Raman spectrum identification method, device and system for drug residues in water body based on neural network algorithm

Abstract

The invention relates to a Raman spectrum identification method, a Raman spectrum identification device and a Raman spectrum identification system for drug residues in a water body based on a neural network algorithm, and relates to the technical field of Raman spectrum analysis. The method comprises the following steps: performing Raman spectrum detection analysis on a to-be-detected target solution containing at least one target by using a Raman spectrum enhanced detection substrate to obtain Raman spectrum data, then performing target type analysis on the Raman spectrum data through a neural network algorithm model to obtain key features of the target, and further determining the target type of the target on the basis of the key features. And performing category distinguishing in combination with a preset key feature sample to obtain a target object detection and recognition result. According to the method, the Raman enhanced spectrum technology is used for detecting the metal enhanced substrate, and the target object in the actual water sample is analyzed and recognized in combination with the model, so that the detection sensitivity is improved, the problem that the mixed target object in the water environment is difficult to recognize in the prior art is solved, and a stable and reliable scheme is provided for recognizing the water body target object.

Description

Technical Field

[0001] The present application relates to the technical field of Raman spectroscopy analysis, and in particular to a method, device and system for Raman spectroscopy identification of drug residues in water based on a neural network algorithm. Background Art

[0002] Raman spectroscopy is a non-destructive and rapid detection technology that can obtain fingerprint information of target molecules. It has the characteristics of specific identification and has been applied in chemical analysis, biomedical testing and other fields.

[0003] Existing target object detection methods based on Raman spectroscopy detection technology are mainly based on quantitative detection of known objects. In terms of Raman spectroscopy data processing, existing methods are prone to misjudgment when processing spectral data of complex samples and mixtures due to lack of sufficient flexibility, and there is a problem of difficulty in identifying multiple pollutants in water. Summary of the invention

[0004] The present application provides a method, device and system for identifying drug residues in water by Raman spectroscopy based on a neural network algorithm. On the one hand, the present application utilizes surface enhanced Raman spectroscopy (SERS) technology to significantly enhance the Raman signal of target molecules adsorbed on the surface by means of the plasma resonance enhancement effect of precious metal nanomaterials, thereby improving the stability of the detection substrate and thus improving its detection sensitivity; on the other hand, the present application utilizes a neural network algorithm model to perform feature extraction and analysis and category distinction on the detected Raman spectrum to achieve the identification of unknown substances. For the case of multiple mixtures, a neural network algorithm model capable of identifying multiple target objects is used for mixture detection in the environment, thereby achieving the identification of multiple pollutants in aquaculture water.

[0005] In a first aspect, the present application provides a method for Raman spectroscopy identification of drug residues in water based on a neural network algorithm, comprising:

[0006] Using Raman spectroscopy technology, using the prepared Raman spectroscopy enhanced detection substrate to perform Raman spectroscopy detection and analysis on a target solution to be detected, to obtain Raman spectroscopy data, wherein the Raman spectroscopy data includes a Raman spectrum of at least one target to be detected and identified;

[0007] By using a neural network algorithm model, the Raman spectroscopy data is analyzed for target species to obtain key features of the target;

[0008] Based on the key features, combined with preset key feature samples, category differentiation is performed to obtain target object detection and recognition results;

[0009] Among them, the target solution to be detected is a target water solution detected by a metal substrate, the metal substrate is an aluminum foil SERS substrate, and the target solution to be detected contains at least one target to be detected and identified among levofloxacin, fleroxacin, pefloxacin, sulfadiazine, methylene blue and malachite green.

[0010] Optionally, Raman spectroscopy is used to perform Raman spectroscopy detection and analysis on the target solution using the prepared Raman spectroscopy enhanced detection substrate to obtain Raman spectroscopy data, including:

[0011] By using Raman spectroscopy technology, raw Raman spectroscopy data is collected from a target solution to be tested using a Raman spectroscopy enhanced detection substrate, wherein the raw Raman spectroscopy data includes a single target Raman spectrum or a mixed target Raman spectrum;

[0012] Performing preprocessing based on the original Raman spectrum data to obtain Raman spectrum data;

[0013] The aluminum foil SERS substrate in the Raman enhanced detection substrate is obtained by a Raman enhanced detection substrate preparation method, in which a metal foil is reacted with a silver nitrate solution containing hydrofluoric acid to form a nanosilver structure on the metal foil.

[0014] Optionally, raw Raman spectroscopy data is collected from a solution of a target object to be detected by using Raman spectroscopy technology and a Raman spectroscopy enhanced detection substrate, including:

[0015] Under the target temperature condition, the target solution to be tested is detected by Raman spectroscopy detection method using a Raman spectroscopy enhanced detection substrate to obtain Raman spectroscopy characteristic peak information at the solution concentration corresponding to the target as the original Raman spectroscopy data;

[0016] The Raman spectrum characteristic peak information includes Raman spectrum characteristic peak intensity and / or peak area intensity.

[0017] Optionally, the Raman spectral data is analyzed for target type through a neural network algorithm model to obtain key features of the target to be detected and identified, including:

[0018] Inputting the Raman spectrum data into a neural network algorithm model to classify targets to obtain a first classification result, wherein the first classification result includes a single target classification result or a multi-target classification result;

[0019] For the first classification result, an analyzability analysis is performed using the neural network algorithm model, and a key feature of at least one target object corresponding to the identified Raman spectrum data is output.

[0020] Optionally, performing an analyzability analysis on the first classification result by using the neural network algorithm model includes:

[0021] Based on the first classification result, selecting a target neuron for identification and analysis from the neural network algorithm model;

[0022] The Raman spectrum data is subjected to feature analysis by the target neuron to obtain key features.

[0023] Optionally, before performing target species analysis on the Raman spectroscopy data using a neural network algorithm model, the method further includes:

[0024] By using Raman spectroscopy technology, the sample solution is detected and analyzed using the prepared Raman spectroscopy enhancement substrate to obtain a Raman spectroscopy sample set, wherein the Raman spectroscopy sample set includes a single target spectroscopy sample and a mixed target spectroscopy sample;

[0025] The Raman spectrum sample set is sorted and divided by a preset random function to obtain a training set, a validation set and a test set;

[0026] According to the composition characteristics of the Raman spectrum sample set, a multi-layer perceptron algorithm model based on a neural network is established;

[0027] The multi-layer perceptron algorithm model is trained using the training set, and is verified using the validation set during the model training process, and the model is tested using the test set until the multi-layer perceptron algorithm model training is completed, thereby obtaining a performance evaluation result of the model and a trained neural network algorithm model;

[0028] Based on the performance evaluation results, the model is analyzed using the SHAP analysis method to obtain key feature samples of each category of target objects;

[0029] The sample solution contains at least one target solution sample of concentration, and the key characteristic samples are concentrated at 447-450cm -1 、1104-1114cm -1 and 571cm -1 The key feature sample is within the Raman shift range, and the key feature sample coincides with the characteristic peak of the target object, so as to distinguish the spectra of different categories of targets.

[0030] Optionally, according to the composition characteristics of the Raman spectrum sample set, a multi-layer perceptron algorithm model based on a neural network is established, including:

[0031] Performing spectral feature number analysis according to the Raman spectrum sample set to obtain spectral feature numbers;

[0032] A multi-layer perceptron algorithm model based on a neural network is established, and the number of input layer neurons of the multi-layer perceptron algorithm model is set to correspond to the number of spectral features, and the output layer neurons of the multi-layer perceptron algorithm model are set to correspond to the target object recognition category.

[0033] Optionally, the sample solution is detected and analyzed using the prepared Raman spectrum enhancement substrate by Raman spectroscopy technology to obtain a Raman spectrum sample set, including:

[0034] Under room temperature conditions, Raman spectroscopy technology is used to collect Raman spectroscopy data of sample solutions of different concentrations in the sample using a Raman spectroscopy enhancement substrate, and a working curve of the relationship between the target object at different concentrations and its Raman spectroscopy characteristic peak intensity or peak area intensity is obtained as a Raman spectroscopy sample set, and a Raman enhanced detection method for the target object is established;

[0035] In the substrate preparation method of the Raman spectroscopy enhanced detection substrate, the reaction solution is prepared by reacting hydrofluoric acid of a preset concentration and a silver nitrate solution of at least one concentration according to the detection requirements of the target object.

