An oil-in-water spectrum analysis and data transmission system based on AI technology

Abstract

The present invention provides an oil-in-water spectral analysis and data transmission system based on AI technology, which relates to the field of oil quality detection technology. The system includes a spectral detection module for emitting multi-wavelength light to a water sample and collecting optical signals; an AI processing module connected to the spectral detection module for performing spectral analysis based on the optical signals, and using an association algorithm group including a light absorption-scattering coupling model, a self-attention feature enhancement model, and a multimodal classification prediction model to output oil concentration and distribution information; a data recording module connected to the AI ​​processing module for storing detection data; an auxiliary function module connected to the data recording module for transmitting detection data, issuing voice prompts, and powering the system; light scattering interference is eliminated through a scattering correction unit and a coupling model, and the weight of oil characteristic wavelengths is automatically enhanced through a self-attention mechanism to improve the accuracy of mixed oil identification. At the same time, wireless charging, real-time cloud transmission, voice alarm functions, and oil-in-water monitoring are combined.

Description

Technical Field

[0001] The present invention relates to the technical field of oil quality detection, and in particular to an oil-in-water spectrum analysis and data transmission system based on AI technology. Background Art

[0002] In recent years, with the increasing popularity of fiber optic sensors, fluorescence sensors, and deep learning algorithms, many researchers in the oil quality testing field have been researching and attempting to combine neural networks, support vector machines (SVMs), and other technologies with spectral data to improve the anti-interference capabilities of oil detection. They also aim to overcome the problem that traditional testing equipment, which mostly uses wired methods such as RS-485 and USB to transmit data and store it on local hard drives, makes remote monitoring and real-time warning difficult.

[0003] However, when existing spectral detection equipment analyzes water samples containing suspended matter or emulsified oil, the scattering of light by particulate matter will lead to errors in the measurement of transmitted light intensity, especially in the short wavelength region. The scattering interference makes the calculation error of the absorption coefficient large, resulting in distorted concentration detection results; and for mixed oil water samples (such as a mixture of crude oil and fuel oil), traditional spectral analysis can only be based on a single wavelength or a few characteristic peaks, and cannot accurately distinguish the proportion and distribution of different oils; moreover, traditional detection equipment is mostly desktop machines, which require manual sampling and then analysis in the laboratory, and cannot meet the needs of field or online monitoring; although some portable devices support on-site detection, they lack functions such as voice prompts and wireless charging, are complicated to operate and have limited battery life.

[0004] Therefore, this field urgently needs an oil-in-water spectral analysis and data transmission system based on AI technology to solve the above problems. Summary of the Invention

[0005] The present invention provides an AI-based oil-in-water spectral analysis and data transmission system, which aims to eliminate light scattering interference through a scattering correction unit and a coupling model, automatically enhance the weight of oil characteristic wavelengths through a self-attention mechanism, and improve the accuracy of mixed oil identification. At the same time, it combines wireless charging, real-time cloud transmission and voice alarm functions to achieve full-scene, intelligent oil-in-water monitoring.

[0006] The present invention provides an oil-in-water spectrum analysis and data transmission system based on AI technology, comprising: a spectrum detection module, an AI processing module, a data recording module, and an auxiliary function module;

[0007] The spectrum detection module is used to emit multi-wavelength light to the water sample and collect light signals;

[0008] The AI ​​processing module is connected to the spectral detection module and is used to perform spectral analysis based on the optical signal, using an association algorithm group including a light absorption-scattering coupling model, a self-attention feature enhancement model, and a multimodal classification prediction model to output oil concentration and distribution information;

[0009] The data recording module is connected to the AI ​​processing module and is used to store detection data;

[0010] The auxiliary function module is connected to the data recording module and is used for transmitting detection data, issuing voice prompts and supplying power to the system.

