A coal ash online detection method and system based on multi-modal data fusion and digital twin simulation
Patent Information
- Application Number
- CN202611095845.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-22
- Publication Date
- 2026-10-27
AI Technical Summary
[0003]然而,现有的无源测灰方法中,直接采用单一探测器进行伽马计数,可能会导致信噪比低、抗干扰能力差,或者模型僵化、无法自适应煤种变化,从而影响灰分检测的精度与稳定性,难以满足现代工业对高精度闭环控制的需求
[0011]本发明实施例的基于多模态数据融合与数字孪生仿真的煤炭灰分在线检测方法及系统,采集煤流的多模态传感数据,多模态传感数据包括天然伽马射线能谱数据、煤流厚度数据和煤流密度数据;基于煤流厚度数据和煤流密度数据,对天然伽马射线能谱数据进行物理参数校正,生成标准化伽马能谱数据;基于数字孪生仿真模型,当前煤流厚度和煤流密度生成理论仿真能谱,并将标准化伽马能谱数据与理论仿真能谱进行残差分析,基于得到的残差分析结果对标准化伽马能谱数据进行修正,得到修正后的标准化伽马能谱数据;提取修正后的标准化伽马能谱数据中特征核素的特征峰信息,并将特征峰信息与煤流厚度数据、煤流密度数据输入至目标机器学习模型中,得到灰分值。本发明通过多模态数据融合与数字孪生仿真,提高了灰分检测的精度与稳定性,实现了高精度、高稳定性、自适应煤种变化的在线灰分检测。
Smart Images

Figure CN122983340A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of coal ash content detection technology, and in particular to an online coal ash content detection method and system based on multimodal data fusion and digital twin simulation. Background Technology
[0002] Coal, as an important energy resource, has ash content as a key indicator of its quality, and it is widely used in thermal power generation, coal washing, and chemical industries. Among related technologies, a passive ash content measurement system has been constructed through the coordinated operation of natural gamma-ray detection, energy spectrum analysis, and physical parameter correction. Specifically, this technology covers the entire process from coal flow radiation signal acquisition to ash content prediction, including key aspects such as detector arrays, signal processing, and algorithm modeling.
[0003] However, existing passive ash detection methods that directly use a single detector for gamma counting may result in low signal-to-noise ratio, poor anti-interference ability, or rigid models that cannot adapt to changes in coal type, thus affecting the accuracy and stability of ash detection and making it difficult to meet the needs of modern industry for high-precision closed-loop control. Summary of the Invention
[0004] The present invention aims to at least partially solve one of the technical problems in the related art.
[0005] To achieve the above objectives, this invention proposes an online coal ash content detection method based on multimodal data fusion and digital twin simulation, the method comprising: Collect multimodal sensing data of coal flow, including natural gamma-ray energy spectrum data, coal flow thickness data, and coal flow density data; Based on the coal flow thickness data and the coal flow density data, physical parameter correction is performed on the natural gamma-ray energy spectrum data to generate standardized gamma-ray energy spectrum data. Based on the digital twin simulation model, the current coal flow thickness and coal flow density are used to generate a theoretical simulation energy spectrum. The standardized gamma energy spectrum data and the theoretical simulation energy spectrum are then subjected to residual analysis. Based on the obtained residual analysis results, the standardized gamma energy spectrum data is corrected to obtain the corrected standardized gamma energy spectrum data. The characteristic peak information of the characteristic nuclides in the corrected standardized gamma spectrum data is extracted, and the characteristic peak information, along with the coal flow thickness data and the coal flow density data, is input into the target machine learning model to obtain the ash content value.
[0006] The online coal ash content detection method based on multimodal data fusion and digital twin simulation in this invention embodiment may also have the following additional technical features: In one embodiment of the present invention, the acquisition of multimodal sensing data of coal flow includes: Natural gamma-ray energy spectrum data were acquired using an arc-shaped NaI(Tl) scintillation detector array. Coal flow thickness data are collected using a laser thickness sensor based on the principle of triangulation. Coal flow density data is collected using an electronic belt scale sensor.
[0007] In one embodiment of the present invention, the step of performing physical parameter correction on the natural gamma-ray energy spectrum data based on the coal flow thickness data and the coal flow density data to generate standardized gamma-ray energy spectrum data includes: The gamma count rate in the natural gamma-ray energy spectrum data is compensated for thickness using a thickness correction formula to obtain the thickness-compensated count rate. The thickness correction formula is as follows: , It is a fitting function established based on the attenuation characteristics of gamma rays in coal medium. The count rate in the natural gamma-ray energy spectrum data. This refers to coal flow thickness data; Density compensation is applied to the thickness-compensated count rate based on the density correction formula to generate standardized gamma-ray spectrum data, wherein the density correction formula is: , The count rate after density compensation. This refers to coal flow quality data.