[0036] In a second aspect, the present application provides a device for identifying Raman spectra of drug residues in water based on a neural network algorithm, comprising:

[0037] A Raman spectrum detection and analysis module, used to perform Raman spectrum detection and analysis on a solution of a target object to be detected through the prepared Raman spectrum enhancement substrate to obtain Raman spectrum data, wherein the Raman spectrum data includes a Raman spectrum of at least one target object to be detected and identified;

[0038] An analysis module, used to analyze the target object type of the Raman spectrum data through a neural network algorithm model to obtain key features of the target object;

[0039] The category differentiation module is used to perform category differentiation based on the key features in combination with preset key feature samples to obtain target detection and identification results; wherein the metal substrate solution is an aluminum foil SERS substrate, and the target water solution contains at least one target to be detected and identified among levofloxacin, fleroxacin, pefloxacin, sulfadiazine, methylene blue and malachite green.

[0040] In a third aspect, the present application provides a Raman spectroscopy identification system for drug residues in water based on a neural network algorithm, comprising:

[0041] A Raman spectrum detection device, used for performing Raman spectrum detection and analysis on a target solution to be detected through the prepared Raman spectrum enhancement substrate to obtain Raman spectrum data, wherein the Raman spectrum data includes a Raman spectrum of at least one target to be detected and identified;

[0042] The detection and analysis equipment is connected to the Raman spectrum detection equipment, and is used to analyze the target type of the Raman spectrum data through a neural network algorithm model to obtain the key features of the target; based on the key features, the category is distinguished in combination with preset key feature samples to obtain the target detection and identification results; wherein the metal substrate solution is an aluminum foil SERS substrate, and the target water solution contains at least one target to be detected and identified among levofloxacin, fleroxacin, pefloxacin, sulfadiazine, methylene blue and malachite green.

[0043] In summary, the embodiments of the present application use Raman spectroscopy technology and a Raman enhanced substrate to perform Raman spectroscopy detection and analysis on a target solution to be detected containing a single target and multiple targets, obtain Raman spectroscopy data, and then use a neural network algorithm model to perform target type analysis on the Raman spectroscopy data to obtain the key features of the target, and then based on the key features, combined with preset key feature samples, classify the categories to obtain target detection and identification results. The present application uses Raman enhanced spectroscopy technology to detect metal enhanced substrates, and combines model analysis to identify targets in actual water samples, thereby improving detection sensitivity, thereby solving the problem that the prior art is difficult to identify mixed targets in a water environment, and proposes a stable and reliable solution for the identification of water targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0046] Figure 1 A schematic diagram of a process for identifying drug residues in water by Raman spectroscopy based on a neural network algorithm provided in an embodiment of the present application;

[0047] Figure 2 This is a schematic diagram of the steps of a method for Raman spectroscopy identification of drug residues in water based on a neural network algorithm provided by an optional embodiment of the present application;

[0048] Figure 3 An optional example of the present application provides a Raman spectrum of a metal substrate and a target on a different metal substrate;

[0049] Figure 4 This is a schematic diagram of a multi-layer perceptron model structure provided by an optional example of this application;

[0050] Figure 5 This is a schematic diagram of the accuracy and loss rate of a multi-layer perceptron model training set and a validation set provided by an optional example of the present application;

[0051] Figure 6 This is a schematic diagram of a SHAP value analysis of the effect of the Raman shift characteristics of a target object on the model output provided by an optional example of the present application;

[0052] Figure 7 A structural block diagram of a device for identifying Raman spectroscopy of drug residues in water based on a neural network algorithm provided in an embodiment of the present application;

[0053] Figure 8 It is a structural schematic diagram of a Raman spectroscopy identification system for drug residues in water based on a neural network algorithm provided in an embodiment of the present application. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0055] To facilitate the understanding of the embodiments of the present application, further explanation will be given below in conjunction with the drawings and specific embodiments. The embodiments do not constitute a limitation on the embodiments of the present application.

[0056] Figure 1 A schematic diagram of a process for identifying Raman spectra of drug residues in water based on a neural network algorithm provided in an embodiment of the present application. Figure 1 As shown, the Raman spectroscopy identification method for drug residues in water based on a neural network algorithm provided in the embodiment of the present application may specifically include the following steps:

[0057] Step 110 , using Raman spectroscopy technology, the prepared Raman spectroscopy enhanced detection substrate is used to perform Raman spectroscopy detection and analysis on the target solution to obtain Raman spectroscopy data.

[0058] Wherein, the Raman spectral data includes the Raman spectrum of at least one target to be detected and identified, the target solution to be detected is an aqueous solution of the target detected by a metal substrate, the metal substrate is an aluminum foil SERS substrate, and the target solution to be detected includes at least one target to be detected and identified among levofloxacin, fleroxacin, pefloxacin, sulfadiazine, methylene blue and malachite green.

[0059] Specifically, the target is usually residual drugs in water, and the target water solution usually contains at least one target, such as a single target (i.e., only one residual drug to be detected and identified), or a mixture of multiple targets (e.g., containing at least two residual drugs to be detected and identified). This embodiment does not limit the types of targets contained in the target water solution. In this embodiment, the metal substrate solution is used as the prepared Raman spectroscopy enhanced detection substrate (or Raman spectroscopy enhanced substrate), which can have a variety of preparation methods. It is mainly prepared by reacting a metal and a silver nitrate solution containing hydrofluoric acid, such as an aluminum foil SERS substrate (also called an aluminum foil SERS detection substrate) is mainly prepared by reacting an aluminum foil and a silver nitrate solution containing hydrofluoric acid. The aluminum foil SERS detection substrate can be used for the detection of drug residues in water. This embodiment uses this substrate to establish a Raman enhanced detection method for multiple drug residues in water to achieve quantitative analysis, and is used for the identification and detection of unknown targets.

[0060] It should be noted that, in addition to being prepared with aluminum foil, the Raman spectrum enhancement substrate in this embodiment can also be prepared with one of copper or zinc sheets in combination with a silver nitrate solution containing hydrofluoric acid. This embodiment selects cheap metal materials and prepares the SERS substrate by chemical replacement self-assembly. The chemical replacement method quickly generates uniform silver nanostructures with Raman enhancement effects on the metal foil, thereby realizing a fast and low-cost detection substrate preparation method, which is an effective way to solve the stability of the existing detection substrate.

[0061] In a specific implementation, this embodiment can use a Raman spectrometer or other Raman spectrum detection equipment to detect and analyze the Raman spectrum of the solution, which mainly uses the prepared Raman spectrum enhancement substrate to detect and analyze the Raman spectrum of the solution of the target object to be detected. By detecting the Raman signal of the target object, the Raman spectrum of the target object is collected as the Raman spectrum of the target object.

[0062] In actual implementation, since the Raman signal of the target object is enhanced on the aluminum foil SERS substrate, when the detection and analysis is performed through the Raman spectrum enhancement detection substrate, the collected Raman spectrum is actually the Raman spectrum of the target object obtained on the aluminum foil SERS substrate. For the case of only one target object, when the detection and analysis is performed through the Raman enhanced spectroscopy technology, the collected Raman spectrum is the spectrum of the target object obtained on the aluminum foil SERS substrate; for the case of mixed targets, when the detection and analysis is performed through the Raman enhanced spectroscopy technology, the Raman spectra corresponding to multiple targets can be collected on the aluminum foil SERS substrate.

[0063] In the relevant technology, methylene blue, malachite green and various antibiotics in water bodies are commonly used disinfectants in aquaculture. However, these substances have certain toxicity and potential hazards. Residual drugs in water bodies can enter the food chain through aquatic products, thereby endangering human health. At the same time, drug residues in water bodies enter the environment through tailwater discharge, which can cause harm to the ecological environment. At present, the main detection methods for these target objects in water are high performance liquid chromatography (HPLC) and liquid chromatography mass spectrometry (LC / MS), but these instruments are complicated to operate and have high detection costs. They generally require complex sample pretreatment processes, which limits their detection applications. In addition, existing laboratory detection methods are mainly based on quantitative detection of known objects, and screening of unknown objects requires higher requirements for instruments and personnel. In addition, there is also a solution for using traditional Raman spectroscopy detection technology for target object detection.

