technology based on sensor array has good feasibility in wheat mildew identification, providing a new perspective for early mildew detection of wheat and other grains.

[0094] Note: The above embodiments are only used to illustrate the design ideas and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, any equivalent changes or modifications made based on the principles and design ideas disclosed by the present invention are within the protection scope of the present invention.

Claims

1. An electronic nose system, characterized in that: The electronic nose system is composed of a data storage module (1), a wireless communication module (2), a conditioning circuit board (3), a single chip computer (4), an air inlet pipe (5), a sensor array (6), a sensor air chamber (7), a connecting pipe (8), an air outlet pipe (9) and an air pump (10); The wireless communication module (2) and the data storage module (1) are located below the conditioning circuit board (3), a single chip computer (4) is provided on the surface of the conditioning circuit board (3), the conditioning circuit board (3) is located below the sensor air chamber (7), one end of the sensor air chamber (7) is connected to the air inlet pipe (5), and the other end is connected to the air pump (10) through the connecting pipe (8); at the same time, one end of the air pump (10) is also connected to the air outlet pipe (9); A sensor array (6) is provided inside the sensor air chamber (7), and the sensor array (6) is electrically connected to the data storage module (1), so that the detected data can be stored through the data storage module (1); In addition, the data storage module (1) is electrically connected to the wireless communication module (2) so that the data in the data storage module (1) can be uploaded to a database in the terminal using the wireless communication module (2); The sensor air chamber (7) is provided with Y guide plates inside, where Y is a positive integer. The guide plates are independent of each other and do not contact each other, and one end of the guide plate is connected to the side wall of the air chamber, while the other end has a gap with the side wall of the air chamber, so as to facilitate gas flow and extend the flow path of the volatile gas of wheat in the air chamber, thereby ensuring that the sensor is in full contact with the gas. The guide plate divides the interior of the sensor air chamber (7) into N independent parts, and n sensor arrays are correspondingly arranged in the N independent parts; N and n are both positive integers; The sensor array (6) is a gas sensor; The sensor array (6), the data storage module (1) and the wireless communication module (2) are all electrically connected to the single chip computer (4).

2. An electronic nose system according to claim 1, characterized in that: The terminals include mobile phones, computers, and tablet computers.

3. An electronic nose system according to claim 1, characterized in that: The data storage module comprises a data storage card; preferably, the sensor air chamber (7) is in the shape of a hexagonal prism with an open bottom surface, and has a built-in sensor array; the two ends of the sensor air chamber (7) are respectively an air inlet pipe (5) and a connecting pipe (8); the bottom surface (7-4) of the sensor air chamber (7) is engaged with the conditioning circuit board (3), and contains the sensor array.

4. An electronic nose system according to claim 1, characterized in that: A fixing structure is provided at each connection of the six sides of the hexagonal prism of the sensor air chamber (7); the fixing structure is fixedly connected to the conditioning circuit board (3) so as to seal the sensor air chamber (7); so that air can only be taken in and out through the air intake pipe (5) and the connecting pipe (8).

5. An electronic nose system according to claim 4, characterized in that: The fixing structure is a cylindrical structure, and the cylindrical structure has built-in screws fixedly connected to the conditioning circuit board (3).

6. An electronic nose system according to claim 1, characterized in that: The guide plate is a T-shaped structure, a cross structure or a staggered cross structure; there are four guide plates, which divide the interior of the sensor air chamber (7) into eight independent parts, that is, N=8, corresponding to eight sensor arrays, namely TGS2611, TGS2600, MQ6, TGS2620, TGS2610, TGS2602, MQ3, and MQ138.