[0008] In one embodiment of the present invention, the theoretical simulation energy spectrum is generated based on the digital twin simulation model using the current coal flow thickness and coal flow density. Residual analysis is then performed between the standardized gamma energy spectrum data and the theoretical simulation energy spectrum. Based on the obtained residual analysis results, the standardized gamma energy spectrum data is corrected to obtain corrected standardized gamma energy spectrum data, including: Constructing a digital twin simulation model based on the Monte Carlo method; Based on the aforementioned digital twin simulation model, the current coal flow thickness and density are simulated in real time, generating a theoretical simulation energy spectrum. ; The standardized gamma spectrum data was calculated using the residual formula. With the theoretically simulated energy spectrum The residuals are used to obtain the residual analysis results, wherein the residual formula is: ; When the residual analysis results If the data exceeds a preset threshold, the corresponding data point is removed or a weighted correction is performed to obtain the corrected standardized gamma spectrum data.
[0009] In one embodiment of the present invention, the step of extracting the characteristic peak information of characteristic nuclides in the corrected normalized gamma-ray spectral data includes: The corrected standardized gamma spectrum data is smoothed and preprocessed to remove background, and then peak-finding algorithms are used to locate peaks. 40 The characteristic peak of K at 1460 keV, 238 The characteristic peak of 1764 keV in the U series and 232 The characteristic peak of 2614 keV in the Th series; Integrate the area of the three characteristic peaks A K A U A Th This was identified as characteristic peak information.
[0010] To achieve the above objectives, another aspect of the present invention proposes an online coal ash content detection system based on multimodal data fusion and digital twin simulation. The system includes a multimodal signal acquisition module, a parameter correction module, an intelligent data processing module, and an integrated mounting frame. The multimodal signal acquisition module is used to acquire multimodal sensing data of coal flow, including natural gamma-ray energy spectrum data, coal flow thickness data, and coal flow density data. The parameter correction module is connected to the multimodal signal acquisition module via a signal bus, and is used to perform physical parameter correction on the natural gamma-ray energy spectrum data based on the coal flow thickness data and the coal flow density data to generate standardized gamma-ray energy spectrum data. The intelligent data processing module is used to generate a theoretical simulation energy spectrum based on the current coal flow thickness and coal flow density using a digital twin simulation model, and to perform residual analysis on the standardized gamma energy spectrum data and the theoretical simulation energy spectrum. Based on the obtained residual analysis results, the standardized gamma energy spectrum data is corrected to obtain the corrected standardized gamma energy spectrum data. The intelligent data processing module is also used to extract the characteristic peak information of the characteristic nuclides in the corrected standardized gamma energy spectrum data, and input the characteristic peak information, the coal flow thickness data, and the coal flow density data into the target machine learning model to obtain the ash content value; The integrated mounting frame is used to support and fix the multimodal signal acquisition module, the parameter correction module, and the intelligent data processing module.
[0011] This invention discloses an online coal ash content detection method and system based on multimodal data fusion and digital twin simulation. The method involves collecting multimodal sensing data of coal flow, including natural gamma-ray energy spectrum data, coal flow thickness data, and coal flow density data. Based on the coal flow thickness and density data, physical parameters of the natural gamma-ray energy spectrum data are corrected to generate standardized gamma-ray energy spectrum data. Using a digital twin simulation model, a theoretical simulation energy spectrum is generated based on the current coal flow thickness and density. Residual analysis is performed between the standardized gamma-ray energy spectrum data and the theoretical simulation energy spectrum. Based on the residual analysis results, the standardized gamma-ray energy spectrum data is corrected to obtain corrected standardized gamma-ray energy spectrum data. Characteristic peak information of characteristic nuclides is extracted from the corrected standardized gamma-ray energy spectrum data. This characteristic peak information, along with the coal flow thickness and density data, is input into a target machine learning model to obtain the ash content value. This invention improves the accuracy and stability of ash content detection through multimodal data fusion and digital twin simulation, achieving high-precision, high-stability, and adaptive online ash content detection for coal type changes.
[0012] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0013] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart illustrating an online coal ash content detection method based on multimodal data fusion and digital twin simulation according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of an online coal ash content detection system based on multimodal data fusion and digital twin simulation according to an embodiment of the present invention. Detailed Implementation
[0014] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0015] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0016] The following description, with reference to the accompanying drawings, describes an online coal ash content detection method and system based on multimodal data fusion and digital twin simulation, according to embodiments of the present invention.
[0017] Figure 1 This is a flowchart illustrating the online coal ash content detection method based on multimodal data fusion and digital twin simulation, according to an embodiment of the present invention.
[0018] like Figure 1 As shown, the method may include the following steps: Step 101: Collect multimodal sensing data of coal flow, including natural gamma-ray energy spectrum data, coal flow thickness data, and coal flow density data.
[0019] In one embodiment of the present invention, when performing online detection of coal ash content, it is necessary to acquire multimodal sensing data to obtain multidimensional raw information characterizing the physical state and nuclear radiation properties of the coal flow, providing a basic data source for subsequent correction, simulation, and prediction. Specifically, multimodal sensing data, including at least natural gamma-ray energy spectrum data, coal flow thickness data, and coal flow density data, can be simultaneously acquired by deploying various types of sensors along the coal flow transport path.