[0064] Compared with the existing target detection technology, this embodiment uses Raman spectroscopy enhanced detection technology, namely surface enhanced Raman spectroscopy (SERS) technology. The SERS technology uses the plasma resonance enhancement effect of metal nanomaterials to significantly enhance the Raman signal of the target molecules adsorbed on the surface, thereby improving the detection sensitivity. This embodiment uses surface enhanced Raman spectroscopy technology to achieve quantitative analysis of known objects, and at the same time, combined with data processing algorithms and recognition models, it can achieve recognition of unknown objects.

[0065] It should be noted that in the present application, Raman spectroscopy is a non-destructive and rapid detection technology that can obtain fingerprint information of target molecules and has the characteristics of specific identification. It has been applied in the fields of chemical analysis, biomedical testing, etc.

[0066] Optionally, the above-mentioned Raman spectroscopy technology is used to perform Raman spectroscopy detection and analysis on the target solution to be detected using the prepared Raman spectroscopy enhanced detection substrate to obtain Raman spectroscopy data, which may include the following sub-steps:

[0067] Sub-step 1101 , using Raman spectroscopy technology and a Raman spectroscopy enhanced detection substrate to collect raw Raman spectroscopy data from a target solution to be detected.

[0068] The original Raman spectrum data includes a single target Raman spectrum or a mixed target Raman spectrum.

[0069] Sub-step 1102: pre-processing is performed based on the original Raman spectrum data to obtain Raman spectrum data.

[0070] The aluminum foil SERS substrate in the Raman enhanced detection substrate is obtained by a Raman enhanced detection substrate preparation method, in which a metal foil is reacted with a silver nitrate solution containing hydrofluoric acid to form a nanosilver structure on the metal foil.

[0071] Sub-steps 1101 and 1102 are described uniformly:

[0072] In a specific implementation, after collecting raw Raman spectral data from a target solution to be detected through a Raman spectroscopy enhancement substrate, this embodiment can perform a series of preprocessing on the raw Raman spectral data, and the preprocessing process includes but is not limited to: processing using displacement clipping, airpls baseline processing, whittaker smoothing filtering, and normalization methods to obtain Raman spectral data.

[0073] In actual implementation, this embodiment can collect Raman spectra under certain temperature conditions, such as room temperature / normal temperature conditions.

[0074] In an optional embodiment, this embodiment uses Raman spectroscopy technology to collect original Raman spectroscopy data from a target solution to be tested using a Raman spectroscopy enhanced detection substrate, which may include: under target temperature conditions, using a Raman spectroscopy detection method and a Raman spectroscopy enhanced detection substrate to detect the target solution to be tested, and obtaining Raman spectroscopy characteristic peak information at a solution concentration corresponding to the target as the original Raman spectroscopy data; wherein the Raman spectroscopy characteristic peak information includes Raman spectroscopy characteristic peak intensity and / or peak area intensity.

[0075] Preferably, the target temperature is room temperature / normal temperature, corresponding to 25 degrees Celsius, i.e. 25°C.

[0076] For example, in this embodiment, the Raman spectrum of the target object can be collected at 25° C. using a Raman detection system equipped with a 785 nm laser. The Raman laser power is 500 mW, and the Raman signal is collected twice, each time for 5 seconds, to obtain the Raman spectrum corresponding to the target object.

[0077] Step 120: Analyze the target object type on the Raman spectrum data through a neural network algorithm model to obtain key features of the target object.

[0078] In this embodiment, the key features mainly refer to the features of the Raman spectrum of the target object, wherein the key features of a single target object and the key features of a mixed target object may be different. The key features may be mainly concentrated in the range of 447-450cm -1 、1104-1114cm -1 and 571cm -1 within the Raman shift range.

[0079] In a specific implementation, this embodiment can pre-train a neural network algorithm model, then input Raman spectrum data into the neural network algorithm model, and use the model to identify and extract features from the Raman spectrum data. This embodiment can use the SHAP model interpretation tool to perform an analyzability analysis on the model, and use SHAP to identify key features of different categories of spectra from Raman spectrum data.

[0080] Optionally, this embodiment uses a neural network algorithm model to analyze the target type of the Raman spectrum data to obtain the key features of the target, which may include the following sub-steps:

[0081] Sub-step 1201: input the Raman spectrum data into a neural network algorithm model to classify the target object and obtain a first classification result.

[0082] Wherein, the first classification result includes a single target classification result or a multi-target classification result;

[0083] Sub-step 1202, performing an analyzability analysis on the first classification result through the neural network algorithm model, and outputting a key feature of at least one target object corresponding to the identified Raman spectrum data.

[0084] Sub-steps 1201 and 1202 are described uniformly:

[0085] In a specific implementation, the neural network algorithm model can identify the number of target types contained in the target solution to be detected from the input Raman spectrum data, and the number of types is used as the classification result to distinguish whether the target water solution contains a single target or a mixed target. Among them, the multi-target classification result can also be subdivided into two mixed target classification results and more than two mixed target classification results, which are not limited in this embodiment. By identifying the number of target types, the model can perform further analytical analysis on the number of types to obtain the key features of the target.

[0086] Optionally, this embodiment performs a parsability analysis on the first classification result through the neural network algorithm model, which may specifically include: selecting a target neuron for identification and analysis from the neural network algorithm model for the first classification result; performing feature analysis on the Raman spectral data through the target neuron to obtain key features.

[0087] In the prior art, in the processing of Raman spectral data, the existing hit quality index (HQI) similarity is used to evaluate the similarity of spectral data. However, the HQI method lacks sufficient flexibility and robustness when processing spectral data of complex samples and mixtures, which leads to misjudgment. Therefore, its application in Raman spectral recognition is subject to certain limitations. The environment (such as aquaculture) usually contains a mixture of multiple antibiotics. To detect these antibiotic mixtures, it is necessary to train a model that can recognize multiple targets for the detection of mixtures in the environment.

[0088] To solve the above problems, the neural network algorithm model in this embodiment may include multiple output layer neurons, and there may be a certain correspondence between the output layer neurons and the target objects. Preferably, the output layer may have 7 neurons corresponding to 7 categories, and the 7 categories of neurons are mainly divided into neurons for single target recognition and neurons for mixed target recognition (such as neurons for recognition of two mixed targets, and neurons for recognition of more than two mixed targets). In this embodiment, for different categories of targets, corresponding neurons can be selected for recognition analysis to distinguish the spectra of different categories of targets.

[0089] Exemplarily, in the above neural network algorithm model, the 7 categories of neurons can be divided as follows: 3 neurons correspond to single target recognition and classification, 3 neurons correspond to recognition and classification in the case of a mixture of two targets, and 1 neuron corresponds to recognition and classification in the case of a mixture of three or more targets.

[0090] Step 130, based on the key features, combined with preset key feature samples, classify the categories to obtain target object detection and recognition results.

[0091] In this embodiment, the key characteristic samples are characteristic samples established by analyzing the sample solution of the target object in advance. The key characteristic samples are mainly concentrated in 447-450cm -1 、1104-1114cm -1 and 571cm -1 The present embodiment uses key features and key feature samples for comparative analysis to determine the target detection and identification results. In the present embodiment, each feature sample in the key feature sample matches the characteristic peak of the target and plays a key role in distinguishing the spectra of different categories of targets.

