Industrial big data informatization system for industrial production

Abstract

The invention relates to the technical field of industrial production data optimization, in particular to an industrial big data informatization system for industrial production, which comprises a dust concentration monitoring unit, a physical model building unit, a machine learning algorithm unit and a dynamic compensation mechanism unit, the dust concentration monitoring unit obtains dust concentration, particle size and density information through a laser beam and a photoelectric detector, the physical model building unit builds an illumination intensity correction model and a pressure sensor mechanical model based on the Mie scattering theory, the Lambert-Beer law and the like, and real-time optimization is conducted through the self-adaptive optimization module. After the machine learning algorithm unit preprocesses sensor and dust data, a support vector regression algorithm is utilized to predict data deviation, the signal compensation module adjusts a compensation value in combination with a physical model correction result, and the dynamic compensation mechanism unit corrects an original output signal of the sensor in real time according to a model calculation and machine learning correction result. The industrial production efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial production data optimization, and more specifically, to an industrial big data informatization system for industrial production. Background Art

[0002] Industrial production data optimization is an important technology. In modern industrial production, it widely exists in industrial fields with high dust pollution such as mine exploitation and cement production. These industries rely on various sensors for monitoring the production process, monitoring the equipment status, and controlling the product quality. The data collected by the sensors plays a crucial role in ensuring production safety, improving production efficiency, and product quality.

[0003] However, the high-dust pollution environment poses great challenges to the normal operation of sensors. When exposed to such an environment for a long time, sensors are extremely prone to problems such as decreased sensitivity. At the mine exploitation site, the dust concentration is relatively high. The light sensor may have deviations in measuring the light intensity due to dust coverage, which in turn affects the automatic adjustment of the lighting system. Under the action of dust accumulation, the pressure sensor measures inaccurate equipment pressure data and cannot timely reflect the true operating state of the equipment. These data deviations will lead to inaccurate industrial production data, thus misleading production decisions, causing waste of production resources, and reducing production efficiency. To solve this technical problem, we provide an industrial big data informatization system for industrial production. Summary of the Invention

[0004] The purpose of the present invention is to provide an industrial big data informatization system for industrial production to solve the problems raised in the above background art.

[0005] To achieve the above purpose, an industrial big data informatization system for industrial production is provided, including a dust concentration monitoring unit, a physical model construction unit, a machine learning algorithm unit, and a dynamic compensation mechanism unit;

[0006] The dust concentration monitoring unit emits a laser beam with a specific wavelength and intensity, and uses a photodetector to obtain the scattered light intensity and angle of the laser beam, and then calculates the concentration, particle size, and density information of the dust around the light sensor and the pressure sensor in combination with the built-in algorithm;

[0007] The physical model construction unit establishes an illumination intensity correction model based on Mie scattering theory and Lambert-Beer law. The illumination intensity correction model analyzes the optical properties of dust particles, calculates the attenuation coefficient of light in the dust environment, and uses the attenuation coefficient to correct the original data. Based on the principles of fluid mechanics and solid mechanics, a mechanical model of the pressure sensor in the dust environment is established. The mechanical model takes into account the force exerted by dust accumulation on the diaphragm of the pressure sensor and the influence of the friction between dust particles on pressure conduction. Through simulation and calculation, the correction coefficient of the pressure sensor is obtained;

[0008] The machine learning algorithm unit preprocesses the collected sensor data and dust concentration data, selects a machine learning algorithm to train the preprocessed data, and uses the trained model to predict the deviation of the sensor data in real time. Combining with the correction result of the built-in physical model, the compensation value of the sensor output signal is dynamically adjusted;

[0009] The dynamic compensation mechanism unit first calculates the theoretical deviation value of the sensor output signal in the current dust environment according to the physical model, then uses the machine learning algorithm to correct the calculation result of the physical model to obtain the deviation prediction value, and finally uses the deviation prediction value as the compensation value to correct the original output signal of the sensor in real time.