[0020] In one embodiment of the present invention, the aforementioned natural gamma-ray energy spectrum data reflects the energy distribution and count rate of gamma rays emitted by natural radionuclides (such as potassium-40, uranium series, and thorium series) in the coal flow, and is the core basis for ash content analysis; the coal flow thickness data characterizes the geometric dimensions of the coal flow in the direction perpendicular to the conveyor belt, and is used to compensate for the attenuation effect of gamma rays in the coal medium; the coal flow density data reflects the bulk density or mass per unit area of the coal flow, and is used to correct the count rate deviation caused by the looseness or compactness of the coal. Furthermore, in one embodiment of the present invention, the above three types of data need to be collected synchronously in the time dimension to ensure the spatiotemporal consistency of subsequent multimodal fusion.
[0021] In one embodiment of the present invention, the method for acquiring multimodal sensing data of coal flow described above may include the following steps: Step 1011: Natural gamma-ray energy spectrum data are acquired using an arc-shaped NaI(Tl) scintillation detector array; Step 1012: Collect coal flow thickness data using a laser thickness sensor based on the principle of triangulation. Step 1013: Collect coal flow density data using an electronic belt scale sensor.
[0022] In one embodiment of the invention, natural gamma-ray energy spectrum data is acquired by an arc-shaped NaI(Tl) scintillation detector array. This array comprises 3 to 5 cylindrical detectors, mounted in an arc within a shield, covering the left, right, top, and bottom directions along the width of the coal conveyor belt, ensuring comprehensive detection of the entire coal flow cross-section. Each detector converts the received gamma rays into electrical pulse signals, which, after pre-amplification and analog-to-digital conversion, output raw gamma-ray energy spectrum data, presented as spectral lines representing the count rate versus energy distribution.
[0023] In one embodiment of the invention, coal flow thickness data is acquired by a laser thickness sensor installed 300mm to 500mm directly above the coal flow. Utilizing the principle of triangulation, the sensor scans the coal flow surface at a frequency of 100Hz. By measuring the angular difference between the laser emission point and the reflection point, and combining this with a known baseline distance, the thickness value H at each point on the coal flow surface is calculated in real time and output as a continuous thickness data stream. In another embodiment, coal flow density data is acquired by an electronic belt scale sensor installed below the conveyor belt weighing idler. By measuring the vertical pressure on the weighing idler and combining this with the belt speed, the mass per unit area σ, i.e., the coal flow density data, is calculated in real time. In another embodiment, all three sensors are connected to a parameter calibration module via shielded twisted-pair cables or an RS485 bus to ensure synchronous signal acquisition. The acquired natural gamma-ray energy spectrum data Craw, coal flow thickness data H, and coal flow density data σ are output as multimodal sensing data to subsequent steps.
[0024] In one embodiment of the present invention, a curved detector array achieves full coverage detection of the coal flow cross-section, avoiding the measurement blind spots of a single detector; high-frequency scanning by a laser thickness sensor provides accurate coal flow thickness data; and an electronic belt scale sensor provides real-time coal flow density data. Through the coordinated operation of these three components, a high-precision, multi-dimensional raw data foundation is provided for subsequent physical parameter correction and intelligent analysis, improving anti-interference capability and detection accuracy under complex working conditions.
[0025] Step 102: Based on the coal flow thickness data and coal flow density data, perform physical parameter correction on the natural gamma-ray energy spectrum data to generate standardized gamma-ray energy spectrum data.
[0026] In one embodiment of the present invention, after obtaining coal flow thickness data and coal flow density data through the above steps, physical parameter correction can be performed on the natural gamma-ray energy spectrum data based on the coal flow thickness data and coal flow density data to generate standardized gamma energy spectrum data, so as to eliminate the interference of coal flow thickness and density fluctuations on gamma count rate and energy spectrum morphology.
[0027] Specifically, in one embodiment of the present invention, thickness correction formulas and density correction formulas can be constructed based on coal flow thickness data and coal flow density data, respectively, to perform channel-by-channel or global compensation processing on the natural gamma-ray energy spectrum data. The thickness correction formula is established based on the exponential decay characteristics of gamma rays in the coal medium and is used to compensate for differences in ray absorption caused by variations in coal flow thickness. The density correction formula is based on the correlation between coal flow bulk density and ray scattering and absorption and is used to correct count rate deviations caused by loose or compacted coal. Furthermore, through the sequential or parallel correction of the above dual physical parameters, the original energy spectrum data is converted into standardized gamma-ray energy spectrum data independent of the physical state of the coal flow, thereby providing a stable and comparable data foundation for subsequent energy spectrum analysis and ash content prediction.
[0028] In one embodiment of the present invention, the method for generating standardized gamma-ray energy spectrum data by performing physical parameter correction on natural gamma-ray energy spectrum data based on coal flow thickness data and coal flow density data may include the following steps: Step 1021: Based on the thickness correction formula, perform thickness compensation on the gamma count rate in the natural gamma-ray energy spectrum data to obtain the thickness-compensated count rate, where the thickness correction formula is: , It is a fitting function established based on the attenuation characteristics of gamma rays in coal medium. The count rate in natural gamma-ray energy spectrum data. This refers to coal flow thickness data; Step 1022: Perform density compensation on the count rate after thickness compensation based on the density correction formula to generate standardized gamma-ray spectrum data, wherein the density correction formula is: , The count rate after density compensation. This refers to coal flow quality data.