[0011] According to the present invention, an oil-in-water spectral analysis and data transmission system based on AI technology is provided, wherein the spectral detection module includes:

[0012] A light source unit, which is used to emit continuous spectrum light of preset different wavelength combinations to provide an excitation light source for detecting the spectral characteristics of oil substances in water samples;

[0013] The sample pool unit adopts a flow-through design and has a built-in stirring device to homogenize the water sample, and is used to contain the water sample and allow light to pass through;

[0014] A spectral sensor unit comprising a cooled CCD array for collecting incident light intensity and transmitted light intensity;

[0015] The scattering correction unit is used to calculate the scattering coefficient to correct the interference of light signals at different wavelengths caused by scattering of particulate matter in the water sample; the scattering coefficient The calculation formula is ,in is the average oil droplet volume, measured based on a laser particle size analyzer. is the relative refractive index of oil and water, obtained based on the oil parameter table, is the scattering angle, is the design angle of the scattered light collection channel in the spectral sensor unit, is the wavelength of incident light.

[0016] According to the present invention, an oil-in-water spectrum analysis and data transmission system based on AI technology is provided, wherein the AI ​​processing module includes:

[0017] The feature extraction unit is used to establish the light absorption-scattering coupling model and derive the absorption coefficient of oil substances at different wavelengths. The light absorption-scattering coupling model is:

[0018] in is the transmitted light intensity, is the incident light intensity, is the absorption coefficient of the oil substance being sought, is the scattering coefficient determined by the scattering correction unit; is the optical path length;

[0019] At the same time, through the formula Convert the transmitted light intensity into absorbance; is the absorbance spectrum; the absorbance at different wavelengths is organized into spectral feature vectors ,in is the number of wavelength points;

[0020] A prediction unit that calculates feature weights using the self-attention feature enhancement model, where:

[0021] in is the query representation of the spectral feature vector, used to identify the key wavelengths, is the key representation of the spectral feature vector, used to match important features, The original value of the spectral feature vector, used for weighted aggregation, They are all generated by linear transformation of spectral features; represent and Multiply the transpose of ; The dimensions of , is the feature dimension; the self-attention feature enhancement model realizes feature weight distribution through matrix operation, calculates the importance weights of different wavelength features, and makes the model target the wavelength corresponding to the characteristic peak of oil;

[0022] The type and concentration analysis unit is connected to the feature extraction unit and is based on the absorption coefficient of the oil substance. Determine the type of oil in the current water sample based on the oil parameter table; Combined with the Lambert-Beer law, the concentration of the oil sample in the current water sample is obtained by reverse calculation;

[0023] The probability distribution analysis unit is connected to the feature extraction unit and the prediction unit to convert the spectral feature vector and feature weights Substitute into the multimodal classification prediction model to determine the probability distribution of various oil samples in the water sample.

[0024] According to an AI-based oil-in-water spectrum analysis and data transmission system provided by the present invention, in the probability distribution analysis unit, the multimodal classification prediction model is:

[0025] in, is the probability distribution of the i-th type of oil, is the weight parameter of category i for feature j, determined based on training; Weighting attention The jth wavelength characteristic after The feature weight The result of normalization;

[0026] Determine the probability distribution of various oil samples through the multimodal classification prediction model .

[0027] According to an AI-based oil-in-water spectrum analysis and data transmission system provided by the present invention, the auxiliary function module includes a wireless charging unit, a voice prompt unit, and a transmission unit.

[0028] According to the AI-based oil-in-water spectral analysis and data transmission system provided by the present invention, the wireless charging unit is based on the wireless charging standard. The wireless charger generates an alternating magnetic field through an alternating current. After the receiving coil at the detector senses the alternating magnetic field, it generates an induced electromotive force to charge the battery inside the detector.

[0029] When the wireless charging unit detects a valid wireless charging signal, it automatically switches to wireless charging mode and cuts off the battery power supply. When the wireless charging unit ends wireless charging or the charging signal disappears, the wireless charging unit automatically switches back to battery power supply mode.