[0092] It can be seen that the embodiment of the present application uses Raman spectroscopy technology and Raman spectroscopy enhanced detection substrate to perform Raman spectroscopy detection and analysis on a target solution to be detected containing a single or multiple targets, obtains Raman spectroscopy data, and then uses a neural network algorithm model to analyze the target type of the Raman spectroscopy data to obtain the key features of the target, and then based on the key features, combined with the preset key feature samples, the category is distinguished to obtain the target detection and identification results. On the one hand, the present application uses Raman enhanced spectroscopy technology to detect metal enhanced substrates to solve the problem that the current SERS substrate is complex to make and it is difficult to identify multiple pollutants in water. On the other hand, this embodiment proposes a SERS identification method based on a deep learning algorithm. On the basis of the collected Raman spectrum, the algorithm model is combined to analyze and identify the target in the actual water sample. While realizing the identification of multiple pollutants in the water body, the detection sensitivity is further improved, thereby solving the problem that the existing technology is difficult to identify mixed targets in the water environment, and a stable and reliable solution is proposed for the identification of water targets.

[0093] Reference Figure 2 , showing a flowchart of the steps of a method for Raman spectroscopy identification of drug residues in water based on a neural network algorithm provided by an optional embodiment of the present application. Specifically, the method for Raman spectroscopy identification of drug residues in water based on a neural network algorithm provided in an embodiment of the present application can first train the neural network algorithm model before performing target object type analysis on Raman spectroscopy data through a neural network algorithm model, and establish key feature samples for different target objects. In the model training stage, the following steps can be specifically included:

[0094] Step 210 , using the prepared Raman spectrum enhancement substrate to detect and analyze the sample solution through Raman spectrum technology to obtain a Raman spectrum sample set.

[0095] Wherein, the sample solution comprises a target solution sample of at least one concentration, and the Raman spectrum sample set comprises a single target spectrum sample and a mixed target spectrum sample;

[0096] In this embodiment, the prepared enhanced detection substrate is composed of an aluminum foil SERS substrate. The scheme of this embodiment is a Raman detection method for drug residues in water based on the detection substrate. The sample solution mainly includes single target solution and mixed target solution of different concentrations. For example, the concentration of 1×10 -4 ~1×10 -8mol / L of drug solutions of different concentrations, and placing the Raman enhanced detection substrate in drug solutions of different concentrations to form multiple substrate solution samples. Among them, the target objects targeted by the Raman detection method for drug residues in water mainly include but are not limited to: one or more of levofloxacin, fleroxacin, pefloxacin, sulfadiazine, methylene blue and malachite green.

[0097] Compared with the unknown substance method based on metal substrate used in the prior art, the metal substrate used in this embodiment is a Raman spectroscopy enhanced substrate. In view of the difficulties in preparing the metal substrate in the prior art, the substrate preparation method proposed in this embodiment has the characteristics of simple operation, fast preparation speed, and scalable preparation, which solves the problems of the complex and time-consuming preparation process of the existing detection substrate, and solves the problem that the current machine learning model has difficulty in identifying mixed targets in the water environment. The specific method is as follows:

[0098] The preparation method of the aluminum foil SERS detection substrate comprises the following steps: placing an aluminum foil of a certain size in a silver nitrate solution containing hydrofluoric acid and soaking it for 2 to 20 seconds (s). Preferably, the concentration of silver nitrate is 0.005 to 0.4 mol / L (mole per liter), and the concentration of hydrofluoric acid is 0.1 mol / L. After the aluminum foil turns into a yellow-brown color with a metallic luster, take it out, rinse it with deionized water to remove the hydrofluoric acid remaining on the surface, and then blow it dry with air to complete the preparation of the aluminum foil SERS substrate.

[0099] In the related art, the prior art also uses a method of using silver nanosol for detection when detecting unknown substances. Compared with the prior method, this embodiment uses hydrofluoric acid to remove the oxide film on the metal foil, and generates a uniform silver nanostructure with Raman enhancement effect through a substitution reaction between a metal element and silver nitrate. The preparation cost of the detection substrate is low and the preparation time is short. The metal enhanced substrate preparation method provided in this embodiment forms a fixed "hot spot", which makes the reproducibility of Raman enhanced detection better. Different concentrations of fleroxacin, pefloxacin and levofloxacin were added to the actual water sample for a spike recovery experiment. The results show that the spike recovery of the water sample of the substrate is between 97.8-110.4%, and RSD<6.73%.

[0100] In actual processing, for the use of aluminum foil SERS detection substrate for the detection of drug residues in water, this embodiment uses fleroxacin, levofloxacin, pefloxacin, sulfadiazine, methylene blue and malachite green as typical targets, and uses target solutions of different concentrations in combination with the aluminum foil SERS detection substrate to form a solution sample for the detection substrate of drug residues in water.

[0101] For example, fleroxacin, levofloxacin, pefloxacin, sulfadiazine, methylene blue and malachite green were used as typical targets, and the concentrations were 1×10 -4 ~1×10 -8 mol / L target solution, add the test solution to the detection substrate, and detect it with a Raman spectrometer at room temperature after the solution evaporates. The laser wavelength of the Raman spectrometer used is 785nm (nanometer), the laser power is 50-500mW (milliwatt), and the sample collection time is 3-5 seconds. Through instrument collection, the working curve of the relationship between the Raman spectrum characteristic peak intensity or peak area corresponding to the target at different concentrations is obtained, thereby establishing a Raman enhanced detection method for the target. For the above-mentioned target, the minimum detection limit of this method is 10.62-23μg / L (micrograms per liter), and the linear range is 4 orders of magnitude.

[0102] In practical applications, the stability and reliability of the substrate limit the widespread application of SERS technology. The SERS substrate is the core of SERS technology. A good SERS substrate should have the advantages of simple preparation method, short preparation time, low cost, good reproducibility and high sensitivity.

[0103] The following is an explanation of the preparation process of the relevant substrate solution samples involved in the preparation of the aluminum foil SERS substrate and the rapid detection of targets in water:

[0104] ① For the preparation of aluminum foil SERS substrate and rapid detection of methylene blue residue in water, in the preparation stage of sample solution, exemplary, it can be achieved by the following contents:

[0105] First, prepare the Raman enhanced detection substrate. The process is as follows: prepare 40mL (milliliters) of 5mmol / L (millimoles per liter) silver nitrate solution, add 80μL (microliters) of hydrofluoric acid solution and stir well. Soak the cut aluminum foil in the above solution for 10 seconds, take it out when the color of the aluminum foil changes from silver to yellow-brown, rinse the residual hydrofluoric acid and silver nitrate on the surface with deionized water, and dry it for use. The concentration is 1×10 -8 ~1×10 -4 mol / L (3.199μg / L~31.99mg / L) methylene blue standard solution; during the detection process, the aluminum foil SERS substrate was immersed in the target solution of different concentrations for 10 minutes. At room temperature (25℃), the Raman spectrum of methylene blue was collected using a Raman detection system equipped with a 785nm laser. The Raman laser power was 500mW, and the Raman signal was collected twice, with each collection time of 5s. The characteristic peak 446cm -1 For example, its Raman signal shows a significant enhancement (such as Figure 3As shown). Silver nitrate solutions with concentrations of 0.005 mol / L, 0.01 mol / L, 0.02 mol / L, 0.03 mol / L, and 0.04 mol / L were used to obtain aluminum foil SERS substrates. The obtained substrates were used to detect 1×10 -4 mol / L methylene blue, 0.02mol / L silver nitrate solution made of substrate has the best effect on methylene blue enhancement, the characteristic peak of methylene blue is 446cm -1 The enhancement factor is 4.2×10 5 ,like Figure 3 As shown. Among them, Figure 3 In the figure, MB represents methylene blue, AlF represents aluminum foil substrate, and AlF SERS represents aluminum foil SERS substrate.

[0106] ②Preparing an aluminum foil SERS substrate and detecting fleroxacin in water, in the sample solution preparation stage, exemplary, can be achieved by the following contents:

[0107] The experimental conditions mentioned in ① above are as follows: the concentration of silver nitrate is 40mmol / L, the reaction time of silver nitrate and aluminum foil is 2-10s to obtain the aluminum foil SERS substrate, and the substrate is used to detect 1×10 -4 mol / L of fleroxacin, and the Raman spectrum obtained, the recovery rate of fleroxacin on the substrate ranged from 91.74-116.34%, and the RSD ranged from 3.36-4.48%.