[0029] In one embodiment of the present invention, according to the thickness correction formula Thickness compensation is performed on the gamma count rate in natural gamma-ray energy spectrum data, whereby... It is a fitting function established based on the attenuation characteristics of gamma rays in coal medium. This function is obtained through prior experimental calibration and is used to compensate for the differences in ray absorption caused by variations in coal flow thickness, thereby eliminating the influence of thickness fluctuations on the count rate and outputting the thickness-compensated count rate. .
[0030] In one embodiment of the present invention, according to the density correction formula Density compensation is applied to the count rate after thickness compensation, where, It is a fitting function established based on the looseness or compactness of coal, used to correct for count rate deviations caused by changes in bulk density, and outputs standardized gamma-ray spectrum data. Among them, the That is, standardized gamma-ray spectral data This method eliminates the interference from the two major physical parameters of coal flow thickness and density, providing a stable and reliable input for subsequent energy spectrum feature extraction and ash content prediction. Furthermore, through this dual physical correction, the influence of coal flow condition fluctuations on the gamma-ray detection signal is effectively suppressed, improving the accuracy and stability of ash content detection.
[0031] In one embodiment of the present invention, by introducing real-time physical parameters of coal flow thickness and density for correction, the interference of changes in the physical state of coal flow on the gamma-ray detection signal is fundamentally eliminated, the signal-to-noise ratio and stability of energy spectrum data are improved, a reliable data foundation is laid for subsequent high-precision ash content prediction, and the system's sensitivity to complex working conditions is reduced, thereby enhancing the robustness of the method.
[0032] Step 103: Based on the digital twin simulation model, the current coal flow thickness and coal flow density are used to generate a theoretical simulation energy spectrum. The standardized gamma energy spectrum data and the theoretical simulation energy spectrum are then subjected to residual analysis. Based on the obtained residual analysis results, the standardized gamma energy spectrum data is corrected to obtain the corrected standardized gamma energy spectrum data.
[0033] In one embodiment of the present invention, after obtaining standardized gamma spectrum data through the above steps, a theoretical simulation energy spectrum can be generated based on the current coal flow thickness and coal flow density using a digital twin simulation model. The standardized gamma spectrum data and the theoretical simulation energy spectrum are then subjected to residual analysis. Based on the obtained residual analysis results, the standardized gamma spectrum data is corrected to obtain corrected standardized gamma spectrum data. This eliminates measurement interference caused by factors such as sudden environmental changes, the mixing of large pieces of gangue in the coal flow, or changes in equipment status, ensuring that the energy spectrum data used for subsequent processing has higher accuracy and reliability.
[0034] In one embodiment of the present invention, a virtual model of the interaction between gamma rays and the coal flow medium is performed based on physical principles (such as the Monte Carlo method). This model dynamically generates a theoretically simulated energy spectrum corresponding to the current operating conditions based on real-time acquired coal flow thickness and density data. Furthermore, the standardized gamma energy spectrum data obtained after physical parameter correction is compared with this theoretically simulated energy spectrum, and the residual between the two is calculated. This residual reflects the deviation between the actual measured signal and the ideal physical model. Further, based on the residual analysis results, a correction operation is performed on the standardized gamma energy spectrum data. For example, abnormal data points with residuals exceeding a preset threshold are removed, or the corresponding data is weighted and adjusted. This eliminates measurement interference introduced by factors such as sudden environmental changes, the introduction of large pieces of gangue into the coal flow, or changes in equipment status, ensuring that the energy spectrum data used for subsequent processing has higher accuracy and reliability.
[0035] In one embodiment of the present invention, the above-mentioned method for generating a theoretical simulation energy spectrum based on the current coal flow thickness and coal flow density using a digital twin simulation model, and performing residual analysis on the standardized gamma energy spectrum data and the theoretical simulation energy spectrum, and correcting the standardized gamma energy spectrum data based on the obtained residual analysis results, to obtain the corrected standardized gamma energy spectrum data may include the following steps: Step 1031: Construct a digital twin simulation model based on the Monte Carlo method; Step 1032: Based on the digital twin simulation model, simulate the current coal flow thickness and density in real time, and generate the theoretical simulation energy spectrum. ; Step 1033: Calculate the standardized gamma spectrum data using the residual formula. Compared with theoretical simulation energy spectrum The residuals are used to obtain the residual analysis results, where the residual formula is: ; Step 1034, when the residual analysis results If the data exceeds a preset threshold, the corresponding data point is removed or a weighted correction is performed to obtain the corrected standardized gamma spectrum data.
[0036] In one embodiment of the present invention, a corresponding digital twin simulation model is constructed based on the Monte Carlo method. Specifically, this digital twin simulation model is constructed using the GEANT4 Monte Carlo simulation tool to simulate the physical process of the interaction between gamma rays and coal flow. Furthermore, the current coal flow thickness H and coal flow density σ are read in real time based on the digital twin simulation model. These two parameters are acquired by a laser thickness sensor and an electronic belt scale sensor, respectively, and obtained after parameter calibration processing.