[0010] As a further improvement of this technical solution, the specific method for the dust concentration monitoring unit to obtain the concentration, particle size and density information of the dust around the light sensor and the pressure sensor is as follows:

[0011] Emit multiple laser beams with specific wavelengths and intensities at different angles into the dust area, use the photodetectors at the corresponding angles to obtain the scattered light intensity and the scattering angle, and establish a relationship model between the scattered light intensity and the dust particle characteristics according to Mie scattering theory. For monodisperse spherical particles, the Mie scattering coefficient is a function of the particle size parameter and the relative refractive index of the particle;

[0012] By measuring the scattered light intensity at multiple scattering angles, a system of equations is constructed, and the system of equations is solved using the nonlinear least squares method to obtain the particle radius and particle concentration. Then, according to the particle radius and particle concentration, the dust concentration is calculated. Combining with the measurement results of the scattered light intensity of lasers with different wavelengths, the inversion algorithm is used to optimize the calculation of the particle density. By continuously iteratively adjusting the value of the dust particle density, the minimum error between the theoretical scattered light intensity and the actual measured value is obtained.

[0013] As a further improvement of this technical solution, when the physical model construction unit establishes the illumination intensity correction model, the dust particle analysis module uses the following algorithm:

[0014] Obtain the expression of the attenuation relationship between the light intensity and the propagation distance according to the Lambert-Beer law, and obtain the expression of the attenuation coefficient and the optical properties of dust particles in combination with the Mie scattering theory. Considering the non-uniformity of particles in the actual dust environment, introduce a correction factor, obtain the corrected attenuation coefficient according to the correction factor, and measure the light intensity attenuation under different dust concentrations and particle characteristics. Use the least squares method to fit the relationship function between the correction factor and the dust concentration and particle size;

[0015] Use the calculated corrected attenuation coefficient to correct the original light intensity data collected by the light sensor to obtain the corrected light intensity.

[0016] As a further improvement of this technical solution, when the physical model construction unit establishes the mechanical model of the pressure sensor, the mechanical model module adopts the following method:

[0017] Analyze the force exerted by dust accumulation on the diaphragm of the pressure sensor. Regard the dust accumulation as a multi-layer particle accumulation body. According to the principles of particle mechanics, calculate the vertical pressure of each layer of particles on the next layer of particles and the diaphragm;

[0018] Consider the influence of the friction force between dust particles on pressure conduction, set a friction coefficient between particles, obtain the friction force value between adjacent layers of particles according to the friction coefficient between particles and the vertical pressure, and establish the mechanical equilibrium equation of the pressure sensor diaphragm. At the same time, considering the action of the vertical pressure and friction force of dust accumulation, obtain the relationship expression between the actual measured pressure and the true pressure of the pressure sensor through finite element simulation;

[0019] Establish the function relationship expression of the pressure deviation and the dust concentration and particle size through numerical simulation, and calculate the correction coefficient of the pressure sensor according to this function relationship.

[0020] As a further improvement of this technical solution, the physical model construction unit further includes an adaptive optimization module. When the adaptive optimization module performs real-time adaptive optimization on the light intensity correction model and the pressure sensor mechanical model, it adopts the following algorithm:

[0021] Define the model parameter vector. For the light intensity correction model, the model parameter vector includes the correction factor. For the pressure sensor mechanical model, the model parameter vector includes the friction coefficient and the elastic coefficient;

[0022] Adopt the Kalman filtering algorithm to update the model parameters online. According to the parameter estimation value at the previous moment and the dynamic model of the system, predict the parameter value at the current moment. At the same time, predict the covariance matrix, and then calculate the Kalman gain according to the measurement value and the measurement model at the current moment, and then update the parameter estimation value and update the covariance matrix.

[0023] As a further improvement of this technical solution, the machine learning algorithm unit predicts the sensor data deviation through a deviation prediction module:

[0024] Preprocess the collected sensor data and dust concentration data, introduce the support vector regression algorithm to train the preprocessed data. The goal of the support vector regression algorithm is to find an optimal hyperplane that minimizes the error between the predicted value and the true value within a preset range;

[0025] Use the training data set to train the support vector regression model, determine the parameters of the model by minimizing the objective function, introduce a penalty factor to control the balance between the complexity of the model and the error tolerance, and use the trained support vector regression model to predict the real-time sensor data to obtain the deviation prediction value of the sensor data.

[0026] As a further improvement of this technical solution, the machine learning algorithm unit further includes a signal compensation module. When the signal compensation module dynamically adjusts the compensation value of the sensor output signal in combination with the correction result of the built-in physical model, the following method is adopted:

[0027] Respectively set the sensor data correction value calculated by the physical model and the sensor data deviation predicted by the machine learning algorithm, and define a weight coefficient to balance the influence of the physical model correction result and the machine learning prediction result. Calculate the compensation value according to the weight coefficient, data correction value, and data deviation;

[0028] Use the calculated compensation value to correct the original sensor output signal to obtain the corrected sensor output signal.