[0108] ③ Rapid identification of sulfadiazine, methylene blue, malachite green and their mixtures can be achieved by, for example, the following during the preparation of the sample solution:

[0109] Based on the experimental conditions mentioned in ① above, this method was used to establish a Raman spectral data set for sulfadiazine, methylene blue, malachite green and their mixtures. The process is as follows: place the aluminum foil in a solution containing 0.02 mol / L silver nitrate and 0.1 mol / L hydrofluoric acid for 10 seconds, take it out after the color of the aluminum foil changes from silvery white to yellowish brown, rinse the hydrofluoric acid on the surface with deionized water and blow it dry with air to prepare the aluminum foil SERS substrate. The concentration of sulfadiazine, methylene blue and malachite green was prepared to be 1×10 -4 ~1×10 -8 mol / L single and multiple mixed target solutions, and immerse the aluminum foil SERS substrate in the target to be tested.

[0110] Thus, in this embodiment, a uniform silver nanostructure with Raman enhancement effect is quickly generated on a metal foil by chemical replacement method. The substrate preparation method is simple and achieves high sensitivity, high reproducibility, low cost and large-scale preparation.

[0111] After the preparation of the substrate solution sample is completed in this embodiment, a Raman spectrometer can be used to detect at room temperature to establish a Raman spectrum data set of the sample, and then displacement clipping, airpls baseline processing, whittaker smoothing filtering and normalization methods are used to process it to obtain a Raman spectrum sample set.

[0112] In an optional embodiment, this embodiment uses Raman spectroscopy technology to detect and analyze the sample solution using the prepared Raman spectroscopy enhancement substrate to obtain a Raman spectroscopy sample set, including: at room temperature, using Raman spectroscopy technology, using the Raman spectroscopy enhancement substrate to collect Raman spectroscopy data of sample solutions of different concentrations in the sample, obtaining a working curve of the relationship between the target object and its Raman spectroscopy characteristic peak intensity or peak area intensity at different concentrations as the Raman spectroscopy sample set, and establishing a Raman enhanced detection method for the target object; wherein, in the substrate preparation method of the Raman spectroscopy enhanced detection substrate, the reaction solution is prepared by reacting hydrofluoric acid of a preset concentration and a silver nitrate solution of at least one concentration according to the detection requirements of the target object.

[0113] In a specific implementation, this embodiment can be detected by a Raman spectrometer at room temperature to obtain a working curve of the relationship between the target object and its Raman spectral characteristic peak intensity or peak area intensity at different concentrations, thereby establishing a Raman enhanced detection method for the target object.

[0114] In actual processing, the method for identifying drug residues in water in this embodiment specifically includes the following steps: based on the Raman detection method, the Raman spectra of single and mixed targets such as sulfadiazine, methylene blue and malachite green are collected, and a data set is established. Then, the collected Raman spectra are subjected to background subtraction, displacement clipping, airpls baseline processing, whittaker smoothing filtering, and normalization processing. Preferably, the clipped Raman shift is 400-1750cm -1 , the airpls baseline processing parameters are set to: lamdba = 100000, order = 2, and the whittaker smoothing filter parameters are set to lamdba = 100000, order = 2.

[0115] Step 220: sort and divide the Raman spectrum sample set by a preset random function to obtain a training set, a validation set, and a test set.

[0116] This embodiment can sort and divide the collected Raman spectrum sample set, and divide the Raman spectrum sample set as a data set. Preferably, the division method can be: the training set accounts for 80% and the test set accounts for 20%, and 20% of the training set is selected as the verification set to verify the accuracy and loss rate of the model. Before training the model, the data can be normalized, and the np.random.permutation function can be used to randomly disrupt the data order, and the one-hot encoding (One-HotEncoding) is used to convert the category label. So that the divided data can be put into the multi-layer perceptron algorithm to train the model later.

[0117] Step 230: Establish a multi-layer perceptron algorithm model based on a neural network according to the composition characteristics of the Raman spectrum sample set.

[0118] In the specific implementation, in view of the problems that the current SERS substrate is complex to make and difficult to identify multiple pollutants in water, this embodiment establishes a multi-layer perceptron algorithm model of the Tensorflow framework, also known as a multi-layer perceptron neural network model. The model has 1 input layer, 3 hidden layers and 1 output layer. The above-mentioned trimmed Raman shift 400-1750cm -1 The interpolation gap is 1cm -1 , there are 1351 features in total, and 1351 neurons in the input layer. The number of layers and neurons in the hidden layer is determined by a grid search procedure on the training set and the validation set. There are 3 layers in total, with 768, 384 and 48 neurons in each layer respectively. ReLU is selected as the activation function for all hidden layers. The output layer contains 7 neurons corresponding to 7 classification categories, and SOFTMAX is selected as the activation function for the output layer. For example, the model structure is as follows Figure 4 shown.

[0119] Optionally, this embodiment establishes a multi-layer perceptron algorithm model based on a neural network according to the composition characteristics of the Raman spectrum sample set, which may include: performing spectral feature number analysis based on the Raman spectrum sample set to obtain spectral feature numbers; establishing a multi-layer perceptron algorithm model based on a neural network, and setting the number of input layer neurons of the multi-layer perceptron algorithm model to correspond to the spectral feature numbers, and setting the output layer neurons of the multi-layer perceptron algorithm model to correspond to the target object recognition category.

[0120] Step 240, using the training set to perform model training on the multilayer perceptron algorithm model, and verifying it through the validation set during the model training process, and using the test set to test the model until the multilayer perceptron algorithm model training is completed, thereby obtaining a performance evaluation result of the model and a trained neural network algorithm model.

[0121] In a specific implementation, this embodiment uses a training set to train the multilayer perceptron algorithm model, and uses a validation set to validate it, and uses a test set to test the model. During the model training process, the multilayer perceptron algorithm model can be optimized in parameters, and the specific parameter optimization includes: using a stochastic gradient descent (SGD) optimizer for iterative optimization, and the SGD parameter settings are as follows: categorical_crossentropy is selected as the loss function, the learning rate of the training model is 0.01, the batch size is 20, the number of iterations is 200, and the iteration is performed until the loss function no longer decreases, that is, the multilayer perceptron neural network algorithm model after parameter optimization is obtained, and the model training is completed.

[0122] In the related art, the prior art also has a solution for unknown object identification method using pollutant identification algorithm, such as the existing technology (taking patent application document CN116678864A as an example), which uses PCA-SVM algorithm to identify different types of quinolone antibiotics, but this technology is not suitable for environmental systems with complex pollutants. PCA-SVM cannot distinguish the Raman spectra of mixed antibiotics, so the data processing method is not suitable for use under complex conditions. The solution of the embodiment of the present application establishes a multi-layer perceptron model for identifying Raman spectra of mixed target objects of various concentrations. The accuracy of the model for identifying different components reaches 97.8%, 5-fold cross validation is 97.4%, and the F1 score is 98.2%, indicating that the model is not only accurate but also has good robustness. In addition, the present embodiment also uses the SHAP interpretation model to evaluate the rationality of the features selected by the model. The aluminum foil SERS substrate proposed in this embodiment can detect and identify the target in the actual water sample in combination with the model, and has good application prospects in actual water samples.

[0123] For example, during the iteration process, the accuracy and loss rate of each iteration of the training set and the validation set can refer to Figure 5 As shown, the trained deep learning neural network model can quickly and accurately identify single and mixed targets of sulfadiazine, methylene blue and malachite green in water. The model accuracy is 97.8%, the 5-fold cross validation is 97.4%, and the F1 score is 98.2%, which proves that the model can accurately identify complex samples.

[0124] Step 250: Analyze the prediction results using a preset analysis method to obtain key feature samples of each category of target objects.

[0125] Among them, the key characteristic samples are concentrated in 447-450cm -1 、1104-1114cm -1 and 571cm -1The key feature sample is within the Raman shift range, and the key feature sample coincides with the characteristic peak of the target object, so as to distinguish the spectra of different categories of targets.