[0030] The receiving coil includes a high-sensitivity receiving coil, a rectifier circuit, a voltage stabilizing circuit and an intelligent charging management chip, and the receiving coil receives charging signals within a standard range.

[0031] According to a water oil spectrum analysis and data transmission system based on AI technology provided by the present invention, the voice prompt unit is used to generate operation voice guidance corresponding to each working state for all working states of the equipment, and at the same time compare the oil sample type, concentration and probability distribution obtained based on the analysis with the oil sample type-concentration-probability distribution threshold mapping table. If the concentration of any oil sample type is greater than its threshold, a voice alarm information is issued, wherein the voice alarm information includes an alarm voice and a warning sound.

[0032] According to the present invention, a water oil spectrum analysis and data transmission system based on AI technology is provided. The transmission unit is used for transmission via wireless communication. The detector uploads real-time monitoring data, equipment working status, voice alarm information and the concentration change curve of each oil sample to a cloud server for storage; the user views the equipment working status and detection results of the detector based on the cloud server through a remote device.

[0033] According to the present invention, an oil-in-water spectrum analysis and data transmission system based on AI technology is provided, wherein the data recording module includes:

[0034] a change curve generating unit, which is used to periodically extract the concentration data and probability distribution data of each oil sample type in the water sample to obtain a data set for each oil sample type;

[0035] Generate a change curve chart of each oil sample based on the data set of each oil sample type, with the time node as the horizontal axis and the concentration and probability distribution as the vertical axis;

[0036] The storage unit is used to automatically store the oil sample change curve diagrams of each cycle.

[0037] According to the present invention, an oil-in-water spectrum analysis and data transmission system based on AI technology also includes a display module, which is connected to the storage unit and is used to display the change curve of each oil sample in each cycle to the manager.

[0038] Compared with the prior art, the present invention has the following advantages:

[0039] 1. The spectral detection module of the present application emits multi-wavelength continuous spectrum light through a light source unit, combines a flow-through sample cell and a stirring device to ensure the uniformity of the water sample, and cooperates with a refrigerated CCD array to collect light intensity signals; on this basis, the AI ​​processing module constructs a light absorption-scattering coupling model, relates the transmitted light intensity to the incident light intensity, scattering coefficient, and optical path length, and derives the absorption coefficient of oil substances, breaking through the limitation of the traditional Lambert-Beer law that only considers the absorption effect; at the same time, the scattering correction unit calculates the scattering coefficient through a formula, comprehensively considers parameters such as oil droplet volume, refractive index, and scattering angle, and corrects the effect of particulate matter on light signals in real time. The self-attention feature enhancement model automatically assigns wavelength feature weights through QKV matrix operations, so that the model focuses on the key wavelengths corresponding to the characteristic peaks of oils, and can enhance the effective signals without manually pre-setting features; the multimodal classification prediction model combines spectral feature vectors and feature weights to calculate the probability distribution of various oil samples in water samples, and realizes the type identification and concentration inversion of mixed oils; this mechanism enables the system to still efficiently process spectral data in complex water quality, shortening the oil concentration detection cycle from tens of minutes of traditional technology to minutes, meeting the needs of rapid on-site analysis.

[0040] 2. The data recording module of the present application periodically extracts oil sample concentration and probability distribution data through the change curve generation unit, automatically generates and saves a change curve graph with time as the horizontal axis, and realizes dynamic visualization of the detection data; the transmission unit uploads real-time monitoring data, equipment status, voice alarm information and concentration change curve to the cloud through wireless communication, which can be viewed by users at any time through remote devices, breaking the limitations of traditional local storage and scheduled upload, and improving the early warning response from hourly level to minute level, which is suitable for emergency handling of sudden pollution incidents; the voice prompt unit in the auxiliary function module generates operation instructions based on the working status of the equipment. When the oil sample concentration exceeds the threshold, a voice alarm and warning sound are issued in combination with the type-concentration-probability distribution threshold mapping table, which is more intuitive than the traditional screen display and reduces the threshold for on-site operation; the wireless charging unit automatically switches the power supply mode based on the principle of electromagnetic induction, cuts off the battery power supply when a valid charging signal is detected, and restores the battery mode after the end, solving the problem of traditional equipment requiring frequent charging or battery replacement, and improving the endurance and applicability of outdoor, underwater and other scenarios.

