Inert gas dynamic monitoring control system for metal powder preparation

Abstract

The invention discloses an inert gas dynamic monitoring control system for metal powder preparation, belongs to the technical field of inert gas monitoring, and solves the problems of signal drift caused by nano-particle adsorption and detection result drift or distortion caused by oxygen release of oxide particles. Comprising a parameter set operation module, an information output module in communication connection with the output end of the parameter set operation module, and a core detection module, a nano-particle detection module, an environment detection module and an auxiliary detection module which are in communication connection with the input end of the parameter set operation module, the core detection module collects the oxygen concentration and the moisture dew point. According to the system disclosed by the invention, all detection modules are used for acquiring data, and methods of multi-dimensional processing, real-time tracking, correction and the like are adopted, so that full-chain fine processing is realized, the problems of nano-particle adsorption and oxide oxygen release are effectively solved, a high-precision correction scheme is provided, the calibration period is prolonged, and drift or distortion of a detection result is slowed down.

Description

Technical Field

[0001] The present invention relates to the technical field of inert gas monitoring, and in particular to an inert gas dynamic monitoring and control system for metal powder making. Background Art

[0002] The most widely used metal powder production process currently is atomization, specifically inert gas atomization, which is performed under an inert gas atmosphere. This process requires inert gas protection, and after the metal powder production is complete, the inert gas must be recovered.

[0003] However, in trace gas purity testing in inert gas (argon) recovery systems, the high-temperature metal vapors (such as aluminum and titanium) generated by the atomization process rapidly condense to form nanoscale metal particles (10-100 nm in diameter). Due to their extremely high surface area and surface activity, these particles easily adsorb to sensor sensitive components and the inner walls of pipelines, causing multiple interference effects. First, the physical adsorption of the nanoparticles can cover the sensor detection interface, such as the platinum electrode of an electrochemical oxygen sensor or the optical window of a laser spectrometer, hindering the diffusion and contact of target gas molecules (O₂, H₂O), resulting in low oxygen detection values. At the same time, water molecules chemically adsorbed on the particle surface desorb and release during temperature and humidity fluctuations, causing abnormally high dew point detection values. Second, the metal particles react with trace oxygen in the argon environment to form oxides. These oxides undergo partial thermal decomposition in the high-temperature region, releasing trace amounts of oxygen. Especially when the system pressure fluctuates, the dynamic equilibrium on the oxide surface is disrupted. The continuously released oxygen causes periodic fluctuations in the detection value, forming "false peaks." In addition, when nanoparticles migrate with the airflow in the pipeline, they are deposited step by step due to van der Waals forces and electrostatic effects, forming a porous adsorption layer, which further aggravates gas retention and cross-contamination, causing the detection baseline to drift slowly.

[0004] It is precisely because of the above-mentioned problems of signal drift caused by nanoparticle adsorption and oxygen release by oxide particles that the detection results drift or are distorted. Therefore, in order to ensure data accuracy, unmanned calibration of parameters and results is required to ensure that the final exported data is accurate during the dynamic monitoring process.

[0005] Therefore, a dynamic monitoring and control system for inert gas used in metal powder making is proposed to solve or alleviate the above problems. Summary of the Invention

[0006] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an inert gas dynamic monitoring and control system for metal powder making.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions: A dynamic monitoring and control system for inert gas used in metal powder making, comprising a parameter set operation module, an information output module communicatively connected to an output end of the parameter set operation module, a core detection module communicatively connected to an input end of the parameter set operation module, a nanoparticle detection module, an environmental detection module, and an auxiliary detection module; The core detection module collects oxygen concentration and moisture dew point, the nanoparticle detection module collects particle concentration, particle composition, and surface deposition amount, the environmental detection module collects multi-point temperature, pressure, and argon flow rate, and the auxiliary detection module collects metal vapor concentration and oxide decomposition rate. The parameter set operation module performs adsorption-release oxygen modeling, state estimation and correction, parameter adaptation and safety control based on the feedback collection parameters, and the information output module feeds back the output results of the parameter set operation module to the outside.