[0037] In one embodiment of the present invention, the digital twin simulation model simulates the theoretical simulation energy spectrum under the current working conditions in virtual space based on the current coal flow thickness H and density σ. This theoretical simulation reflects the gamma-ray energy spectrum distribution produced by naturally occurring radioactive nuclides (such as 40K, 238U, and 232Th) in coal flow under ideal physical conditions. It also calculates standardized gamma-ray energy spectrum data. Compared with theoretical simulation energy spectrum residuals between The residual It characterizes the deviation between the actual measured energy spectrum and the theoretical simulated energy spectrum. This deviation may be caused by abnormal factors such as sudden changes in environmental background radiation, large pieces of gangue mixed in the coal flow, or changes in equipment status.
[0038] In one embodiment of the present invention, the residual Compared with a preset threshold, when the residual When the threshold is exceeded, the current data point is determined to have abnormal interference, and the data point is automatically removed or weighted and corrected to reduce the impact of abnormal data on subsequent ash prediction. The standardized gamma spectrum data after residual analysis correction is output to the subsequent energy spectrum feature extraction step, thereby ensuring that the data input to the machine learning model has higher reliability and accuracy.
[0039] In one embodiment of the present invention, by introducing digital twin simulation and residual analysis mechanism, dynamic verification and real-time correction of measurement data are realized, which enhances the anti-interference ability and data stability of the system under complex working conditions, and provides purer and more reliable input data for subsequent ash content prediction, thereby effectively improving the overall detection accuracy and robustness.
[0040] Step 104: Extract the characteristic peak information of the characteristic nuclides in the corrected standardized gamma spectrum data, and input the characteristic peak information, coal flow thickness data, and coal flow density data into the target machine learning model to obtain the ash content value.
[0041] In one embodiment of the present invention, after obtaining the corrected standardized gamma spectrum data through the above steps, the characteristic peak information of the characteristic nuclides in the corrected standardized gamma spectrum data can be extracted, and the characteristic peak information, along with the coal flow thickness data and coal flow density data, can be input into the target machine learning model to obtain the ash content value.
[0042] In one embodiment of the present invention, characteristic peak information of characteristic nuclides is extracted from the corrected standardized gamma-ray energy spectrum data. This allows for the identification and quantification of characteristic peak information corresponding to nuclides intrinsically correlated with coal ash content. This characteristic peak information reflects the distribution characteristics and intensity of different radionuclides in the energy spectrum. Furthermore, the extracted characteristic peak information, along with synchronously acquired coal flow thickness and density data, are used as multidimensional input parameters and sent to a target machine learning model. This model establishes a nonlinear mapping relationship between the input parameters and ash content based on historical data and physical laws, thereby outputting the current ash content value of the coal flow.
[0043] In one embodiment of the present invention, the method for extracting characteristic peak information of characteristic nuclides from the corrected normalized gamma-ray spectrum data may include: performing smoothing and background removal preprocessing on the corrected normalized gamma-ray spectrum data, and locking the peak information using a peak-finding algorithm. 40 The characteristic peak of K at 1460 keV, 238 The characteristic peak of 1764 keV in the U series and 232 The 2614 keV characteristic peak of the Th series, the area integral of the three characteristic peaks A K A U A Th This was identified as characteristic peak information.
[0044] In one embodiment of the present invention, characteristic nuclides are identified from the corrected normalized gamma-ray spectrum data using a peak-finding algorithm. 40 K, 238 U and 232 The characteristic peaks of the Th series are identified, corresponding to energy positions of 1460 keV, 1764 keV, and 2614 keV, respectively. The area integrals of each characteristic peak are calculated to obtain the eigenvectors. Simultaneously, the real-time thickness of the current coal flow is obtained. and mass per unit area , constituting physical parameters Furthermore, the aforementioned feature vectors are combined with physical parameters to form a five-dimensional input vector, which is then sent to a pre-trained target machine learning model to obtain gray values.
[0045] In one embodiment of the present invention, the target machine learning model can be implemented using either a random forest or a backpropagation (BP) neural network. Specifically, when using a random forest model, it consists of an ensemble of multiple decision trees. Each tree splits based on different feature subsets of the input vector, and the mean of the prediction results from all trees is output as the gray prediction value. When using a backpropagation (BP) neural network model, its structure includes an input layer (5 nodes), several hidden layers (e.g., two layers, each with 16 nodes, using ReLU activation function), and an output layer (1 node, linear activation). Predicted values are obtained through forward propagation. In the offline phase, this machine learning model is trained using a large amount of historical data. The training samples include feature vectors of different coal types and operating conditions, along with corresponding laboratory ash content verification values. Model parameters are optimized by minimizing the mean squared error loss function. In the online prediction phase, the model receives the current input in real time and directly outputs the ash content. This enables a high-precision mapping from multimodal sensing data to ash content. Based on this, by fusing energy spectrum characteristics and physical parameters, and fully utilizing natural gamma-ray information and coal flow state information, the prediction model can effectively cope with the nonlinear effects of coal flow fluctuations and coal type changes.
