A method for simultaneously detecting the chemical composition content and identifying physical impurities in fermentation broth

Abstract

This invention relates to a method for simultaneously detecting the chemical composition content and identifying physical impurities in fermentation broth, comprising the following steps: obtaining anaerobic fermentation broth; acquiring hyperspectral data of the fermentation broth using near-infrared hyperspectral technology; establishing prediction models for total organic carbon and total nitrogen in the anaerobic fermentation broth using the average spectral data of the near-infrared hyperspectral images combined with machine learning algorithms; and automatically identifying physical impurities in the fermentation broth using the spectral and positional information of the near-infrared hyperspectral images and a binary classification model. Compared with existing technologies, the method provided by this invention eliminates the need for sample homogenization and centrifugation pretreatment operations, has a short detection time, and can simultaneously perform chemical property analysis and mechanical impurity identification of anaerobic fermentation broth, providing a multi-dimensional description of the anaerobic fermentation process and thus achieving precise monitoring of the anaerobic fermentation process.

Description

Technical Field

[0001] This invention relates to the field of bio-fermentation broth characteristic detection and bio-fermentation process control technology, and in particular to a method for simultaneously detecting the chemical composition content of fermentation broth and identifying physical impurities. Background Technology

[0002] Anaerobic digestion of biomass waste can produce biogas and soil conditioners such as sludge. In the global context of pollution reduction and carbon reduction in solid waste management, anaerobic digestion technology for biomass waste has been widely applied. However, anaerobic digestion plants may suffer from unstable operation and low efficiency. Monitoring the physicochemical indicators of the fermentation broth is crucial for ensuring stable and efficient operation of anaerobic digestion.

[0003] The analysis of the physical and chemical properties of anaerobic fermentation broth is crucial for the stable operation of anaerobic digestion. Traditional methods include titration, gas chromatography, and electrochemical methods. ① Titration is simple to operate and can simultaneously determine alkalinity and VFA concentrations, but it requires solid-liquid separation pretreatment for samples with high solids content, increasing the complexity of the analytical process. ② Electrochemical methods are not suitable for volatile fatty acid concentrations exceeding 1000 mg / L. ③ Gas chromatography online monitoring requires a complex sample pretreatment system (filtration, dilution), which requires frequent maintenance due to membrane fouling. Furthermore, current methods are limited to the analysis of the chemical properties of the fermentation broth and cannot detect physical instability factors such as impurities and foam. Currently, impurities and foam in the fermentation broth can only be detected manually. Therefore, developing rapid, non-destructive, and multi-indicator simultaneous detection methods for the chemical properties and physical impurities of fermentation broth is of great significance.

[0004] Near-infrared hyperspectral images have high spectral resolution, which can effectively identify the chemical composition of the target object. In addition, based on the spatial distribution characteristics of the spectrum of near-infrared hyperspectral images, the morphology and texture details of the target object (debris) can be obtained. This allows near-infrared hyperspectral images to identify foreign objects in the fermentation broth while detecting the chemical composition of the fermentation broth sample. Chinese patent CN116306322A discloses a method for inverting the total phosphorus concentration in water based on hyperspectral data. Its features include: (1) acquiring hyperspectral data of the water body to be detected and preprocessing it; (2) using a variational autoencoder to extract features from the preprocessed hyperspectral data of the water body. However, this method only uses hyperspectral detection to detect the low concentration of phosphorus in natural water bodies. It cannot be used for the composition analysis of high-concentration fermentation broth in industrial fermentation devices, nor can it be used for the identification of physical foreign objects in the fermentation broth.