[0008] Preferably, the core detection module includes a laser absorption spectrometer and a chilled mirror dew point meter, the nanoparticle detection module includes an electromigration particle size spectrometer, an aerosol mass spectrometer, and a quartz crystal microbalance, the environmental detection module includes a platinum resistance temperature sensor PT1000 array, a piezoresistive pressure sensor, and a vortex flowmeter, the auxiliary detection module includes an atomic emission spectrometer and a high-temperature in-situ mass spectrometer, the parameter set operation module includes a processor, and the information output module includes a display.

[0009] Preferably, the parameter set operation module performs adsorption-release oxygen modeling, state estimation and correction, parameter adaptation and safety control according to the feedback collection parameters, including the following steps: Simultaneously collect oxygen concentration, water dew point, particle concentration, particle composition, surface deposition, temperature, pressure, flow rate, metal vapor concentration and oxide decomposition rate parameters; Based on the adsorption kinetics and thermal decomposition mechanism, a coupled interference model of nanoparticle adsorption and oxide oxygen release was established; The real gas concentration is dynamically estimated through extended Kalman filtering, and the detection error is corrected by fusing forward prediction and residual feedback; Online adaptive update of model parameters, combined with periodic self-checking and surface cleanliness monitoring to maintain system accuracy; Trigger hierarchical alarms based on multi-factor dynamic thresholds, and execute fault tracing and redundancy control strategies.

[0010] Preferably, the synchronous collection of oxygen concentration, water dew point, particle concentration, particle composition, surface deposition amount, temperature, pressure, flow rate, metal vapor concentration and oxide decomposition rate parameters includes the following steps: All sensor data is hardware-triggered and synchronized through the precision time protocol to ensure data timestamp alignment, and cubic spline interpolation is used to fill time gaps in asynchronous data. The oxygen and moisture signals are decomposed into five layers of wavelet decomposition to filter out high-frequency noise layer by layer, where the noise threshold of each layer is dynamically adjusted according to the signal standard deviation; A reference noise model is constructed based on the vibration signal collected by the triaxial accelerometer, and the coupling interference of mechanical vibration on the detection signal is eliminated through the least mean square adaptive filter.

[0011] Preferably, the nanoparticle adsorption-oxide oxygen release coupling interference model is established based on adsorption kinetics and thermal decomposition mechanism, comprising the following steps: In the nanoparticle adsorption kinetic model, the rate of change of surface coverage over time is equal to the particle concentration multiplied by the product of the adsorption rate constant and the uncovered surface area, minus the product of the desorption rate constant and the current coverage, where the adsorption rate constant decreases exponentially with increasing temperature and the desorption rate constant increases exponentially with increasing temperature; In the oxide thermal decomposition model, the oxygen release of aluminum oxide and titanium oxide is determined by activation energy, real-time temperature and gas residence time. The oxygen release rate is exponentially related to temperature and is proportional to the oxide concentration and the square root of the residence time.

[0012] Preferably, the method of dynamically estimating the true gas concentration by using an extended Kalman filter and fusing forward prediction with residual feedback to correct the detection error comprises the following steps: The state vector of the extended Kalman filter includes the true oxygen concentration, the true water dew point, the sensor surface coverage, the coverage change rate, the sensor surface temperature and the amount of oxygen released by the oxide; The rate of change of the true oxygen concentration is negatively correlated with the current concentration. The rate of change of the water dew point is negatively correlated with the current dew point. The rate of change of the surface coverage is determined by the dynamic balance of adsorption and desorption rates. The rate of change of the surface temperature is determined by the difference between the heat input and the ambient heat loss. In the nonlinear observation equation, the original value of oxygen detection is equal to the actual oxygen concentration multiplied by one minus the product of the shielding coefficient and the surface coverage rate. The original value of moisture detection is equal to the actual moisture dew point minus the product of the hydrolysis reaction coefficient and the surface coverage rate of change. The short-term trend of oxygen concentration is predicted based on historical data through a forward prediction filter and fused with the Kalman filter estimate according to dynamic weights, with the weights allocated according to the inverse of the variance of the historical residuals.