[0041] 3. This application adopts a modular architecture of spectral detection, AI processing, data recording, and auxiliary functions, and each module operates independently and works together; the flow-through sample pool and scattering correction unit of the spectral detection module are adapted to complex water samples containing suspended matter, and the algorithm group of the AI ​​processing module can flexibly adapt to different types of oil such as crude oil and fuel oil through the oil parameter table, and can achieve multi-scenario applications without replacing hardware; the display module is connected to the storage unit to intuitively present the oil sample change curve of each cycle to the manager to assist in decision-making analysis; this application avoids the cumbersome process of traditional detection equipment relying on manual sampling and laboratory analysis, and realizes full-link automation from on-site detection to data display, reducing operation and maintenance costs while improving system reliability.

[0042] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0043] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0045] Figure 1 This is a structural diagram of an oil-in-water spectral analysis and data transmission system based on AI technology provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0047] Example 1:

[0048] The present invention provides an oil-in-water spectrum analysis and data transmission system based on AI technology. Figure 1 ,include:

[0049] The present invention provides an oil-in-water spectrum analysis and data transmission system based on AI technology, comprising: a spectrum detection module, an AI processing module, a data recording module, and an auxiliary function module;

[0050] The spectrum detection module is used to transmit multi-wavelength light to the water sample and collect light signals;

[0051] The AI ​​processing module is connected to the spectral detection module to perform spectral analysis based on optical signals. It uses an association algorithm group including a light absorption-scattering coupling model, a self-attention feature enhancement model, and a multimodal classification prediction model to output oil concentration and distribution information.

[0052] The data recording module is connected to the AI ​​processing module to store the detection data;

[0053] The auxiliary function module is connected to the data recording module and is used to transmit detection data, issue voice prompts and provide power to the system.

[0054] The spectrum detection module includes:

[0055] A light source unit, which is used to emit continuous spectrum light of preset different wavelength combinations to provide an excitation light source for detecting the spectral characteristics of oil substances in water samples;

[0056] The sample pool unit adopts a flow-through design and has a built-in stirring device to homogenize the water sample, and is used to contain the water sample and allow light to pass through;

[0057] A spectral sensor unit comprising a cooled CCD array for collecting incident light intensity and transmitted light intensity;

[0058] The scattering correction unit is used to calculate the scattering coefficient to correct the interference of light signals at different wavelengths caused by scattering of particulate matter in the water sample; the scattering coefficient The calculation formula is ,in is the average oil droplet volume, measured based on a laser particle size analyzer. is the relative refractive index of oil and water, obtained based on the oil parameter table, is the scattering angle, is the design angle of the scattered light collection channel in the spectral sensor unit, is the wavelength of incident light.

[0059] It should be noted that the light source unit uses a xenon lamp or a halogen tungsten lamp as a broad-spectrum light source, and is paired with a programmable wavelength filter (such as an acousto-optic tunable filter AOTF or a diffraction grating), and emits continuous spectrum light of a preset wavelength combination (such as ultraviolet 200-400nm, visible light 400-700nm, and near-infrared 700-1100nm) through a control system.

[0060] The sample pool unit is made of quartz glass or polytetrafluoroethylene (PTFE), with water sample inlets and outlets at both ends. It cooperates with a peristaltic pump to achieve continuous flow of water samples, avoiding precipitation or stratification caused by static sampling. At the same time, a built-in micro magnetic stirrer or impeller stirrer stirs the oil droplets in the water sample to evenly disperse them through stirring, ensuring the representativeness of the optical signal acquisition.