[0029] As a further improvement of this technical solution, the dynamic compensation mechanism unit adopts the following method to perform real-time correction on the original sensor output signal:

[0030] Calculate the theoretical deviation value of the sensor output signal in the current dust environment according to the light intensity correction model and the pressure sensor mechanical model, and use the machine learning algorithm to correct the calculation results of the light intensity correction model and the pressure sensor mechanical model to obtain the corrected deviation prediction value;

[0031] Use the corrected deviation prediction value as the compensation value to perform real-time correction on the original sensor output signal to obtain the corrected sensor output signal.

[0032] Compared with the prior art, the beneficial effects of the present invention:

[0033] In an industrial big data informatization system for industrial production, the dust concentration monitoring unit accurately obtains dust-related information, providing a basis for subsequent calibration. The physical model construction unit uses professional theories and algorithms to establish a model of the light and pressure sensors, calculates the calibration coefficient considering the influence of various factors, and adjusts the model parameters in real time through an adaptive optimization module to adapt to environmental changes. The machine learning algorithm unit is trained through data preprocessing and the support vector regression algorithm to accurately predict the deviation of sensor data. The signal compensation module fuses the physical model and the machine learning results to optimize the compensation value. The dynamic compensation mechanism unit combines the two to calculate and correct the deviation prediction value, and corrects the original output signal of the sensor in real time, providing accurate data for industrial production and ensuring production safety and improving efficiency. Brief Description of the Drawings

[0034] Figure 1 It is the overall block diagram of the present invention.

[0035] The meanings of the various reference numerals in the figure are as follows:

[0036] 1. Dust concentration monitoring unit; 2. Physical model construction unit; 21. Dust particle analysis module; 22. Mechanical model module; 23. Adaptive optimization module; 3. Machine learning algorithm unit; 31. Deviation prediction module; 32. Signal compensation module; 4. Dynamic compensation mechanism unit. Detailed Embodiments

[0037] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0038] The present invention provides an industrial big data informatization system for industrial production. Please refer to Figure 1 As shown, it includes a dust concentration monitoring unit 1, a physical model construction unit 2, a machine learning algorithm unit 3, and a dynamic compensation mechanism unit 4;

[0039] The dust concentration monitoring unit 1 emits a laser beam with a specific wavelength and intensity, and uses a photodetector to obtain the scattered light intensity and angle of the laser beam, and then calculates the concentration, particle size, and density information of the dust around the light sensor and the pressure sensor in combination with the built-in algorithm.

[0040] The specific method for the dust concentration monitoring unit 1 to obtain the concentration, particle size, and density information of the dust around the light sensor and the pressure sensor is as follows:

[0041] Laser beams with specific wavelengths and intensities are emitted into the dust area at multiple different angles, and the scattered light intensity and scattering angle are obtained using photodetectors corresponding to the angles. Based on Mie scattering theory, a relationship model between the scattered light intensity and the characteristics of dust particles is established. For monodisperse spherical particles, the Mie scattering coefficient is a function of the particle size parameter and the relative refractive index of the particle. The multi-angle laser beam emission and measurement increase the richness and comprehensiveness of the data. The Mie scattering theory model accurately describes the interaction between light and particles, providing a reliable theoretical basis for accurately calculating the characteristics of dust particles.

[0042] By measuring the scattered light intensity at multiple scattering angles, a system of equations is constructed, and the system of equations is solved using the nonlinear least squares method to obtain the particle radius and particle concentration. Suppose scattered light intensities at different scattering angles are measured. According to , equations can be constructed, where is the particle concentration, is the volume of a single particle, is the Mie scattering coefficient, is the incident laser intensity, is the particle size parameter, is the relative refractive index of the particle.

[0043] The goal of the nonlinear least squares method is to minimize the error function ; where is the actually measured scattered light intensity, is the scattered light intensity calculated according to the theoretical model. By using an iterative optimization method to solve, the values of that minimize and are obtained, providing basic data for subsequent calculation of dust concentration and making the calculation of dust concentration more accurate.

[0044] Then, based on the particle radius and the number concentration of particles, the dust concentration is calculated, providing an important indicator for dust monitoring in the industrial production environment. There may be certain errors in the pre-determined dust particle density, which cannot accurately reflect the actual dust situation. Combining the measurement results of the scattered light intensity of lasers with different wavelengths and using an inversion algorithm to optimize the calculation of particle density. By continuously iteratively adjusting the value of the dust particle density, the minimum error between the theoretical scattered light intensity and the actual measured value is obtained, improving the accuracy of particle density calculation and thus enhancing the overall accuracy of dust concentration monitoring.