[0013] Preferably, the online adaptive updating of model parameters, combined with periodic self-checking and surface cleanliness monitoring to maintain system accuracy, comprises the following steps: The recursive least squares method was used to update the adsorption rate constant, desorption rate constant, and shielding coefficient online, with the forgetting factor set to 0.95 to 0.99. The regression vector included particle concentration, temperature, and surface coverage. Switch to the built-in high-purity argon gas source every 2 hours to perform zero point calibration and span calibration. The zero point offset is equal to the original detection value minus the standard reference value; The deposition quality on the sensor surface is monitored in real time by a quartz crystal microbalance. When the deposition exceeds 100 nanograms per square centimeter, ultrasonic cleaning or high-pressure argon reverse purge is triggered. After cleaning, it is verified whether the zero drift has returned to the initial value.

[0014] Preferably, triggering a hierarchical alarm based on a multi-factor dynamic threshold and executing a fault tracing and redundancy control strategy includes the following steps: The oxygen dynamic alarm threshold is the historical mean plus three standard deviations, with temperature compensation and pressure compensation added. The temperature compensation coefficient is 0.02 ppm per degree Celsius, and the pressure compensation coefficient is 0.05 ppm per MPa. The dynamic moisture alarm threshold is the historical mean plus three standard deviations, with a particle concentration change rate correction factor of 0.1 degrees Celsius per cubic centimeter per second. Fault tracing uses a Bayesian network to construct a causal relationship diagram, linking abnormal oxygen increase events with potential fault sources, including pipeline leakage, filter element blockage, or sensor failure; When the main sensor exceeds the threshold value alarm three times in a row, it switches to the backup sensor and triggers the high-pressure argon back-purge process. The purge pressure is 1.2 to 1.5 times the system working pressure.

[0015] The present invention has the following beneficial effects: The system of the present invention can collect data through various detection modules when facing dynamic processes such as nanoparticle adsorption and oxide oxygen release. Through the comprehensive application of methods such as multi-dimensional processing of multi-physical field data, real-time adsorption state tracking, thermodynamic correction of oxide oxygen release, dynamic filtering and state estimation, it realizes the refined processing of the entire chain from data acquisition to fault response, effectively solves the problems of nanoparticle adsorption and oxide oxygen release, and provides a high-precision correction solution for trace-level gas purity detection in inert gas recovery systems, which can extend the manual calibration cycle and significantly reduce the problem of drift or distortion of detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 It is a structural block diagram of the present invention.

[0018] 101. Core detection module; 102. Nanoparticle detection module; 103. Environmental detection module; 104. Auxiliary detection module; 2. Parameter set operation module; 3. Information output module. DETAILED DESCRIPTION

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0020] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0021] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0022] In the description of the present invention, it should be understood that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, or are the orientation or position relationship in which the product of the invention is usually placed when in use, or are the orientation or position relationship commonly understood by those skilled in the art. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0023] Furthermore, the terms “first,” “second,” “third,” etc., are merely used for distinguishing descriptions and are not to be understood as indicating or implying relative importance.

[0024] In the description of the present invention, it should also be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0025] A dynamic monitoring and control system for inert gas used in metal powder production. After the atomized metal powder is prepared, in order to ensure the quality of the metal powder and avoid an increase in the oxygen content of the metal powder during the recovery process of the inert gas, which is actually argon, trace-level gas purity testing is performed to ensure that the purity of the recovered argon is high enough to have no negative impact on the quality of the metal powder, thereby ensuring the performance and quality of the final product.