[0061] The spectral sensor unit uses a cooled CCD array to reduce the interference of thermal noise on the light intensity signal.

[0062] The model in the scattering correction unit is based on Mie scattering theory, which describes the light scattering effect of uniform spherical particles. The average oil droplet volume characterizes the size distribution of oil droplets in a water sample; larger volumes increase the light scattering effect. After measuring the oil droplet size distribution using a laser particle size analyzer, a volume-weighted average is calculated to reflect the macroscopic properties of the scattering particles. The relative refractive index, the ratio of the refractive index of the oil to that of water, determines the intensity of light reflection and refraction at the oil-water interface. When m is close to 1 (e.g., light oil), the scattering effect is weak; when m deviates from 1 (e.g., heavy oil), the scattering effect is enhanced. The scattering angle is the angle between the scattered light collection channel and the direction of the incident light, describing the distribution of scattered light intensity at different angles.

[0063] To further optimize the above embodiment, the AI ​​processing module includes:

[0064] The feature extraction unit is used to establish a light absorption-scattering coupling model to derive the absorption coefficient of oil substances at different wavelengths. The light absorption-scattering coupling model is:

[0065]

[0066] in is the transmitted light intensity, is the incident light intensity, is the absorption coefficient of the oil substance being sought, is the scattering coefficient determined by the scattering correction unit; is the optical path length;

[0067] At the same time, through the formula Convert the transmitted light intensity into absorbance; is the absorbance spectrum; the absorbance at different wavelengths is organized into spectral feature vectors ,in is the number of wavelength points;

[0068] The prediction unit calculates the feature weights through the self-attention feature enhancement model, which is:

[0069]

[0070] in is the query representation of the spectral feature vector, used to identify the key wavelengths, is the key representation of the spectral feature vector, used to match important features, The original value of the spectral feature vector, used for weighted aggregation, They are all generated by linear transformation of spectral features; represent and Multiply the transpose of ; The dimensions of , is the feature dimension; the self-attention feature enhancement model realizes feature weight distribution through matrix operation, calculates the importance weights of features of different wavelengths, and makes the model focus on the wavelength corresponding to the characteristic peak of oil;

[0071] Type and concentration analysis unit, which is connected to the feature extraction unit, is based on the absorption coefficient of oil substances Determine the type of oil in the current water sample based on the oil parameter table; Combined with the Lambert-Beer law, the concentration of the oil sample in the current water sample is obtained by reverse calculation;

[0072] The probability distribution analysis unit is connected with the feature extraction unit and the prediction unit to convert the spectral feature vector and feature weights Substitute it into the multimodal classification prediction model to determine the probability distribution of various oil samples in water samples.

[0073] It should be noted that the light absorption-scattering coupling model describes that when light passes through a water sample, its intensity attenuation is affected by both absorption and scattering effects. The traditional Lambert-Beer law only considers absorption, while this model corrects the scattering interference of particulate matter on light by introducing the scattering coefficient, making it applicable to complex water samples containing suspended matter. Taking the natural logarithm of both sides of the formula and sorting it out, we can obtain: , which converts the absorption coefficient Separation into light intensity attenuation contribution and scatter correction , ensuring that subsequent analysis focuses on the true absorption characteristics of the oil;

[0074] absorbance It is proportional to the concentration of the substance (Lambert-Beer law), but in actual measurement it is affected by scattering. The absorbance values ​​at each wavelength are arranged as vectors. , where n is the number of wavelength points (e.g., n=2048); this vector captures the absorption characteristics of the water sample across the entire spectrum and serves as input for subsequent AI analysis;

[0075] In the self-attention feature enhancement model, A query matrix is ​​used to identify key wavelengths associated with oil characteristics; It is a key matrix that stores the “labels” of all wavelength features and is used to match important features; is a value matrix that stores the original eigenvalues ​​for weighted aggregation; As a learnable linear transformation matrix; its calculation process is:

[0076] 1. Pass Calculate the correlation between each wavelength and other wavelengths (similarity matrix);

[0077] 2. Divide by To prevent gradient disappearance, the similarity is normalized into a weight distribution through softmax;

[0078] 3. Use weights to perform weighted summation on the original features V to enhance the contribution of the key wavelengths;

[0079] Because oil substances have characteristic absorption peaks at specific wavelengths (such as crude oil at 260nm and fuel oil at 290nm), the self-attention mechanism automatically assigns higher weights to these characteristic peaks through learning, suppressing the interference of noise wavelengths. For example, if a wavelength point corresponds to an oil characteristic peak, its value in the weight matrix will be significantly increased, so that the characteristics of this wavelength are strengthened in subsequent analysis;

[0080] Furthermore, the oil parameter table pre-stores the standard absorption coefficient spectra of different oil types at various wavelengths. , such as crude oil, diesel, lubricating oil, etc.; this embodiment calculates the similarity between the measured absorption coefficient and the standard spectrum, such as cosine similarity: , select the oil type with the highest similarity as the recognition result;

[0081] The modified form of the Lambert-Beer law is ,in is the molar absorption coefficient, which is related to the oil type and wavelength and can be obtained from the oil parameter table; is the oil concentration; is the optical path length (known);

[0082] The inversion steps include:

[0083] 1. Select the characteristic peak wavelength (such as 260nm for crude oil) and obtain the wavelength ;

[0084] 2. Substitute into the formula: , Calculated by the feature extraction unit;

[0085] Furthermore, the spectral feature vector and feature weights Substituting into the multimodal classification prediction model, the process of determining the probability distribution of various oil samples in water samples is as follows:

[0086] Establish a multimodal classification prediction model, which is:

[0087]

[0088] in, is the probability distribution of the i-th type of oil, is the weight parameter of category i for feature j, determined based on training; Weighting attention The jth wavelength characteristic after The feature weight The result of normalization;

[0089] Determine the probability distribution of various oil samples through a multimodal classification prediction model .

[0090] It should be noted that the multimodal classification prediction model is used to input the spectral feature vector and self-attention weight into the neural network to construct a conditional probability distribution whose output vector represents the probability of existence of various types of oil in the water sample; for example, if , , it indicates that the water sample may be a mixture of 70% crude oil and 30% diesel, providing a basis for the quantitative analysis of mixed oils;

[0091] In this embodiment, the collaborative workflow of each unit in the AI ​​processing module is as follows:

[0092] 1. Feature extraction: Separate the absorption and scattering effects through a coupled model to generate a spectral feature vector;

[0093] 2. Feature enhancement: The self-attention mechanism calculates wavelength weights and enhances oil characteristic peaks;

[0094] 3. Type identification: Compare with the standard spectrum to determine the oil type;

[0095] 4. Concentration calculation: based on the absorbance of the characteristic peak wavelength and molar absorptivity to invert oil concentration;

[0096] 5. Probability distribution: Combines features and weights to output the probability of existence of each type of oil, supporting mixed oil analysis;

[0097] This architecture achieves high-precision conversion from raw light signals to oil information through the combination of physical models and AI algorithms, and is particularly suitable for detection scenarios of complex water quality and mixed oils.

[0098] In order to further optimize the above embodiment, the auxiliary function module includes a wireless charging unit, a voice prompt unit, and a transmission unit.

[0099] Specifically, the wireless charging unit is based on the wireless charging standard. The wireless charger generates an alternating magnetic field through an alternating current. After the receiving coil at the detector senses the alternating magnetic field, it generates an induced electromotive force to charge the battery inside the detector.