[0045] The physical model construction unit 2 establishes an illumination intensity correction model based on the Mie scattering theory and the Lambert-Beer law. The illumination intensity correction model analyzes the optical characteristics of dust particles, calculates the attenuation coefficient of light in the dust environment, and uses the attenuation coefficient to correct the original data. Based on the principles of fluid mechanics and solid mechanics, a mechanical model of the pressure sensor in the dust environment is established. The mechanical model considers the force exerted by dust accumulation on the diaphragm of the pressure sensor and the influence of the friction between dust particles on pressure conduction. Through simulation and calculation, the correction coefficient of the pressure sensor is obtained.

[0046] When the physical model construction unit 2 establishes the illumination intensity correction model, the dust particle analysis module 21 adopts the following algorithm:

[0047] The Lambert-Beer law can describe the attenuation law of illumination intensity with propagation distance in a homogeneous medium, providing a basic framework for establishing the illumination intensity attenuation model. The Mie scattering theory can accurately describe the interaction between light and particles, and through it, the relationship between the attenuation coefficient and the optical characteristics of dust particles can be obtained. The combination of the two can provide a theoretical basis for accurately calculating the attenuation of illumination intensity subsequently.

[0048] According to the Lambert-Beer law, the expression of the attenuation relationship of illumination intensity with propagation distance is obtained. Combining with the Mie scattering theory, the attenuation coefficient is related to the optical characteristics of dust particles. The combination of the Lambert-Beer law and the Mie scattering theory fully considers the basic law of light propagation and the scattering effect of dust particles on light, making the established illumination intensity attenuation model more accurate and scientific.

[0049] Considering the non-uniformity of particles in the actual dust environment, a correction factor is introduced. The corrected attenuation coefficient is obtained according to the correction factor. The introduction of the correction factor increases the flexibility and adaptability of the model, can better cope with the complexity of the actual dust environment, and improves the applicability of the model to different working conditions.

[0050] The original illumination intensity data collected by the illumination sensor is affected by dust and cannot accurately reflect the real illumination situation. Using the calculated corrected attenuation coefficient to correct the original data can eliminate the interference of dust on the measurement of illumination intensity and obtain a more accurate illumination intensity value, providing reliable data support for illumination-related decisions in subsequent industrial production processes.

[0051] Let the original illumination intensity data collected by the illumination sensor be , the propagation distance be , then the corrected illumination intensity ; where is the corrected attenuation coefficient. Through calibration, it can effectively eliminate the influence of dust on the measurement of light intensity, making the obtained light intensity value closer to the actual situation and improving the quality and reliability of light data.

[0052] When the physical model construction unit 2 establishes the mechanical model of the pressure sensor, the mechanical model module 22 adopts the following method:

[0053] In a dusty environment, the measurement accuracy of the pressure sensor will be significantly affected by dust accumulation. Analyze the force exerted by dust accumulation on the diaphragm of the pressure sensor. Consider dust accumulation as a multi-layer particle accumulation body. According to the principle of particle mechanics, calculate the vertical pressure of each layer of particles on the next layer of particles and the diaphragm. For the layer of particles, its vertical pressure ; where is the mass of the layer of particles, is the acceleration due to gravity, laying a foundation for subsequent consideration of other factors such as friction and establishing a complete mechanical model, and helping to improve the accuracy of pressure sensor calibration.

[0054] Merely considering the vertical pressure of dust accumulation is not sufficient to comprehensively reflect the influence of dust on the pressure sensor. The friction between dust particles will hinder the transmission of pressure, changing the distribution of pressure on the sensor diaphragm. Set a coefficient of friction between particles, obtain the friction force value between adjacent layers of particles based on the coefficient of friction between particles and the vertical pressure, and establish a mechanical equilibrium equation for the pressure sensor diaphragm.

[0055] Let the friction force between adjacent layers of particles be ; where is the coefficient of friction between particles. Let the total pressure received by the pressure sensor diaphragm be , the elastic modulus of the diaphragm be , the deformation of the diaphragm be , according to Hooke's law , is the elastic coefficient of the diaphragm. Considering the action of the vertical pressure and friction force of dust accumulation, through theoretical derivation, the relationship between the actual measured pressure of the pressure sensor and the true pressure can be obtained, ; where is the pressure deviation caused by dust influence, ; which provides a key basis for the calculation of subsequent correction coefficients and can effectively improve the measurement accuracy of the pressure sensor in a dusty environment.