[0026] In trace-level gas purity detection, nanoparticles generated by the atomization process are easily adsorbed on the sensor and the inner wall of the pipeline, causing interference. Their physical adsorption covers the detection interface, hindering the contact of gas molecules, resulting in low oxygen detection and abnormally high dew point detection. Metal particles react with oxygen to form oxides, which decompose at high temperatures, especially when the pressure fluctuates, continuously releasing oxygen to form false peaks. Nanoparticles migrate and deposit, aggravating gas retention and contamination, causing the detection baseline to drift. These influences will cause the detection results of trace-level gas purity detection to drift or be distorted.

[0027] In order to solve this problem, a dynamic monitoring and control system for inert gas used in metal powder making is proposed to calibrate the collected parameters and calculated purity results to ensure that the final gas purity in the inert gas recovery process meets the requirements.

[0028] A dynamic monitoring and control system for inert gas used in metal powder making, such as Figure 1 As shown, it includes a parameter set operation module 2, an information output module 3 communicatively connected to the output end of the parameter set operation module, a core detection module 101 communicatively connected to the input end of the parameter set operation module 2, a nanoparticle detection module 102, an environment detection module 103, and an auxiliary detection module 104; The core detection module 101 collects oxygen concentration and moisture dew point, the nanoparticle detection module 102 collects particle concentration, particle composition, and surface deposition amount, the environmental detection module 103 collects multi-point temperature, pressure, and argon flow rate, and the auxiliary detection module 104 collects metal vapor concentration and oxide decomposition rate. The parameter set operation module 2 performs adsorption-release oxygen modeling, state estimation and correction, parameter adaptation and safety control based on the feedback collection parameters, and the information output module 3 feeds back the output results of the parameter set operation module 2 to the outside.

[0029] The core detection module 101 includes a laser absorption spectrometer and a chilled mirror dew point meter, the nanoparticle detection module 102 includes an electromigration particle size spectrometer, an aerosol mass spectrometer, and a quartz crystal microbalance, the environmental detection module 103 includes a platinum resistance temperature sensor PT1000 array, a piezoresistive pressure sensor, and a vortex flowmeter, the auxiliary detection module 104 includes an atomic emission spectrometer and a high-temperature in-situ mass spectrometer, the parameter set operation module 2 includes a processor, and the information output module 3 includes a display.

[0030] In the above technical solution, full-chain data collection and processing are achieved through the collaboration of multiple modules, and parameters such as oxygen concentration, water dew point, particle concentration, temperature, and pressure are collected separately. Through adsorption-oxygen release modeling, extended Kalman filter dynamic estimation, parameter adaptive update and safety control strategy, the system can correct the detection errors caused by nanoparticle adsorption and oxide oxygen release in real time, and solve the signal drift problem caused by nanoparticle adsorption and oxide oxygen release. The collaborative actions of multiple modules cover the entire process from data collection to fault response, improve detection accuracy, and reduce the frequency of manual calibration and maintenance costs.

[0031] Preferably, the parameter set operation module 2 performs adsorption-release oxygen modeling, state estimation and correction, parameter adaptation and safety control according to the feedback collection parameters, including the following steps: Simultaneously collect oxygen concentration, water dew point, particle concentration, particle composition, surface deposition, temperature, pressure, flow rate, metal vapor concentration and oxide decomposition rate parameters; Based on the adsorption kinetics and thermal decomposition mechanism, a coupled interference model of nanoparticle adsorption and oxide oxygen release was established; The real gas concentration is dynamically estimated through extended Kalman filtering, and the detection error is corrected by fusing forward prediction and residual feedback; Online adaptive update of model parameters, combined with periodic self-checking and surface cleanliness monitoring to maintain system accuracy; Trigger hierarchical alarms based on multi-factor dynamic thresholds, and execute fault tracing and redundancy control strategies.

[0032] The aforementioned method, based on multi-source data fusion and dynamic modeling correction, aims to eliminate interference from nanoparticle adsorption and oxide oxygen release on the detection system. First, by synchronously collecting key parameters such as oxygen concentration, water dew point, particle concentration, and temperature, ensuring data timestamp alignment, and using wavelet decomposition and adaptive filtering to eliminate noise interference, the method provides input data with a high signal-to-noise ratio for subsequent modeling.