[0100] When the wireless charging unit detects a valid wireless charging signal, it automatically switches to wireless charging mode and cuts off the battery power supply. When the wireless charging unit ends wireless charging or the charging signal disappears, the wireless charging unit automatically switches back to battery power supply mode.

[0101] Among them, the receiving coil includes a high-sensitivity receiving coil, a rectifier circuit, a voltage stabilizing circuit and an intelligent charging management chip. The receiving coil receives charging signals within the standard range.

[0102] The voice prompt unit is used to generate operation voice guidance corresponding to each working state for all working states of the equipment. At the same time, the oil sample type, concentration and probability distribution obtained based on the analysis are compared with the oil sample type-concentration-probability distribution threshold mapping table. If the concentration of any oil sample type is greater than its threshold, a voice alarm information is issued, where the voice alarm information includes alarm voice and warning sound.

[0103] The transmission unit is used for transmission via wireless communication. The detector uploads real-time monitoring data, equipment working status, voice alarm information and the concentration change curve of each oil sample to the cloud server for storage; users view the equipment working status and detection results of the detector based on the cloud server through remote devices.

[0104] In order to further optimize the above embodiment, the data recording module includes:

[0105] a change curve generating unit, which is used to periodically extract the concentration data and probability distribution data of each oil sample type in the water sample to obtain a data set for each oil sample type;

[0106] Generate a change curve chart of each oil sample based on the data set of each oil sample type, with the time node as the horizontal axis and the concentration and probability distribution as the vertical axis;

[0107] The storage unit is used to automatically store the oil sample change curve diagrams of each cycle.

[0108] Furthermore, it also includes a display module, which is connected to the storage unit and is used to display the oil sample change curve of each cycle to the manager.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An oil-in-water spectrum analysis and data transmission system based on AI technology, characterized in that: include: Spectral detection module, AI processing module, data recording module, auxiliary function module; The spectrum detection module is used to emit multi-wavelength light to the water sample and collect light signals; The AI ​​processing module is connected to the spectral detection module and is used to perform spectral analysis based on the optical signal, using an association algorithm group including a light absorption-scattering coupling model, a self-attention feature enhancement model, and a multimodal classification prediction model to output oil concentration and distribution information; The data recording module is connected to the AI ​​processing module and is used to store detection data; The auxiliary function module is connected to the data recording module and is used to transmit detection data, issue voice prompts and power the system; wherein, the spectrum detection module includes: a scattering correction unit, which is used to calculate the scattering coefficient to correct the interference of light signals at different wavelengths caused by scattering of particulate matter in the water sample; the scattering coefficient The calculation formula is ,in is the average oil droplet volume, measured based on a laser particle size analyzer. is the relative refractive index of oil and water, obtained based on the oil parameter table, is the scattering angle, is the design angle of the scattered light collection channel in the spectral sensor unit, is the wavelength of incident light; the AI ​​processing module includes: The feature extraction unit is used to establish the light absorption-scattering coupling model and derive the absorption coefficient of oil substances at different wavelengths. The light absorption-scattering coupling model is: ； in is the transmitted light intensity, is the incident light intensity, is the absorption coefficient of the oil substance being sought, is the scattering coefficient determined by the scattering correction unit; is the optical path length; is the absorbance spectrum; the absorbance at different wavelengths is organized into spectral feature vectors ,in is the number of wavelength points; A prediction unit that calculates feature weights using the self-attention feature enhancement model ,in is the query representation of the spectral feature vector, used to identify the key wavelengths, is the key representation of the spectral feature vector, used to match important features, The original value of the spectral feature vector, used for weighted aggregation, The type and concentration analysis unit is connected to the feature extraction unit and is based on the absorption coefficient of the oil substance. Determine the type of oil in the current water sample based on the oil parameter table; Combined with the Lambert-Beer law, the concentration of the oil sample in the current water sample is obtained by reverse calculation; The probability distribution analysis unit is connected to the feature extraction unit and the prediction unit to convert the spectral feature vector and feature weights Substitute into the multimodal classification prediction model to determine the probability distribution of various oil samples in the water sample; the multimodal classification prediction model is: ； in, is the probability distribution of the i-th type of oil, is the weight parameter of category i for feature j, determined based on training; Weighting attention The jth wavelength characteristic after The feature weight The result of normalization; Determine the probability distribution of various oil samples through the multimodal classification prediction model .