[0005] In summary, there is an urgent need to provide a method for simultaneously detecting the chemical composition content of fermentation broth and identifying physical impurities based on near-infrared hyperspectral imaging, laying the foundation for multi-dimensional real-time online process monitoring of anaerobic digestion. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art by providing a method that is time-efficient, eliminates the need for fermentation broth drying, simplifies the chemical property detection steps, and simultaneously detects the chemical composition content of fermentation broth and identifies physical impurities.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] This invention provides a method for simultaneously detecting the chemical composition content and identifying physical impurities in fermentation broth, comprising the following steps:

[0009] S1: Select a bioreactor for testing fermentation broth samples;

[0010] S2: Collect fermentation broth samples from the bioreactor in S1;

[0011] S3: Obtain the near-infrared hyperspectral image of the fermentation broth sample to be tested in S2;

[0012] S4: Process the near-infrared hyperspectral image of the fermentation broth sample obtained in S3, including the following steps:

[0013] S4-1 uses an organic carbon and total nitrogen analyzer to test the total organic carbon and total nitrogen content in the fermentation broth sample of S3 after near-infrared hyperspectral testing;

[0014] S4-2 takes the near-infrared hyperspectral images acquired in S3, selects a region of a preset size in each image as the region of interest, calculates and derives the average spectrum in the region of interest, preprocesses the average spectrum, and divides the processed spectral data into a training set and a test set.

[0015] S4-3 selects regions of interest (ROIs) for physical impurities and fermentation broth matrix on the near-infrared hyperspectral image acquired in S3, obtains standardized spectral curves for these two ROIs, and uses a binary classification model to identify physical impurities. The detection accuracy is calculated based on the actual number of physical impurities and the detection results. Furthermore, the detection accuracy is the ratio of the number of impurities detected by the binary classification model to the actual number of impurities.

[0016] S5: Based on the total organic carbon and total nitrogen content of the fermentation broth samples obtained from S4-1 in the training set, and the average spectral information of the fermentation broth samples obtained from S4-2, establish a regression prediction model for the total organic carbon and total nitrogen content of the fermentation broth samples, and use the test set to detect the model accuracy, and further optimize the prediction model for the organic carbon and total nitrogen content of the fermentation broth.

[0017] S6: Collect anaerobic fermentation broth samples containing physical impurities, obtain hyperspectral data of the fermentation broth samples using near-infrared hyperspectral technology, and identify the quantity and area of ​​impurities using the binary classification model constructed in S4-3;

[0018] S7: Collect the fermentation broth sample to be tested, obtain the hyperspectral data of the fermentation broth sample using near-infrared hyperspectral technology, and predict the organic carbon content and total nitrogen content of the fermentation broth sample using the prediction model of organic carbon and total nitrogen content constructed in S5.

[0019] Furthermore, in S1, four bioreactors are set up, and the initial organic load of organic waste fed into the bioreactors is 1-3 kg-VS / L / d. The bioreactors feed and discharge materials every 1-2 days, and collect fermentation broth samples.

[0020] Two of the four bioreactors, C-1 and C-2, are fed only with organic matter; two parallel reactors, N-1 and N-2, are fed with an additional 50-100 mg of ammonium chloride per liter of effective reactor volume.

[0021] All four bioreactors operate in a semi-continuous mode.

[0022] Furthermore, in S1, the organic waste feed includes one or more of the following: kitchen waste, catering waste, sewage sludge, agricultural and forestry waste, and poultry and livestock manure;

[0023] The number of fermentation broth samples collected was greater than 100.

[0024] Furthermore, in S1, the total organic carbon content of the collected fermentation broth sample ranges from 1253 mg-C / L to 12507 mg-C / L.

[0025] Furthermore, in S1, the total nitrogen content of the collected fermentation broth sample ranges from 2237 mg-N / L to 12168 mg-N / L.

[0026] Furthermore, in S2, a fermentation broth sample of more than 15 mL was taken from the four bioreactors in S1 and placed in a glass petri dish as the fermentation broth sample to be tested.

[0027] In S3, a near-infrared hyperspectral camera is used to acquire near-infrared hyperspectral images of the fermentation broth sample to be tested in S2. By adjusting the object distance, exposure time, frame rate range, and light source intensity, the clearest near-infrared hyperspectral image of the fermentation broth sample to be tested is obtained.

[0028] Furthermore, in S4-2, the method for dividing the training set and the test set is to divide the data of each sample into a training set and a test set based on a Random function;

[0029] The training set comprises 60% to 80% of all sample data, and the test set comprises 20% to 40% of all sample data.