[0056] To effectively correct the measured values of the pressure sensor, it is necessary to clarify the quantitative relationship between the pressure deviation and key parameters such as dust concentration and particle size. Then, based on this relationship, the correction coefficient is calculated. The functional relationship expression between the pressure deviation and dust concentration and particle size is established through numerical simulation, and the correction coefficient of the pressure sensor is calculated according to this functional relationship, as follows:

[0057] Input different dust concentrations and particle sizes , and simulate to obtain the corresponding pressure deviation . Assume that the functional relationship between the pressure deviation and dust concentration and particle size obtained through simulation fitting is . According to the relationship between the pressure deviation, true pressure, and measured pressure, the correction coefficient ; Substitute to obtain the specific functional relationship between the correction coefficient and , , which is convenient for calculating the correction coefficient in real time according to the dust parameters during actual measurement, enabling the pressure sensor to automatically correct the measured value according to the real-time dust concentration and particle size, greatly improving the accuracy and stability of pressure measurement in a dust environment.

[0058] The physical model construction unit 2 also includes an adaptive optimization module 23. The following algorithm is used when the adaptive optimization module 23 performs real-time adaptive optimization on the light intensity correction model and the pressure sensor mechanical model:

[0059] When performing real-time adaptive optimization on the light intensity correction model and the pressure sensor mechanical model, it is necessary to clarify which parameters are variable and need to be optimized, and define the model parameter vector. For the light intensity correction model, the model parameter vector includes the correction factor. For the pressure sensor mechanical model, the model parameter vector includes the friction coefficient and the elastic coefficient.

[0060] Assume that the parameter vector of the light intensity correction model is , where is the correction factor. Assume that the parameter vector of the pressure sensor mechanical model is , where is the friction coefficient, is the elastic coefficient, which provides clear data input for the subsequent Kalman filter algorithm, helping to accurately update the model parameters online, thereby improving the adaptability and accuracy of the model.

[0061] The dust situation in the industrial production environment is dynamically changing, and the model parameters will also change accordingly. The Kalman filter algorithm is used to update the model parameters online. According to the parameter estimation value at the previous moment and the dynamic model of the system, the parameter value at the current moment is predicted, and at the same time, the prediction covariance matrix.

[0062] Let the model parameter vector be , for the light intensity correction model , for the mechanical model of the pressure sensor , the dynamic model of the system is assumed to be linear, i.e., ; where is the state transition matrix, is the parameter estimate value at the previous moment, is the predicted parameter value at the current moment, and the covariance matrix represents the uncertainty of the parameter estimate, and the predicted covariance matrix , where is the covariance matrix at the previous moment, is the process noise covariance matrix, which reflects the uncertainty of the system dynamic model, enables the model to better track the dynamic changes of the parameters, and can adjust the model parameters in time when the environment changes, maintaining the accuracy and adaptability of the model.

[0063] The predicted parameter value is only a preliminary estimate and needs to be corrected by combining the actual measurement value at the current moment. Then, according to the measurement value and measurement model at the current moment, the Kalman gain is calculated, and then the parameter estimate value is updated, and the covariance matrix is updated. The measurement model is ; where is the measurement value at the current moment, is the measurement matrix, is the measurement noise, then the Kalman gain ; update the parameter estimate value ; update the covariance matrix ; where is the identity matrix, enabling the light intensity correction model and the mechanical model of the pressure sensor to adapt to the changes of the industrial production environment in real time, accurately correct the light intensity and pressure, and improve the performance and reliability of the entire industrial big data informatization system.

[0064] The machine learning algorithm unit 3 preprocesses the collected sensor data and dust concentration data, selects a machine learning algorithm to train the preprocessed data, and uses the trained model to predict the deviation of the sensor data in real time, and combines the correction result of the built-in physical model to dynamically adjust the compensation value of the sensor output signal.

[0065] The machine learning algorithm unit 3 predicts the sensor data deviation through the deviation prediction module 31:

[0066] Preprocess the collected sensor data and dust concentration data. Normalization can map data in different ranges to the same interval, eliminate the influence of data scale, enable the model to treat each feature more fairly, accelerate the convergence speed of the model, and improve the training efficiency.