[0126] In the specific implementation, this embodiment uses the Kernel-Explainer method in the SHAP model interpretation tool to interpret the explanatory variables and feature selection of the model and evaluate the rationality of the model. Figure 6 As shown in the SHAP summary diagram, the model accurately identifies the key features of 7 types of targets, and the key features are mainly concentrated in 447-450cm -1 、1104-1114cm -1 and 571cm -1 These key features are consistent with the characteristic peaks of the target, which is of great significance for distinguishing the spectra of different categories of targets, and plays a key role in distinguishing the spectra of different categories of targets, indicating that the features selected by the model are reasonable. Finally, the model was used to detect and identify antibiotics spiked in actual water samples of different matrices, with an accuracy rate of 100%.

[0127] As an example, first, taking the preparation of aluminum foil SERS substrate and the rapid detection and identification of malachite green solution in aquaculture water as an example, referring to the experimental conditions of ① above, 1×10 -5 mol / L of malachite green, put the aluminum foil SERS substrate into the actual aquaculture water and soak it for 10 minutes. After taking it out and waiting for the solution to evaporate, the Raman signal on the aluminum foil SERS substrate is detected. The collected Raman spectrum data is preprocessed, and then the preprocessed Raman spectrum data is input into the multi-layer perceptron model for recognition. The model can accurately identify the target contained in the water as malachite green based on the collected Raman spectrum data.

[0128] Take the preparation of aluminum foil SERS substrate and the rapid detection and identification of mixed solutions of methylene blue, malachite green and sulfadiazine with different concentrations in aquaculture water as an example: referring to the experimental conditions of ① above, 1×10 - 5 mol / L malachite green, 1×10 -5 mol / L sulfadiazine and 1×10 -6 mol / L of methylene blue. Then the aluminum foil SERS substrate was immersed in the mixed solution for 30 minutes, taken out and the Raman signal on the aluminum foil SERS substrate was detected after the solution evaporated. The collected Raman spectrum data was preprocessed and input into the multi-layer perceptron model for recognition. The model can accurately identify the target objects contained in the water as malachite green, sulfadiazine and methylene blue based on the collected Raman spectrum data.

[0129] Therefore, this embodiment proposes Raman spectroscopy identification of drug residues in water based on a neural network algorithm, that is, a Raman spectroscopy identification method for drug residues in aquaculture water based on a multi-layer perceptron deep learning algorithm. The detection substrate is obtained by treating a metal foil with a silver nitrate solution containing hydrofluoric acid, and a Raman enhanced detection method for multiple drug residues in water is established on this substrate to achieve quantitative analysis. The identification of unknown substances is based on the above-mentioned detection method, collecting model data and establishing a data set, using a multi-layer perceptron neural network model, and training through the data set to optimize the parameter model.

[0130] It should be noted that the recognition algorithm of this embodiment uses a deep learning multi-layer perceptron algorithm to model the Raman spectra of single and mixed targets of sulfadiazine, methylene blue and malachite green. The obtained model combined with the detection substrate can be used for rapid detection and identification of drug residues in water environments.

[0131] Furthermore, the method provided in this embodiment can realize the detection of sulfadiazine, methylene blue and malachite green in water, and the aluminum foil SERS substrate can be quickly prepared by chemical replacement method, which reduces the operation cost and shortens the preparation time. By training the model and making the recognition software, the pollutants and their types can be quickly and accurately identified.

[0132] In summary, in terms of model training and construction of key features of samples, this embodiment first uses Raman spectroscopy technology to detect and analyze solution samples using the prepared Raman spectroscopy enhanced detection substrate to obtain Raman spectroscopy sample sets of single targets and mixed targets, and then sorts and divides the Raman spectroscopy sample sets to obtain training sets, validation sets and test sets, and establishes a multi-layer perceptron algorithm model based on a neural network based on the composition characteristics of the Raman spectroscopy sample sets, so as to use the training set for model training, and verify it through the validation set during the model training process, and use the test set for testing to obtain a trained neural network algorithm model, and then parse it through a preset parsing method to obtain key feature samples of each category of target. It can be seen that this patent uses a multi-layer perceptron deep learning model to identify mixed pollutants, and uses SHAP to parse the deep learning model, which provides a new technical solution for high-sensitivity and high-reproducibility detection of surface-enhanced Raman spectroscopy and rapid and accurate identification of drug residues in water by deep learning models. After effective training, the model can identify single and mixed targets such as sulfadiazine, methylene blue and malachite green in water, that is, accurately identify complex samples, and use the SHAP model interpretation tool to quickly and accurately identify pollutants and their types.

[0133] Furthermore, this embodiment uses a multi-layer perceptron algorithm to train a recognition model to obtain a trained neural network recognition model, which can be applied in practice. The model automatically captures the key features of the Raman spectrum data of the target object to be detected and identified, and then performs identification and classification based on the key features.

[0134] It should be noted here that there are multiple options for the ratio of the solutions. The concentrations and ratios of the solutions in the above embodiments are only examples. The present application does not limit the ratios of the target solutions and the metal substrate solutions.

[0135] It should be noted that, for the purpose of simple description, the method embodiments are expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited to the described order of actions, because according to the embodiments of the present application, certain steps may be performed in other orders or simultaneously.

[0136] like Figure 7 As shown, the embodiment of the present application also provides a device 700 for identifying Raman spectra of drug residues in water based on a neural network algorithm, comprising:

[0137] A Raman spectrum detection and analysis module 710 is used to perform Raman spectrum detection and analysis on a target solution to be detected by using a Raman spectrum enhancement detection substrate prepared by Raman spectrum technology to obtain Raman spectrum data, wherein the Raman spectrum data includes a Raman spectrum of at least one target to be detected and identified;

[0138] An analysis module 720 is used to analyze the target object type of the Raman spectrum data through a neural network algorithm model to obtain key features of the target object;

[0139] The category differentiation module 730 is used to perform category differentiation based on the key features in combination with preset key feature samples to obtain target detection and identification results; wherein the target solution to be detected is a target water solution detected by a metal substrate, the metal substrate is an aluminum foil SERS substrate, and the target solution to be detected contains at least one target to be detected and identified among levofloxacin, fleroxacin, pefloxacin, sulfadiazine, methylene blue and malachite green.

[0140] Optionally, the Raman spectrum detection and analysis module 710 includes:

[0141] A collection submodule, for collecting raw Raman spectral data from a target solution to be tested by using Raman spectroscopy technology and a Raman spectroscopy enhanced detection substrate, wherein the raw Raman spectral data includes a single target Raman spectrum or a mixed target Raman spectrum;

[0142] A preprocessing submodule is used to perform preprocessing based on the original Raman spectrum data to obtain Raman spectrum data; wherein the aluminum foil SERS substrate in the Raman spectrum enhanced detection substrate is obtained by a Raman enhanced detection substrate preparation method, using a metal foil to react with a silver nitrate solution containing hydrofluoric acid to form a nanosilver structure on the metal foil.

[0143] Optionally, the acquisition submodule is specifically used to detect the target solution to be tested under target temperature conditions through Raman spectroscopy detection method using a Raman spectroscopy enhanced detection substrate to obtain Raman spectroscopy characteristic peak information at the solution concentration corresponding to the target as the original Raman spectroscopy data; wherein the Raman spectroscopy characteristic peak information includes Raman spectroscopy characteristic peak intensity and / or peak area intensity.

[0144] Optionally, the analysis module 720 includes:

[0145] A classification submodule, used for inputting the Raman spectrum data into a neural network algorithm model to classify targets and obtain a first classification result, wherein the first classification result includes a single target classification result or a multi-target classification result;

[0146] The parsing submodule is used to perform a parsability analysis on the first classification result through the neural network algorithm model, and output a key feature of at least one target object corresponding to the identified Raman spectrum data.

[0147] Optionally, a parsing submodule is specifically used to select a target neuron for identification and analysis from the neural network algorithm model based on the first classification result; and perform feature analysis on the Raman spectral data through the target neuron to obtain key features.