[0030] Furthermore, in S4-3, the binary classification model is obtained through one of the following methods: matched filter, spectral angle mapper, constrained energy minimization, and adaptive consistency estimator.

[0031] Furthermore, in S4-3, the physical impurities include wood chips, plastic sheets, and suspended particles and foam formed during fermentation that are mixed in with the feed.

[0032] Furthermore, in S5, the regression prediction model for total organic carbon and total nitrogen in the fermentation broth is obtained through one of the following methods: random forest regression, support vector regression, partial least squares regression, or continuous projection algorithm combined with support vector regression.

[0033] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0034] 1. The method provided by this invention eliminates the need for steps such as drying the fermentation broth, which simplifies the steps for detecting the chemical properties of samples.

[0035] 2. The method provided by this invention can detect the organic carbon content and total nitrogen content of fermentation broth in situ with low damage.

[0036] 3. The method provided by this invention has a short detection time and can simultaneously obtain the organic carbon and total nitrogen content of the fermentation broth, as well as the quantity and area of ​​physical impurities in the fermentation broth. Attached Figure Description

[0037] Figure 1 This is a flowchart illustrating a method for simultaneously detecting the chemical composition content of fermentation broth and identifying physical impurities.

[0038] Figure 2 The spectrum of background and impurities in the anaerobic fermentation broth in Example 1;

[0039] Figure 3 The results of physical impurity identification in the fermentation broth in Example 1 (bright spots in the petri dish are suspended particulate impurities);

[0040] Figure 4 This is a comparison chart of the predicted and actual values ​​of the total organic carbon content in the fermentation broth samples collected in Example 1;

[0041] Figure 5 This is a comparison chart of the predicted and actual total nitrogen content of the fermentation broth samples collected in Example 1. Detailed Implementation

[0042] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0043] Any preparation methods, materials, structures, algorithm models, or composition ratios not explicitly described in this technical solution are considered common technical features disclosed in the prior art.

[0044] Example 1

[0045] like Figure 1 As shown in the figure, this embodiment proposes a method for simultaneously detecting the chemical composition content of fermentation broth and identifying physical impurities, including the following steps:

[0046] S1: Four bioreactors were set up. The initial organic loading rate of organic waste in the bioreactors was 1 kg-VS / L / d. Two parallel reactors, C-1 and C-2, fed only organic matter, with slurry from kitchen waste and catering waste as the organic matter feed. The other two parallel reactors, N-1 and N-2, added 50 mg-ammonium chloride per liter of effective reactor volume in addition to the 1 kg-VS / L / organic matter feed. All four bioreactors operated in a semi-continuous mode, with feed and discharge every one day. The reaction ran for 100 days, collecting 200 fermentation broth samples, each with a volume greater than 15 mL.

[0047] S2: Place the fermentation broth sample obtained in S1 into a glass petri dish with a diameter of 60 mm as the fermentation broth sample to be tested.

[0048] S3: Hyperspectral images of the fermentation broth samples from S2 were acquired using a near-infrared hyperspectral camera (model: FX 17, brand: Specim, country of manufacture: Finland). To prevent ambient light from interfering with the samples, the image acquisition system was covered by a black screen. The object distance was 18 cm, the exposure time was 3.2 ms, the frame rate was 55 Hz, the light source intensity was 2200 lux, and the conveyor belt speed was 15.6 mm / s to obtain near-infrared hyperspectral images for each sample.

[0049] S4: Process the hyperspectral image of the fermentation broth sample obtained in S3, including the following steps:

[0050] S4-1 uses a total organic carbon and total nitrogen analyzer (model: TOC-VCPH, brand: Shimadzu, country of manufacture: Japan) to test the total organic carbon and total nitrogen content in the sample of S3 after hyperspectral testing;

[0051] S4-2 selects a preset region of interest (ROI) in each of the near-infrared hyperspectral images (60mm in diameter) acquired in S3, for example, a 50×50 pixel region. The average spectrum of each pixel within the ROI is calculated and exported using ENVI software (version 5.3.1, company: Exelis Visual Information Solutions, Inc., country of manufacture: USA). Savitzky-Golay smoothing, standard normal variables (SNV), and multivariate scattering correction (MSC) are used to preprocess the average spectrum. Based on a random function, the data for each sample is divided into a training set and a test set, with the training set comprising 80% of the total data and the test set comprising 20%.