[0046] In one embodiment of the present invention, by jointly inputting the characteristic peak area, coal flow thickness, and density into a random forest or BP neural network model, deep fusion and nonlinear mapping of multi-source information are achieved, which improves the accuracy and robustness of ash content prediction and provides a reliable prediction basis for subsequent self-calibration mechanisms.
[0047] In one embodiment of the present invention, before inputting the characteristic peak information, coal flow thickness data, and coal flow density data into the target machine learning model to obtain the ash content value, the above steps can also perform self-calibration of the machine learning model and the digital twin simulation model based on the true ash content value to obtain the target machine learning model. Specifically, when obtaining the true value of manual analysis... Then, the true value is compared with the gray prediction value output by the machine learning model. The two are compared, and the error between them is calculated, for example, using mean squared error as the loss function. Then, the gradient of the loss function with respect to the weights of each layer of the neural network is calculated using the backpropagation algorithm, and the neural network weight parameters are fine-tuned accordingly to make the model closer to the true value in subsequent predictions. Simultaneously, the physical parameters involved in the fine-tuning process, such as... and The coefficients in the model, as well as the medium attenuation coefficients and geometric parameters used in the Monte Carlo simulation in the digital twin simulation model, are fed back to the digital twin simulation model. The digital twin simulation model updates the parameters of its virtual model based on the new physical parameters, thereby generating a theoretical simulation energy spectrum that is more consistent with the current working conditions in subsequent real-time simulations. Through the above steps, collaborative self-calibration of the machine learning model and the digital twin simulation model is achieved, enabling the ash content prediction accuracy to continuously improve over time without the need for frequent manual calibration.
[0048] In one embodiment of the present invention, high-precision nonlinear prediction of ash content is achieved by jointly inputting energy spectrum feature information and physical parameters into the machine learning model. At the same time, a two-way self-calibration mechanism based on the true value of ash content is introduced, enabling the machine learning model and the digital twin simulation model to evolve together, improving the long-term stability and adaptability under complex working conditions, and avoiding the accuracy decay of traditional static models caused by changes in working conditions.
[0049] This invention discloses an online coal ash content detection method based on multimodal data fusion and digital twin simulation. The method collects multimodal sensing data of coal flow, including natural gamma-ray energy spectrum data, coal flow thickness data, and coal flow density data. Based on the coal flow thickness and density data, the natural gamma-ray energy spectrum data is physically corrected to generate standardized gamma-ray energy spectrum data. Using a digital twin simulation model, a theoretical simulation energy spectrum is generated based on the current coal flow thickness and density. Residual analysis is performed between the standardized gamma-ray energy spectrum data and the theoretical simulation energy spectrum. Based on the residual analysis results, the standardized gamma-ray energy spectrum data is corrected to obtain corrected standardized gamma-ray energy spectrum data. Characteristic peak information of characteristic nuclides in the corrected standardized gamma-ray energy spectrum data is extracted and input along with the coal flow thickness and density data into a target machine learning model to obtain the ash content value. This invention improves the accuracy and stability of ash content detection through multimodal data fusion and digital twin simulation, achieving high-precision, high-stability, and adaptive online ash content detection for coal type changes.
[0050] Figure 2 This is a schematic diagram of the structure of the online coal ash content detection system based on multimodal data fusion and digital twin simulation according to an embodiment of the present invention.
[0051] like Figure 2 As shown, the system may include a multimodal signal acquisition module, a parameter correction module, an intelligent data processing module, and an integrated mounting rack, wherein: The multimodal signal acquisition module 201 is used to acquire multimodal sensing data of coal flow, including natural gamma-ray energy spectrum data, coal flow thickness data and coal flow density data. The parameter correction module 202 is connected to the multi-mode signal acquisition module via a signal bus. It is used to perform physical parameter correction on natural gamma-ray energy spectrum data based on coal flow thickness data and coal flow density data to generate standardized gamma-ray energy spectrum data. The intelligent data processing module 203 is used to generate a theoretical simulation energy spectrum based on the current coal flow thickness and coal flow density using a digital twin simulation model, and to perform residual analysis on the standardized gamma energy spectrum data and the theoretical simulation energy spectrum. Based on the obtained residual analysis results, the standardized gamma energy spectrum data is corrected to obtain the corrected standardized gamma energy spectrum data. The intelligent data processing module 203 is also used to extract the characteristic peak information of the characteristic nuclides in the corrected standardized gamma energy spectrum data, and input the characteristic peak information, coal flow thickness data, and coal flow density data into the target machine learning model to obtain the ash content value; The integrated mounting rack is used to support and fix the multimodal signal acquisition module, parameter correction module, and intelligent data processing module.
[0052] In one embodiment of the present invention, the multimodal signal acquisition module 201 includes a NaI(Tl) scintillation detector array, a laser thickness sensor, and an electronic belt scale sensor. The NaI(Tl) scintillation detector array comprises 3 to 5 cylindrical detectors arranged in an arc within a shield, covering the width of the belt, used for gamma spectrum detection of the entire coal flow cross-section, outputting the gamma count rate from the natural gamma ray energy spectrum data. The laser thickness sensor is installed directly above the coal flow, using the triangulation principle to scan the coal flow surface at a set frequency, outputting real-time coal flow thickness and density data. The electronic belt scale sensor is installed below the conveyor belt weighing idler, outputting coal flow density data.