[0067] The support vector regression algorithm has good performance in dealing with small sample data and nonlinear problems. Introduce the support vector regression algorithm to train the preprocessed data. The goal of the support vector regression algorithm is to find an optimal hyperplane, and the hyperplane minimizes the error between the predicted value and the true value within a preset range for the given training data set , where is the preprocessed sensor data and dust concentration data, is the corresponding true deviation value. The goal of the support vector regression model is to find a function , where is the weight vector, is the function that maps the input features to a high-dimensional space, is the bias term, in order to more accurately capture the relationship between the sensor data, dust concentration data and the deviation of the sensor data, and improve the accuracy of deviation prediction.

[0068] To enable the support vector regression model to accurately predict the deviation of sensor data, it is necessary to use the training data set to train the model and determine the parameters and of the model. Use the training data set to train the support vector regression model, determine the parameters of the model by minimizing the objective function, and introduce a penalty factor to control the balance between the model complexity and the error tolerance.

[0069] The objective function is ; where is the penalty factor, which is used to control the balance between the model complexity and the error tolerance. By solving the minimum value of the objective function, the optimal and can be obtained. The trained support vector regression model has good generalization performance and can accurately predict the deviation of unknown data.

[0070] The support vector regression model obtained through training has learned the relationship between the sensor data, dust concentration data and the deviation of the sensor data. Using this model to predict the real-time sensor data can obtain the deviation prediction value of the sensor data in a timely manner, providing a basis for subsequent sensor data compensation.

[0071] Substitute the real-time sensor data into the trained model Among them, the deviation prediction value is calculated Based on the trained model for prediction, the deviation prediction value of the sensor data can be obtained quickly and accurately, providing an effective means for the real-time compensation of the sensor data and improving the reliability and accuracy of the sensor data.

[0072] The machine learning algorithm unit 3 further includes a signal compensation module 32. When the signal compensation module 32 dynamically adjusts the compensation value of the sensor output signal in combination with the correction result of the built-in physical model, the following method is adopted:

[0073] In the industrial production environment, the physical model and the machine learning algorithm each have their own advantages and limitations. The correction value of the sensor data calculated by the physical model is set as The deviation of the sensor data predicted by the machine learning algorithm is And a weight coefficient is defined To balance the influence of the correction result of the physical model and the prediction result of the machine learning, so as to improve the reliability of the compensation value and further improve the accuracy of the correction of the sensor output signal.

[0074] In order to obtain a compensation value that can effectively correct the original output signal of the sensor, it is necessary to calculate according to a certain rule based on the set correction value of the physical model, the prediction deviation of the machine learning algorithm and the weight coefficient. The compensation value is calculated according to the weight coefficient, the data correction value and the data deviation The calculation formula is This formula simply and intuitively realizes the fusion of the results of the physical model and the machine learning algorithm, and dynamically adjusts the contribution ratio of the two through the weight coefficient to adapt to the needs of different scenarios.

[0075] Due to the interference of many factors such as the dust environment, the original output signal of the sensor often cannot accurately reflect the true value of the measured quantity. Using the calculated compensation value To the original output signal of the sensor Perform correction to obtain the corrected sensor output signal The calculation formula is The corrected sensor output signal is closer to the true physical quantity, providing solid data support for monitoring and control decisions in industrial production, and ensuring the efficiency and safety of the production process.

[0076] The dynamic compensation mechanism unit 4 first calculates the theoretical deviation value of the sensor output signal under the current dust environment according to the physical model, then uses the machine learning algorithm to correct the calculation result of the physical model to obtain the deviation prediction value, and finally uses the deviation prediction value as the compensation value to correct the original output signal of the sensor in real time.

[0077] The dynamic compensation mechanism unit 4 adopts the following method to perform real-time correction on the original output signal of the sensor:

[0078] In a dust environment, the output signal of the sensor will be affected by dust and deviate. According to the light intensity correction model and the mechanical model of the pressure sensor, calculate the theoretical deviation value of the sensor output signal in the current dust environment, and use machine learning algorithms to correct the calculation results of the light intensity correction model and the mechanical model of the pressure sensor to obtain the corrected deviation prediction value.

[0079] Let the theoretical deviation value of the light sensor calculated by the light intensity correction model be , and the theoretical deviation value of the pressure sensor calculated by the mechanical model of the pressure sensor be . Let the predicted value of the deviation of the light sensor by the machine learning algorithm be , and the predicted value of the deviation of the pressure sensor be . Introduce the correction coefficient . The corrected predicted value of the light sensor deviation ; The corrected predicted value of the pressure sensor deviation ; The obtained corrected deviation prediction value can better reflect the actual sensor deviation situation, providing a more reliable basis for subsequent signal correction and helping to improve the accuracy of sensor measurement.