[0052] In S4-3, regions of interest (ROIs) for impurities and fermentation broth matrix are selected from the near-infrared hyperspectral images acquired in S3. Standardized spectral curves for these two ROIs are obtained, and a binary classification model is used to identify physical impurities. The detection accuracy is calculated based on the actual number of physical impurities and the detection results. The detection accuracy is the ratio of the number of impurities detected by the binary classification model to the actual number of impurities. The results are as follows: Figure 2 As shown.

[0053] S5: Based on the total organic carbon and total nitrogen content of the fermentation broth obtained in S4-1 of the training set and the average spectral information of the fermentation broth samples obtained in S4-2, establish a partial least squares regression (PLSR) prediction model for the total organic carbon and total nitrogen content of the fermentation broth, and use the test set to test the model accuracy, and further optimize the prediction model for the organic carbon and total nitrogen content of the fermentation broth.

[0054] S6: Collect anaerobic fermentation broth samples containing physical impurities. Near-infrared hyperspectral imaging (NIR) is used to acquire hyperspectral data of the samples. The matched filter (MF), spectral angle mapper (SAM), constrained energy minimization (CEM), and adaptive coherence estimator (ACE) binary classification model constructed in S4-3 are used to identify the impurities. The detection results for one sample are shown in Table 1. Specifically, the constrained energy minimization (CEM) algorithm yields the impurity target recognition image as shown below. Figure 3 As shown.

[0055] Table 1. Number of targets and total area identified by the binary classification model.

[0056]

[0057] S7: Collect the fermentation broth sample to be tested, acquire the hyperspectral data of the sample using near-infrared hyperspectral spectroscopy, and predict the organic carbon and total nitrogen content of the sample using the prediction model for fermentation broth organic carbon and total nitrogen content constructed in S5. The comparison results of the predicted value and the actual value of total organic carbon content are as follows: Figure 4 As shown, from Figure 4 As can be seen from the regression R, the total organic carbon content... 2 The predicted value can reach 0.824, with a root mean square error (RMSE) of 1187 mg-C / L. The comparison between the predicted and actual values ​​of total nitrogen content is as follows: Figure 5 As shown, from Figure 5 As can be seen from the regression R, the total nitrogen content... 2 It can reach 0.860, and the root mean square error (RMSE) is 706 mg-N / L.

[0058] The above description of the embodiments is provided to enable those skilled in the art to understand and use the invention. It will be apparent to those skilled in the art that various modifications can be made to these embodiments, and the general principles described herein can be applied to other embodiments without inventive effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the invention should be within the protection scope of the present invention.