[0033] Secondly, a coupled interference model is constructed based on the adsorption kinetics and thermal decomposition mechanism of oxides to quantify the adsorption rate of nanoparticles on the sensor surface and the kinetic process of oxygen release by thermal decomposition of oxides, thereby converting the dynamic effects of physical adsorption and chemical oxygen release into a calculable mathematical relationship. This enables the system to dynamically quantify the coupling effect of nanoparticle adsorption and oxide oxygen release, correct the detection deviation from the root, and avoid the limitations of traditional single parameter calibration.

[0034] Subsequently, the extended Kalman filter is used to dynamically estimate the real gas concentration. Through the state vector, such as the real oxygen concentration, surface coverage, etc. and the nonlinear observation equation, the detection deviation caused by the sensor shielding effect and oxygen release interference is corrected in real time, overcoming the nonlinear error caused by the sensor surface coverage and oxygen release fluctuations, and significantly improving the authenticity and stability of oxygen and moisture dew point detection.

[0035] In addition, the above method steps introduce a forward prediction and residual feedback fusion mechanism, combine historical data trend prediction and real-time filtering results, dynamically adjust the weights to optimize the estimation accuracy, and the online adaptive update module optimizes the model parameters online through the recursive least squares method, specifically the adsorption rate constant, and combines periodic self-tests, specifically zero-point calibration and surface cleanliness monitoring, so as to maintain the long-term stability of the system. At the same time, it can also track system status changes in real time, reduce the frequency of manual calibration, and extend the maintenance cycle. Finally, the multi-factor dynamic threshold alarm strategy is based on historical data statistics and temperature and pressure compensation, triggering a graded response and executing redundant control to switch backup sensors, while achieving accurate graded alarms and quickly locating potential fault sources to ensure reliable operation of the system under faults.

[0036] Preferably, the simultaneous collection of oxygen concentration, water dew point, particle concentration, particle composition, surface deposition, temperature, pressure, flow rate, metal vapor concentration, and oxide decomposition rate parameters comprises the following steps: All sensor data is hardware-triggered and synchronized through the precision time protocol to ensure data timestamp alignment, and cubic spline interpolation is used to fill time gaps in asynchronous data. The oxygen and moisture signals are decomposed into five layers of wavelet decomposition to filter out high-frequency noise layer by layer, where the noise threshold of each layer is dynamically adjusted according to the signal standard deviation; A reference noise model is constructed based on the vibration signal collected by the triaxial accelerometer, and the coupling interference of mechanical vibration on the detection signal is eliminated through the least mean square adaptive filter.

[0037] Through the above method and steps, the data of all modules are hardware-triggered and synchronized through the precision time protocol to ensure that the timestamps of each parameter are strictly aligned and avoid the timing misalignment problem caused by asynchronous data. For asynchronous data that cannot be fully synchronized, the cubic spline interpolation method is used to fill the time gap and ensure data continuity.

[0038] Secondly, to address the high-frequency noise in the oxygen and moisture detection signals, a five-layer wavelet decomposition technique is used to filter out the noise layer by layer. The noise threshold of each layer is dynamically adjusted based on the signal standard deviation, effectively retaining the useful signal while suppressing noise interference. Furthermore, a three-axis accelerometer is used to collect mechanical vibration signals, construct a reference noise model, and use a least mean square adaptive filter to eliminate the coupling interference of vibration on the detection signal.

[0039] The combination of precision time protocol and cubic spline interpolation solves the timing deviation problem in multi-sensor data acquisition and avoids data distortion caused by traditional asynchronous processing. Wavelet decomposition technology dynamically adjusts the noise threshold, accurately filtering out high-frequency noise while retaining key signal characteristics, improving the accuracy of oxygen and moisture detection. The reference noise model based on the vibration signal and adaptive filtering technology effectively isolate the interference of mechanical vibration on the detection signal.