[0053] In one embodiment of the present invention, the intelligent data processing module 203 includes an embedded AI chip; the embedded AI chip is specifically used for: The corrected standardized gamma spectrum data is smoothed and preprocessed to remove background, and then peak-finding algorithms are used to locate peaks. 40 The characteristic peak of K at 1460 keV, 238 The characteristic peak of 1764 keV in the U series and 232 The characteristic peak of 2614 keV in the Th series; Integrate the area of the three characteristic peaks A K A U A Th This was identified as characteristic peak information.
[0054] In one embodiment of the present invention, the intelligent data processing module further includes a digital twin simulation unit; the digital twin simulation unit is specifically used for: Constructing a digital twin simulation model based on the Monte Carlo method; Based on a digital twin simulation model, the current coal flow thickness and density are simulated in real time, generating a theoretical simulation energy spectrum. ; Standardized gamma spectrum data were calculated using the residual formula. Compared with theoretical simulation energy spectrum The residuals are used to obtain the residual analysis results, where the residual formula is: ; When the residual analysis results If the data exceeds a preset threshold, the corresponding data point is removed or a weighted correction is performed to obtain the corrected standardized gamma spectrum data.
[0055] In one embodiment of the present invention, the integrated mounting frame is welded with high-strength stainless steel and has a shock-absorbing mounting seat inside. The shock-absorbing mounting seat uses rubber damping or spring shock absorbers to isolate belt vibration. The outer shell of the integrated mounting frame is coated with a nano dustproof and hydrophobic coating and has an integrated temperature control device inside to ensure that electronic components work stably in an environment of -20°C to 60°C.
[0056] This invention discloses an online coal ash content detection system based on multimodal data fusion and digital twin simulation. The system collects multimodal sensing data of coal flow, including natural gamma-ray energy spectrum data, coal flow thickness data, and coal flow density data. Based on the coal flow thickness and density data, the system performs physical parameter correction on the natural gamma-ray energy spectrum data to generate standardized gamma-ray energy spectrum data. Using a digital twin simulation model, the system generates a theoretically simulated energy spectrum based on the current coal flow thickness and density. Residual analysis is performed between the standardized gamma-ray energy spectrum data and the theoretically simulated energy spectrum. Based on the residual analysis results, the standardized gamma-ray energy spectrum data is corrected to obtain corrected standardized gamma-ray energy spectrum data. Characteristic peak information of characteristic nuclides is extracted from the corrected standardized gamma-ray energy spectrum data. This characteristic peak information, along with the coal flow thickness and density data, is input into a target machine learning model to obtain the ash content value. This invention improves the accuracy and stability of ash content detection through multimodal data fusion and digital twin simulation, achieving high-precision, high-stability, and adaptive online ash content detection for coal type changes.
[0057] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0058] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.
Claims
1. A method for online detection of coal ash content based on multimodal data fusion and digital twin simulation, characterized in that, The method includes: Collect multimodal sensing data of coal flow, including natural gamma-ray energy spectrum data, coal flow thickness data, and coal flow density data; Based on the coal flow thickness data and the coal flow density data, physical parameter correction is performed on the natural gamma-ray energy spectrum data to generate standardized gamma-ray energy spectrum data. Based on the digital twin simulation model, the current coal flow thickness and coal flow density are used to generate a theoretical simulation energy spectrum. The standardized gamma energy spectrum data and the theoretical simulation energy spectrum are then subjected to residual analysis. Based on the obtained residual analysis results, the standardized gamma energy spectrum data is corrected to obtain the corrected standardized gamma energy spectrum data. The characteristic peak information of the characteristic nuclides in the corrected standardized gamma spectrum data is extracted, and the characteristic peak information, along with the coal flow thickness data and the coal flow density data, is input into the target machine learning model to obtain the ash content value.
2. The method according to claim 1, characterized in that, The multimodal sensing data of the collected coal flow includes: Natural gamma-ray energy spectrum data were acquired using an arc-shaped NaI(Tl) scintillation detector array. Coal flow thickness data are collected using a laser thickness sensor based on the principle of triangulation. Coal flow density data is collected using an electronic belt scale sensor.
3. The method according to claim 1, characterized in that, The step of performing physical parameter correction on the natural gamma-ray energy spectrum data based on the coal flow thickness data and the coal flow density data to generate standardized gamma-ray energy spectrum data includes: The gamma count rate in the natural gamma-ray energy spectrum data is compensated for thickness using a thickness correction formula to obtain the thickness-compensated count rate. The thickness correction formula is as follows: , It is a fitting function established based on the attenuation characteristics of gamma rays in coal medium. The count rate in the natural gamma-ray energy spectrum data. This refers to coal flow thickness data; Density compensation is applied to the thickness-compensated count rate based on the density correction formula to generate standardized gamma-ray spectrum data, wherein the density correction formula is: , The count rate after density compensation. This refers to coal flow quality data.