Claims

1. A method for simultaneous detection of chemical content and identification of physical tramp in a fermentation broth, characterized in that, Includes the following steps: S1: Select a bioreactor for testing fermentation broth samples; S2: Collect fermentation broth samples from the bioreactor in S1; S3: Obtain the near-infrared hyperspectral image of the fermentation broth sample to be tested in S2; S4: Process the near-infrared hyperspectral image of the fermentation broth sample obtained in S3, including the following steps: S4-1 uses a total organic carbon and total nitrogen analyzer to test the total organic carbon and total nitrogen content in the fermentation broth sample of S3 after near-infrared hyperspectral testing; S4-2 takes the near-infrared hyperspectral images acquired in S3, selects a region of a preset size in each image as the region of interest, calculates and derives the average spectrum in the region of interest, preprocesses the average spectrum, and divides the processed spectral data into a training set and a test set. S4-3 selects regions of interest for physical impurities and fermentation broth matrix on the near-infrared hyperspectral image acquired in S3, obtains standardized spectral curves for these two types of regions of interest, and uses a binary classification model to identify physical impurities. The detection accuracy is calculated based on the actual number of physical impurities and the detection results. S5: Based on the total organic carbon and total nitrogen content of the fermentation broth samples obtained from S4-1 in the training set, and the average spectral information of the fermentation broth samples obtained from S4-2, establish a regression prediction model for the total organic carbon and total nitrogen content of the fermentation broth samples, and use the test set to detect the model accuracy, and further optimize the prediction model for the total organic carbon and total nitrogen content of the fermentation broth. S6: Collect anaerobic fermentation broth samples containing physical impurities, obtain hyperspectral data of the fermentation broth samples using near-infrared hyperspectral technology, and identify the quantity and area of ​​impurities using the binary classification model constructed in S4-3; S7: Collect the fermentation broth sample to be tested, obtain the hyperspectral data of the fermentation broth sample using near-infrared hyperspectral technology, and predict the total organic carbon and total nitrogen content of the fermentation broth sample using the prediction model for total organic carbon and total nitrogen content constructed in S5. In S4-3, the binary classification model is obtained through one of the following methods: matched filter, spectral angle mapper, constrained energy minimization, and adaptive consistency estimator. In S5, the regression prediction model for total organic carbon and total nitrogen in the fermentation broth is obtained by one of the following methods: random forest regression, support vector regression, partial least squares regression, or continuous projection algorithm combined with support vector regression.

2. The method for simultaneous detection of chemical components content and identification of physical trashes in fermentation broth according to claim 1, characterized in that, In S1, four bioreactors are set up. The initial organic load of organic waste fed into the bioreactors is 1~3 kg-VS / L / d. The bioreactors feed and discharge materials every 1~2 days and collect fermentation broth samples. Two of the four bioreactors, C-1 and C-2, are fed only with organic matter; two parallel reactors, N-1 and N-2, are fed with an additional 50-100 mg of ammonium chloride per liter of effective reactor volume. All four bioreactors operate in a semi-continuous mode.

3. The method for simultaneous detection of chemical components content and identification of physical trashes in fermentation broth according to claim 2, characterized in that, In S1, the organic waste feed includes one or more of the following: kitchen waste, catering waste, sewage sludge, agricultural and forestry waste, and poultry and livestock manure; The number of fermentation broth samples collected was greater than 100.

4. The method for simultaneously detecting the chemical composition content and identifying physical impurities in fermentation broth according to claim 2, characterized in that, In S1, the total organic carbon content of the collected fermentation broth sample ranges from 1253 mg-C / L to 12507 mg-C / L.

5. The method for simultaneously detecting the chemical composition content and identifying physical impurities in fermentation broth according to claim 2, characterized in that, In S1, the total nitrogen content of the collected fermentation broth sample ranges from 2237 mg-N / L to 12168 mg-N / L.

6. According to claim 2, in the method for simultaneously detecting the chemical composition content of fermentation broth and identifying physical impurities, in S2, a fermentation broth sample of more than 15 mL is taken from the four bioreactors in S1 and placed in a glass petri dish as the fermentation broth sample to be tested. In S3, a near-infrared hyperspectral camera is used to acquire near-infrared hyperspectral images of the fermentation broth sample to be tested in S2. By adjusting the object distance, exposure time, frame rate range, and light source intensity, a clear near-infrared hyperspectral image of the fermentation broth sample to be tested is obtained.

7. The method for simultaneously detecting the chemical composition content and identifying physical impurities in fermentation broth according to claim 1, characterized in that, In S4-2, the method for dividing the training set and the test set is to divide the data of each sample into a training set and a test set based on a Random function; The training set comprises 60% to 80% of all sample data, and the test set comprises 20% to 40% of all sample data.

8. The method for simultaneously detecting the chemical composition content and identifying physical impurities in fermentation broth according to claim 1, characterized in that, In S4-3, the physical impurities include wood chips, plastic sheets, suspended particles and foam formed during fermentation that are mixed in with the feed.

Images