[0148] Optionally, the device 700 for identifying Raman spectra of drug residues in water based on a neural network algorithm further includes:

[0149] A sample detection and analysis module is used to detect and analyze the sample solution by using Raman spectroscopy technology and the prepared Raman spectroscopy enhancement substrate to obtain a Raman spectroscopy sample set, wherein the Raman spectroscopy sample set includes a single target spectroscopy sample and a mixed target spectroscopy sample;

[0150] A sorting and partitioning module, used to sort and partition the Raman spectrum sample set by a preset random function to obtain a training set, a validation set and a test set;

[0151] A model building module, used to establish a multi-layer perceptron algorithm model based on a neural network according to the composition characteristics of the Raman spectrum sample set;

[0152] A model training and verification module is used to perform model training on the multi-layer perceptron algorithm model using the training set, and to verify the model using the verification set during the model training process, and to perform model testing using the test set until the multi-layer perceptron algorithm model training is completed, thereby obtaining a performance evaluation result of the model and a trained neural network algorithm model;

[0153] The parsing module is used to perform model parsing based on the performance evaluation results through the SHAP parsing method to obtain key feature samples of each category of target objects;

[0154] Among them, the key characteristic samples are concentrated in 447-450cm -1 、1104-1114cm -1 and 571cm -1 The key characteristic sample is within the Raman shift range, and the key characteristic sample is consistent with the characteristic peak of the target object, which is used to distinguish the spectra of different categories of targets;

[0155] Optionally, a model building module is specifically used to perform spectral feature number analysis based on the Raman spectrum sample set to obtain spectral feature numbers; establish a multi-layer perceptron algorithm model based on a neural network, and set the number of input layer neurons of the multi-layer perceptron algorithm model to correspond to the spectral feature numbers, and set the output layer neurons of the multi-layer perceptron algorithm model to correspond to the target object recognition category;

[0156] Optionally, a sample detection and analysis module is specifically used to collect Raman spectroscopy data of sample solutions of different concentrations in a sample at room temperature through Raman spectroscopy technology using a Raman spectroscopy enhanced substrate, obtain a working curve of the relationship between the target object and its Raman spectroscopy characteristic peak intensity or peak area intensity at different concentrations as a Raman spectroscopy sample set, and establish a Raman enhanced detection method for the target object; wherein, in the substrate preparation method of the Raman spectroscopy enhanced detection substrate, the reaction solution is prepared by reacting hydrofluoric acid of a preset concentration and a silver nitrate solution of at least one concentration according to the detection requirements of the target object.

[0157] It should be noted that the Raman spectroscopy identification device for drug residues in water based on a neural network algorithm provided in each embodiment of this patent can execute the Raman spectroscopy identification method for drug residues in water based on a neural network algorithm provided in any embodiment of this application, and has the corresponding functions and beneficial effects of the execution method.

[0158] In the specific implementation, Figure 8 As shown, this embodiment also provides a water drug residue detection and identification system for implementing the water drug residue Raman spectrum identification method based on a neural network algorithm, the system comprising:

[0159] A Raman spectrum detection device 81 is used to perform Raman spectrum detection and analysis on a target solution to be detected by using a prepared Raman spectrum enhanced detection substrate through Raman spectrum technology to obtain Raman spectrum data, wherein the Raman spectrum data includes a Raman spectrum of at least one target to be detected and identified;

[0160] The detection and analysis device 82 is connected to the Raman spectrum detection device, and is used to analyze the target type of the Raman spectrum data through a neural network algorithm model to obtain the key features of the target; based on the key features, combined with preset key feature samples, the category is distinguished to obtain the target detection and identification results; wherein the target solution to be detected is a target water solution detected by a metal substrate, the metal substrate is an aluminum foil SERS substrate, and the target solution to be detected contains at least one target to be detected and identified among levofloxacin, fleroxacin, pefloxacin, sulfadiazine, methylene blue and malachite green.

[0161] Optionally, the Raman spectrum detection device 81 includes:

[0162] A first Raman spectrometer is used to collect original Raman spectral data from a target solution to be detected by using Raman spectroscopy technology and a Raman spectroscopy enhanced detection substrate, wherein the original Raman spectral data includes a single target Raman spectrum or a mixed target Raman spectrum;

[0163] A spectral data processor is used to perform preprocessing based on the original Raman spectral data to obtain Raman spectral data; wherein the aluminum foil SERS substrate in the Raman spectral enhanced detection substrate is obtained by a Raman enhanced detection substrate preparation method, using a metal foil to react with a silver nitrate solution containing hydrofluoric acid to form a nanosilver structure on the metal foil.

[0164] Optionally, the first Raman spectrometer is specifically used to detect the target solution to be tested under target temperature conditions through Raman spectroscopic detection method using a Raman spectroscopic enhanced detection substrate to obtain Raman spectroscopic characteristic peak information at the solution concentration corresponding to the target as the original Raman spectroscopic data; wherein the Raman spectroscopic characteristic peak information includes Raman spectroscopic characteristic peak intensity and / or peak area intensity.

[0165] Optionally, the detection and analysis equipment 82 includes:

[0166] A classification device, used for inputting the Raman spectrum data into a neural network algorithm model to classify targets and obtain a first classification result, wherein the first classification result includes a single target classification result or a multi-target classification result;

[0167] The parsing device is used to perform a parsability analysis on the first classification result through the neural network algorithm model, and output a key feature of at least one target object corresponding to the identified Raman spectrum data.

[0168] Optional, assorted equipment, including:

[0169] A classifier, used for inputting the Raman spectrum data into a neural network algorithm model to classify targets and obtain a first classification result, wherein the first classification result includes a single target classification result or a multi-target classification result;

[0170] A parser is used to perform a parsability analysis on the first classification result through the neural network algorithm model, and output a key feature of at least one target object corresponding to the identified Raman spectrum data.

[0171] Optionally, the classifier is specifically used to select a target neuron for identification and analysis from the neural network algorithm model based on the first classification result; and perform feature analysis on the Raman spectral data through the target neuron to obtain key features.

[0172] Optional, water drug residue detection and identification system also includes:

[0173] A second Raman spectrometer is used to detect and analyze the sample solution by using the Raman spectroscopy technique and the prepared Raman spectroscopy enhancement substrate to obtain a Raman spectroscopy sample set, wherein the Raman spectroscopy sample set includes a single target spectroscopy sample and a mixed target spectroscopy sample;

[0174] A model building and training device is connected to the second Raman spectrometer and is used to establish a multi-layer perceptron algorithm model based on a neural network according to the composition characteristics of the Raman spectrum sample set; the multi-layer perceptron algorithm model is trained using the training set, and is verified by the verification set during the model training process; the model is tested using the test set until the multi-layer perceptron algorithm model is trained, thereby obtaining a performance evaluation result of the model and a trained neural network algorithm model;

[0175] The analysis device is connected to the model building and training device, and is used to analyze the model by the SHAP analysis method according to the performance evaluation results to obtain key feature samples of each category of target; wherein the sample solution contains at least one concentration of target solution sample, and the key feature samples are concentrated in 447-450cm -1 、1104-1114cm -1 and 571cm -1 The key feature sample is within the Raman shift range, and the key feature sample coincides with the characteristic peak of the target object, so as to distinguish the spectra of different categories of targets.

[0176] Optional, model building and training equipment, including:

[0177] A feature number analyzer, connected to the second Raman spectrum detector, for performing spectrum feature number analysis according to the Raman spectrum sample set to obtain spectrum feature numbers;

[0178] A model builder is used to establish a multi-layer perceptron algorithm model based on a neural network, and to set the number of input layer neurons of the multi-layer perceptron algorithm model to correspond to the number of spectral features, and to set the output layer neurons of the multi-layer perceptron algorithm model to correspond to the target object recognition category.

[0179] Optionally, a second Raman spectrometer is specifically used to collect Raman spectroscopy data of sample solutions of different concentrations in a sample at room temperature through Raman spectroscopy technology using a Raman spectroscopy enhanced substrate, to obtain a working curve of the relationship between the target object and its Raman spectroscopy characteristic peak intensity or peak area intensity at different concentrations as a Raman spectroscopy sample set, and to establish a Raman enhanced detection method for the target object; wherein, in the substrate preparation method of the Raman spectroscopy enhanced detection substrate, the reaction solution is prepared by reacting hydrofluoric acid of a preset concentration and a silver nitrate solution of at least one concentration according to the detection requirements of the target object.