7. Use of the electronic nose system according to any one of claims 1 to 6 for detecting different degrees of mildew in grains, characterized in that: Here are the steps: (1) Sample processing: Firstly, the grain sample is soaked in a sampling bottle containing sterile water, and allowed to stand to balance the moisture to obtain a sample with balanced moisture; then the grain sample with balanced moisture is transferred to a constant temperature and humidity chamber for storage; Samples were collected at different storage times (t); The collected samples are recorded as moldy samples, and are marked as moldy 0, moldy 1, moldy 2, ... moldy t, where t is a positive integer; (2) Performing sample analysis based on the electronic nose system includes the following steps: S1. During the headspace process, place the moldy sample in a sampling bottle and let it stand for a period of time for future use; S2, after standing and volatilizing, the air inlet pipe (5) is connected to the sampling bottle, and the air pump (10) is used to transport the headspace gas through the air inlet pipe (5) to the sensor air chamber (7), so that the headspace gas reacts with the sensor array (6) in the sensor air chamber (7), obtains a voltage value signal, and transmits and stores it through the data storage module (1) and the wireless communication module (2), thereby completing the information collection of the mold degree sample; When the information collection of a mildew degree sample is completed, before the information collection of the next sample is performed, the sensor air chamber (7) is purged by connecting the air inlet pipe (5) with clean air to ensure that the response curve of the sensor array returns to the baseline state; S3. Determination of total colony count: the total colony count of the sample corresponding to the degree of mold; according to the total colony count, the grain samples can be divided into three stages: no mold [<2.7log(CFU / g)], beginning to mold [2.7~4log(CFU / g)] and severe mold [>4log(CFU / g)]; S4, data preprocessing: performing data preprocessing on the voltage value signal obtained in step S2, that is, extracting features from the voltage value signal obtained by the sensor array (6), and selecting a voltage stability value as gas sample feature data; S5. Build a predictive regression model: S5.

1. The characteristics of samples with different degrees of mildew are different. The collected voltage stability values ​​are used as characteristic values, and the characteristic values ​​are used as input data to apply to the parameters of the LDA model, and the LDA is reduced in dimension to obtain a visualized three-dimensional LDA graph. S5.2, using the gas sample characteristic data obtained in S4 as input data of SVM, training the SVM model for qualitative identification, and classifying grains with different degrees of mold; S5.

3. Quantitative prediction of the total number of bacterial colonies in grains using the BPNN model includes the following steps: S5.3.

1. Set an H×N element electronic nose data input matrix, where the columns of the matrix include the voltage stability values ​​of N sensors, and H is the steady-state voltage value of H samples; at the same time, the total number of colonies of the samples corresponding to the degree of mold is stored in an H×1 matrix; S5.3.

2. Use BPNN method to build regression model: A linear relationship was formed between the voltage stability value and the total number of bacterial colonies in the sample with the corresponding degree of mold, and a global calibration curve of the voltage response signal of the electronic nose of grain was obtained. A prediction regression model was constructed based on the BPNN method. S6. Actual warehouse warning: The electronic nose system was arranged at different positions in the granary, and the voltage stability values ​​of samples at different positions were measured. The voltage stability values ​​were then input into the constructed prediction regression model to predict the total colony count of the grain sample. The moldy state of the grain was judged based on the total colony count. If the total colony count is <2.7log(CFU / g), it means that the grain is not moldy; if the total colony count is in the range of 2.7 to 4log(CFU / g), it means that the grain has begun to mold; if the total colony count is >4log(CFU / g), it means that the grain is seriously moldy.

8. The use according to claim 6, characterized in that The cereal samples in step (1) include wheat, rice, and oats; the temperature of the static moisture equilibrium is 4° C., and the time is 36-48 hours; the temperature of the constant temperature and humidity chamber is 35° C., and the humidity is 95%; and collecting samples at different time periods refers to collecting samples on the 0th day, the 2nd day, the 4th day, the 6th day, and the 8th day; The sample on day 0 refers to the sample at equilibrium moisture.

9. The use according to claim 6, characterized in that: The total bacterial count determination in step S3 refers to GB4789.2-2016 standard, and the operation method is as follows: weigh 25g of the moldy sample in a beaker, add 225mL of phosphate buffer or saline; use an oscillator to shake the grain sample for 60s to make 10 -1 The bacterial suspension was diluted to different degrees and then spread on agar medium after gradient dilution, and cultured at 36℃ for 28±2h for statistics.

10. The use according to claim 6, characterized in that In step S4, the voltage value signal of 120s is selected as the gas sample characteristic data.

Images