[0080] The original output signal of the sensor contains deviations caused by factors such as dust, which will affect the data accuracy and decision reliability in the industrial production process. Use the corrected deviation prediction value as a compensation value to correct the original output signal of the sensor in real time to obtain the corrected sensor output signal.

[0081] Let the original output signal of the light sensor be , and the corrected output signal of the light sensor be ; Let the original output signal of the pressure sensor be , and the corrected output signal of the pressure sensor be ; Real-time correction can respond to changes in sensor signals in a timely manner, ensuring the timeliness and accuracy of data. By subtracting the corrected deviation prediction value, it simply and effectively eliminates the influence of factors such as dust on the sensor signal. The corrected sensor output signal more accurately reflects the true physical quantity in the industrial production process, helping to ensure the safe and stable operation of production.

[0082] In the present invention, the dust concentration monitoring unit 1 uses a laser beam and a photodetector to obtain dust concentration, particle size, and density information. The physical model construction unit 2 establishes a light intensity correction model and a pressure sensor mechanical model based on Mie scattering theory and Lambert-Beer's law, and optimizes them in real time through the adaptive optimization module 23. After preprocessing the sensor and dust data by the machine learning algorithm unit 3, the support vector regression algorithm is used to predict data deviation. The signal compensation module 32 adjusts the compensation value in combination with the correction result of the physical model. The dynamic compensation mechanism unit 4 corrects the original output signal of the sensor in real time according to the model calculation and the machine learning correction result, improving the industrial production efficiency.

[0083] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. An industrial big data information system for industrial production, characterized in that: It includes a dust concentration monitoring unit (1), a physical model building unit (2), a machine learning algorithm unit (3) and a dynamic compensation mechanism unit (4); The dust concentration monitoring unit (1) emits a laser beam of a specific wavelength and intensity, and uses a photodetector to obtain the scattered light intensity and angle of the laser beam, and then uses a built-in algorithm to calculate the concentration, particle size and density information of the dust around the light sensor and the pressure sensor; The physical model building unit (2) establishes a light intensity correction model based on Mie scattering theory and Lambert-Beer law. The light intensity correction model calculates the attenuation coefficient of light in a dusty environment by analyzing the optical properties of dust particles. The attenuation coefficient is used to correct the original data. Based on the principles of fluid mechanics and solid mechanics, a mechanical model of the pressure sensor in a dusty environment is established. The mechanical model takes into account the force of dust accumulation on the diaphragm of the pressure sensor and the influence of friction between dust particles on pressure conduction. The correction coefficient of the pressure sensor is obtained through simulation and calculation. The machine learning algorithm unit (3) pre-processes the collected sensor data and dust concentration data, selects a machine learning algorithm to train the pre-processed data, and uses the trained model to predict the deviation of the sensor data in real time, and dynamically adjusts the compensation value of the sensor output signal in combination with the correction result of the built-in physical model; The dynamic compensation mechanism unit (4) first calculates the theoretical deviation value of the sensor output signal under the current dust environment according to the physical model, then uses the machine learning algorithm to correct the physical model calculation result to obtain the deviation prediction value, and finally uses the deviation prediction value as the compensation value to correct the original output signal of the sensor in real time.

2. The industrial big data information system for industrial production according to claim 1, characterized in that: The specific method of the dust concentration monitoring unit (1) obtaining the concentration, particle size and density information of the dust around the light sensor and the pressure sensor is as follows: Multiple laser beams of specific wavelength and intensity are emitted at different angles to the dust area, and the scattered light intensity and scattering angle are obtained using photodetectors at corresponding angles. A relationship model between the scattered light intensity and the characteristics of dust particles is established based on the Mie scattering theory. For monodisperse spherical particles, the Mie scattering coefficient is a function of the particle size parameter and the relative refractive index of the particles. By measuring the scattered light intensity at multiple scattering angles, a group of equations is constructed, and the nonlinear least squares method is used to solve the group of equations to obtain the particle radius and particle concentration. The dust concentration is then calculated based on the particle radius and particle concentration. Combined with the scattered light intensity measurement results of lasers with different wavelengths, the inversion algorithm is used to optimize the calculation of the particle density. The value of the dust particle density is adjusted by continuous iteration to obtain the minimum error between the theoretical scattered light intensity and the actual measured value.