4. The method according to claim 1, characterized in that, The method uses a digital twin simulation model to generate a theoretical simulation energy spectrum based on the current coal flow thickness and density. Residual analysis is then performed between the standardized gamma energy spectrum data and the theoretical simulation energy spectrum. Based on the obtained residual analysis results, the standardized gamma energy spectrum data is corrected to obtain corrected standardized gamma energy spectrum data, including: Constructing a digital twin simulation model based on the Monte Carlo method; Based on the aforementioned digital twin simulation model, the current coal flow thickness and density are simulated in real time, generating a theoretical simulation energy spectrum. ; The standardized gamma spectrum data was calculated using the residual formula. With the theoretically simulated energy spectrum The residuals are used to obtain the residual analysis results, wherein the residual formula is: ; When the residual analysis results If the data exceeds a preset threshold, the corresponding data point is removed or a weighted correction is performed to obtain the corrected standardized gamma spectrum data.
5. The method according to claim 1, characterized in that, The extraction of characteristic peak information of characteristic nuclides from the corrected normalized gamma-ray spectral data includes: The corrected standardized gamma spectrum data is smoothed and preprocessed to remove background, and then peak-finding algorithms are used to locate peaks. 40 The characteristic peak of K at 1460 keV, 238 The characteristic peak of 1764 keV in the U series and 232 The characteristic peak of 2614 keV in the Th series; Integrate the area of the three characteristic peaks A K A U A Th This was identified as characteristic peak information.
6. An online coal ash content detection system based on multimodal data fusion and digital twin simulation, characterized in that, The system includes a multimodal signal acquisition module, a parameter correction module, an intelligent data processing module, and an integrated mounting rack; The multimodal signal acquisition module is used to acquire multimodal sensing data of coal flow, including natural gamma-ray energy spectrum data, coal flow thickness data, and coal flow density data. The parameter correction module is connected to the multimodal signal acquisition module via a signal bus, and is used to perform physical parameter correction on the natural gamma-ray energy spectrum data based on the coal flow thickness data and the coal flow density data to generate standardized gamma-ray energy spectrum data. The intelligent data processing module is used to generate a theoretical simulation energy spectrum based on the current coal flow thickness and coal flow density using a digital twin simulation model, and to perform residual analysis on the standardized gamma energy spectrum data and the theoretical simulation energy spectrum. Based on the obtained residual analysis results, the standardized gamma energy spectrum data is corrected to obtain the corrected standardized gamma energy spectrum data. The intelligent data processing module is also used to extract the characteristic peak information of the characteristic nuclides in the corrected standardized gamma energy spectrum data, and input the characteristic peak information, the coal flow thickness data, and the coal flow density data into the target machine learning model to obtain the ash content value; The integrated mounting frame is used to support and fix the multimodal signal acquisition module, the parameter correction module, and the intelligent data processing module.
7. The system according to claim 6, characterized in that, The multimodal signal acquisition module includes a NaI(Tl) scintillation detector array, a laser thickness sensor, and an electronic belt scale sensor; The NaI(Tl) scintillation detector array contains 3 to 5 cylindrical detectors, which are installed in an arc shape inside the shielding body, covering the width direction of the belt, and are used to perform gamma energy spectrum detection on the entire cross-section of coal flow, and output the gamma count rate in the natural gamma ray energy spectrum data. The laser thickness sensor is installed directly above the coal flow and uses the triangulation principle to scan the surface of the coal flow at a set frequency, outputting real-time coal flow thickness data and coal flow density data. The electronic belt scale sensor is installed below the weighing idler of the conveyor belt and outputs coal flow density data.
8. The system according to claim 6, characterized in that, The intelligent data processing module includes an embedded AI chip; the embedded AI chip is specifically used for: The corrected standardized gamma spectrum data is smoothed and preprocessed to remove background, and then peak-finding algorithms are used to locate peaks. 40 The characteristic peak of K at 1460 keV, 238 The characteristic peak of 1764 keV in the U series and 232 The characteristic peak of 2614 keV in the Th series; Integrate the area of the three characteristic peaks A K A U A Th This was identified as characteristic peak information.
9. The system according to claim 6, characterized in that, The intelligent data processing module further includes a digital twin simulation unit; the digital twin simulation unit is specifically used for: Constructing a digital twin simulation model based on the Monte Carlo method; Based on the aforementioned digital twin simulation model, the current coal flow thickness and density are simulated in real time, generating a theoretical simulation energy spectrum. ; The standardized gamma spectrum data was calculated using the residual formula. With the theoretically simulated energy spectrum The residuals are used to obtain the residual analysis results, wherein the residual formula is: ; When the residual analysis results If the data exceeds a preset threshold, the corresponding data point is removed or a weighted correction is performed to obtain the corrected standardized gamma spectrum data.
10. The system according to claim 6, characterized in that, The integrated mounting frame is welded from high-strength stainless steel and has internal shock-absorbing mounting bases. The shock-absorbing mounting bases use rubber dampers or spring shock absorbers to isolate belt vibration. The outer shell of the integrated mounting frame is coated with a nano dustproof and hydrophobic coating and has an integrated temperature control device to ensure that electronic components operate stably in an environment ranging from -20℃ to 60℃.