[0180] Furthermore, an embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the Raman spectroscopy identification method for drug residues in water based on a neural network algorithm as provided in any of the aforementioned method embodiments are implemented.

[0181] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0182] The above description is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest range consistent with the principles and novel features applied for herein.

Claims

1. A method for identifying drug residues in water by Raman spectroscopy based on a neural network algorithm, characterized in that: include: Using Raman spectroscopy technology, using the prepared Raman spectroscopy enhanced detection substrate to perform Raman spectroscopy detection and analysis on a target solution to be detected, to obtain Raman spectroscopy data, wherein the Raman spectroscopy data includes a Raman spectrum of at least one target to be detected and identified; By using a neural network algorithm model, the Raman spectroscopy data is analyzed for target species to obtain key features of the target; Based on the key features, combined with preset key feature samples, category differentiation is performed to obtain target object detection and recognition results; Among them, the target solution to be detected is a target water solution detected by a metal substrate, the metal substrate is an aluminum foil SERS substrate, and the target solution to be detected contains at least one target to be detected and identified among levofloxacin, fleroxacin, pefloxacin, sulfadiazine, methylene blue and malachite green.

2. The method according to claim 1, characterized in that Through Raman spectroscopy technology, the prepared Raman spectroscopy enhanced detection substrate is used to perform Raman spectroscopy detection and analysis on the target solution to obtain Raman spectroscopy data, including: By using Raman spectroscopy technology, raw Raman spectroscopy data is collected from a target solution to be tested using a Raman spectroscopy enhanced detection substrate, wherein the raw Raman spectroscopy data includes a single target Raman spectrum or a mixed target Raman spectrum; Performing preprocessing based on the original Raman spectrum data to obtain Raman spectrum data; The aluminum foil SERS substrate in the Raman enhanced detection substrate is obtained by a Raman enhanced detection substrate preparation method, in which a metal foil is reacted with a silver nitrate solution containing hydrofluoric acid to form a nanosilver structure on the metal foil.

3. The method according to claim 2, characterized in that Through Raman spectroscopy technology, the original Raman spectroscopy data is collected from the target solution using the Raman spectroscopy enhanced detection substrate, including: Under the target temperature condition, the target solution to be tested is detected by Raman spectroscopy detection method using a Raman spectroscopy enhanced detection substrate to obtain Raman spectroscopy characteristic peak information at the solution concentration corresponding to the target as the original Raman spectroscopy data; The Raman spectrum characteristic peak information includes Raman spectrum characteristic peak intensity and / or peak area intensity.

4. The method according to claim 1, characterized in that: The Raman spectral data is analyzed for target types through a neural network algorithm model to obtain key features of the target to be detected and identified, including: Inputting the Raman spectrum data into a neural network algorithm model to classify targets to obtain a first classification result, wherein the first classification result includes a single target classification result or a multi-target classification result; For the first classification result, an analyzability analysis is performed using the neural network algorithm model, and a key feature of at least one target object corresponding to the identified Raman spectrum data is output.

5. The method according to claim 4, characterized in that For the first classification result, performing an analyzability analysis using the neural network algorithm model includes: Based on the first classification result, selecting a target neuron for identification and analysis from the neural network algorithm model; The Raman spectrum data is subjected to feature analysis by the target neuron to obtain key features.

6. The method according to any one of claims 1 to 5, characterized in that: Before analyzing the target species of the Raman spectroscopy data by using a neural network algorithm model, the method further includes: By using Raman spectroscopy technology, the sample solution is detected and analyzed using the prepared Raman spectroscopy enhancement substrate to obtain a Raman spectroscopy sample set, wherein the Raman spectroscopy sample set includes a single target spectroscopy sample and a mixed target spectroscopy sample; The Raman spectrum sample set is sorted and divided by a preset random function to obtain a training set, a validation set and a test set; According to the composition characteristics of the Raman spectrum sample set, a multi-layer perceptron algorithm model based on a neural network is established; The multi-layer perceptron algorithm model is trained using the training set, and is verified using the validation set during the model training process, and the model is tested using the test set until the multi-layer perceptron algorithm model training is completed, thereby obtaining a performance evaluation result of the model and a trained neural network algorithm model; Based on the performance evaluation results, the model is analyzed using the SHAP analysis method to obtain key feature samples of each category of target objects; The sample solution contains at least one target solution sample of concentration, and the key characteristic samples are concentrated at 447-450cm -1 、1104-1114cm -1 and 571cm -1 The key feature sample is within the Raman shift range, and the key feature sample coincides with the characteristic peak of the target object, so as to distinguish the spectra of different categories of targets.

7. The method according to claim 6, characterized in that According to the composition characteristics of the Raman spectrum sample set, a multi-layer perceptron algorithm model based on a neural network is established, including: Performing spectral feature number analysis according to the Raman spectrum sample set to obtain spectral feature numbers; A multi-layer perceptron algorithm model based on a neural network is established, and the number of input layer neurons of the multi-layer perceptron algorithm model is set to correspond to the number of spectral features, and the output layer neurons of the multi-layer perceptron algorithm model are set to correspond to the target object recognition category.

8. The method according to claim 6, characterized in that Through Raman spectroscopy technology, the sample solution is detected and analyzed using the prepared Raman spectroscopy enhancement substrate to obtain a Raman spectroscopy sample set, including: Under room temperature conditions, Raman spectroscopy technology is used to collect Raman spectroscopy data of sample solutions of different concentrations in the sample using a Raman spectroscopy enhancement substrate, and a working curve of the relationship between the target object at different concentrations and its Raman spectroscopy characteristic peak intensity or peak area intensity is obtained as a Raman spectroscopy sample set, and a Raman enhanced detection method for the target object is established; In the substrate preparation method of the Raman spectroscopy enhanced detection substrate, the reaction solution is prepared by reacting hydrofluoric acid of a preset concentration and a silver nitrate solution of at least one concentration according to the detection requirements of the target object.

9. A device for identifying drug residues in water by Raman spectroscopy based on a neural network algorithm, characterized in that: include: A Raman spectrum detection and analysis module is used to perform Raman spectrum detection and analysis on a target solution to be detected by using a prepared Raman spectrum enhanced detection substrate through Raman spectrum technology to obtain Raman spectrum data, wherein the Raman spectrum data includes at least one Raman spectrum of a target to be detected and identified; An analysis module, used to analyze the target object type of the Raman spectrum data through a neural network algorithm model to obtain key features of the target object; The category differentiation module is used to perform category differentiation based on the key features in combination with preset key feature samples to obtain target detection and identification results; wherein the target solution to be detected is a target water solution detected by a metal substrate, the metal substrate is an aluminum foil SERS substrate, and the target solution to be detected contains at least one target to be detected and identified among levofloxacin, fleroxacin, pefloxacin, sulfadiazine, methylene blue and malachite green.

10. A Raman spectrum identification system for drug residues in water based on a neural network algorithm, characterized in that: The system comprises: A Raman spectroscopy detection device is used to perform Raman spectroscopy detection and analysis on a target solution to be detected by using a prepared Raman spectroscopy enhanced detection substrate through Raman spectroscopy technology to obtain Raman spectroscopy data, wherein the Raman spectroscopy data includes a Raman spectrum of at least one target to be detected and identified; The detection and analysis equipment is connected to the Raman spectrum detection equipment, and is used to analyze the target type of the Raman spectrum data through a neural network algorithm model to obtain the key features of the target; based on the key features, combined with preset key feature samples, the category is distinguished to obtain the target detection and identification results; wherein the target solution to be detected is a target water solution detected by a metal substrate, the metal substrate is an aluminum foil SERS substrate, and the target solution to be detected contains at least one target to be detected and identified among levofloxacin, fleroxacin, pefloxacin, sulfadiazine, methylene blue and malachite green.

Images