3. The industrial big data information system for industrial production according to claim 2 is characterized in that: When the physical model building unit (2) establishes the light intensity correction model, the dust particle analysis module (21) adopts the following algorithm: The attenuation relationship expression of light intensity and propagation distance is obtained according to the Lambert-Beer law, and the expression of attenuation coefficient and optical characteristics of dust particles is obtained by combining Mie scattering theory. Considering the non-uniformity of particles in the actual dust environment, a correction factor is introduced, and the corrected attenuation coefficient is obtained according to the correction factor. The attenuation of light intensity under different dust concentrations and particle characteristics is measured, and the relationship function between the correction factor and dust concentration and particle size is obtained by fitting using the least squares method. The calculated corrected attenuation coefficient is used to correct the original light intensity data collected by the light sensor to obtain the corrected light intensity.

4. The industrial big data information system for industrial production according to claim 3 is characterized by: When the physical model building unit (2) builds the mechanical model of the pressure sensor, the mechanical model module (22) adopts the following method: Analyze the force of dust accumulation on the pressure sensor diaphragm, regard the dust accumulation as a multi-layer particle accumulation, and calculate the vertical pressure of each layer of particles on the next layer of particles and the diaphragm according to the principle of particle mechanics; Considering the influence of friction between dust particles on pressure conduction, a friction coefficient between particles is set, and the friction value between two adjacent layers of particles is obtained according to the friction coefficient between particles and the vertical pressure, and the mechanical equilibrium equation of the pressure sensor diaphragm is established. At the same time, considering the vertical pressure and friction of dust accumulation, the relationship expression between the actual measured pressure of the pressure sensor and the true pressure is obtained through finite element simulation; The functional relationship expression of pressure deviation, dust concentration and particle size is established through numerical simulation, and the correction coefficient of the pressure sensor is calculated based on the functional relationship.

5. The industrial big data information system for industrial production according to claim 4 is characterized in that: The physical model construction unit (2) also includes an adaptive optimization module (23), and the adaptive optimization module (23) uses the following algorithm when performing real-time adaptive optimization on the illumination intensity correction model and the pressure sensor mechanical model: Define a model parameter vector. For the illumination intensity correction model, the model parameter vector includes a correction factor. For the pressure sensor mechanical model, the model parameter vector includes a friction coefficient and an elastic coefficient. The Kalman filter algorithm is used to update the model parameters online. The parameter values ​​at the current moment are predicted based on the parameter estimates at the previous moment and the dynamic model of the system. At the same time, the covariance matrix is ​​predicted. The Kalman gain is calculated based on the current measurement values ​​and the measurement model, and then the parameter estimates are updated and the covariance matrix is ​​updated.

6. The industrial big data information system for industrial production according to claim 5, characterized in that: The machine learning algorithm unit (3) predicts the sensor data deviation through a deviation prediction module (31): The collected sensor data and dust concentration data are preprocessed, and a support vector regression algorithm is introduced to train the preprocessed data. The goal of the support vector regression algorithm is to find an optimal hyperplane that minimizes the error between the predicted value and the true value within a preset range; The support vector regression model is trained using the training data set. The parameters of the model are determined by minimizing the objective function, and a penalty factor is introduced to control the balance between the complexity and error tolerance of the model. The trained support vector regression model is used to predict the real-time sensor data to obtain the deviation prediction value of the sensor data.

7. The industrial big data information system for industrial production according to claim 6, characterized in that: The machine learning algorithm unit (3) further comprises a signal compensation module (32), wherein the signal compensation module (32) adopts the following method when dynamically adjusting the compensation value of the sensor output signal in combination with the correction result of the built-in physical model: Set the sensor data correction value calculated by the physical model and the sensor data deviation predicted by the machine learning algorithm respectively, and define a weight coefficient to balance the impact of the physical model correction result and the machine learning prediction result, and calculate the compensation value based on the weight coefficient, data correction value and data deviation; The calculated compensation value is used to correct the original output signal of the sensor to obtain a corrected sensor output signal.

8. The industrial big data information system for industrial production according to claim 7, characterized in that: The dynamic compensation mechanism unit (4) uses the following method to perform real-time correction on the original output signal of the sensor: The theoretical deviation value of the sensor output signal under the current dust environment is calculated based on the light intensity correction model and the pressure sensor mechanical model. The calculation results of the light intensity correction model and the pressure sensor mechanical model are corrected using a machine learning algorithm to obtain a corrected deviation prediction value. The corrected deviation prediction value is used as a compensation value to correct the original output signal of the sensor in real time to obtain a corrected sensor output signal.

Images