2. The oil-in-water spectrum analysis and data transmission system based on AI technology according to claim 1 is characterized in that: The spectrum detection module also includes: A light source unit, which is used to emit continuous spectrum light of preset different wavelength combinations to provide an excitation light source for detecting the spectral characteristics of oil substances in water samples; The sample pool unit adopts a flow-through design and has a built-in stirring device to homogenize the water sample, and is used to contain the water sample and allow light to pass through; The spectral sensor unit includes a cooled CCD array for collecting incident light intensity and transmitted light intensity.

3. The oil-in-water spectrum analysis and data transmission system based on AI technology according to claim 2 is characterized in that: The self-attention feature enhancement model is: ; represent and Multiply the transpose of ; The dimensions of , is the feature dimension; By formula The transmitted light intensity is converted into absorbance; the self-attention feature enhancement model realizes feature weight distribution through matrix operation, calculates the importance weights of features of different wavelengths, and makes the model target the wavelength corresponding to the characteristic peak of oil.

4. The oil-in-water spectrum analysis and data transmission system based on AI technology according to claim 1 is characterized in that: The auxiliary function module includes a wireless charging unit, a voice prompt unit, and a transmission unit.

5. The oil-in-water spectrum analysis and data transmission system based on AI technology according to claim 4 is characterized in that: The wireless charging unit is based on the wireless charging standard. The wireless charger generates an alternating magnetic field through an alternating current. After the receiving coil at the detector senses the alternating magnetic field, it generates an induced electromotive force to charge the battery inside the detector. When the wireless charging unit detects a valid wireless charging signal, it automatically switches to wireless charging mode and cuts off the battery power supply. When the wireless charging unit ends wireless charging or the charging signal disappears, the wireless charging unit automatically switches back to battery power supply mode. The receiving coil includes a high-sensitivity receiving coil, a rectifier circuit, a voltage stabilizing circuit and an intelligent charging management chip, and the receiving coil receives charging signals within a standard range.

6. The oil-in-water spectrum analysis and data transmission system based on AI technology according to claim 5 is characterized in that: The voice prompt unit is used to generate operation voice guidance corresponding to each working state for all working states of the equipment, and at the same time compare the oil sample type, concentration and probability distribution obtained based on the analysis with the oil sample type-concentration-probability distribution threshold mapping table. If the concentration of any oil sample type is greater than its threshold, a voice alarm information is issued, wherein the voice alarm information includes alarm voice and warning sound.

7. The oil-in-water spectrum analysis and data transmission system based on AI technology according to claim 6 is characterized in that: The transmission unit is used to transmit via wireless communication, and the detector uploads real-time monitoring data, equipment working status, voice alarm information and each oil sample concentration change curve to the cloud server for storage; the user views the detector's equipment working status and detection results based on the cloud server through a remote device.

8. The oil-in-water spectrum analysis and data transmission system based on AI technology according to claim 7 is characterized in that: The data recording module includes: a change curve generating unit, which is used to periodically extract the concentration data and probability distribution data of each oil sample type in the water sample to obtain a data set for each oil sample type; Generate a change curve chart of each oil sample based on the data set of each oil sample type, with the time node as the horizontal axis and the concentration and probability distribution as the vertical axis; The storage unit is used to automatically store the oil sample change curve diagrams of each cycle.

9. The oil-in-water spectrum analysis and data transmission system based on AI technology according to claim 8, characterized in that: It also includes a display module, which is connected to the storage unit and is used to display the oil sample change curve diagram of each cycle to the manager.

Images