[0040] Preferably, based on the adsorption kinetics and thermal decomposition mechanism, a nanoparticle adsorption-oxide oxygen release coupling interference model is established, comprising the following steps: In the nanoparticle adsorption kinetics model, the rate of change of surface coverage over time is equal to the particle concentration multiplied by the product of the adsorption rate constant and the uncovered surface area, minus the product of the desorption rate constant and the current coverage, where the adsorption rate constant decreases exponentially with increasing temperature, and the desorption rate constant increases exponentially with increasing temperature; In the oxide thermal decomposition model, the amount of oxygen released by aluminum oxide and titanium oxide is determined by the activation energy, real-time temperature and gas residence time. The oxygen release rate is exponentially related to the temperature and is proportional to the oxide concentration and the square root of the residence time.

[0041] Specifically, in a clean argon environment, particles of known concentration were injected, the mass change of QCM was recorded, and the adsorption rate was fitted as ,in, To balance the time, To balance the probability, the QCM data and the Langmuir model are combined to generate the nanoparticle adsorption kinetic model: ,in, is the surface coverage, is the adsorption mass, is the maximum adsorption mass, the adsorption rate constant , A is the prefactor, is the adsorption activity energy, R is the gas constant, T is the temperature, is the concentration of nanoparticles in the gas, t is the time, and the desorption rate constant is , B is the pre-factor, is the desorption activation energy, and the oxygen detection value is corrected: , α is the adsorption shielding coefficient, moisture detection value correction: , β is the hydrolysis reaction coefficient; The thermal decomposition model of oxides is ,in, , dynamic updates To match the real-time working conditions, is the oxygen release coefficient of oxide i, is the concentration of oxide i, is the activation energy of thermal decomposition of oxide i, is the total amount of oxygen released during the thermal decomposition of oxides, is an exponential term that represents the temperature dependence of the reaction, is the gas residence time, , L is the length of the pipeline, v is the gas flow rate, is the reference time constant.

[0042] In the above method steps, a mathematical model of the coupled interference between nanoparticle adsorption and oxide oxygen release is established. In the nanoparticle adsorption kinetic model, the rate of change of surface coverage is jointly determined by the particle concentration, adsorption rate constant and desorption rate constant. Among them, the adsorption rate constant decreases exponentially with temperature, and the desorption rate constant increases exponentially with temperature. The dynamic adsorption-desorption equilibrium process of particles on the sensor surface is quantified, and the dynamic coverage process of nanoparticles on the sensor surface is accurately described, which solves the limitation of traditional empirical models that cannot reflect the influence of temperature.

[0043] In the oxide thermal decomposition model, the oxygen release of aluminum oxide and titanium oxide is calculated comprehensively based on the activation energy, real-time temperature and gas residence time. The oxygen release rate is exponentially related to the temperature and is proportional to the oxide concentration and the square root of the residence time. These two models convert the physical adsorption shielding effect and chemical oxygen release interference into calculable mathematical relationships through thermodynamic and kinetic mechanisms, providing a basis for dynamic correction of detection deviations.

[0044] Preferably, the actual gas concentration is dynamically estimated by using an extended Kalman filter, and the detection error is corrected by fusing forward prediction and residual feedback, including the following steps: The state vector of the extended Kalman filter includes the true oxygen concentration, the true water dew point, the sensor surface coverage, the coverage change rate, the sensor surface temperature and the amount of oxygen released by the oxide; The rate of change of the true oxygen concentration is negatively correlated with the current concentration. The rate of change of the water dew point is negatively correlated with the current dew point. The rate of change of the surface coverage is determined by the dynamic balance of adsorption and desorption rates. The rate of change of the surface temperature is determined by the difference between the heat input and the ambient heat loss. In the nonlinear observation equation, the original value of oxygen detection is equal to the actual oxygen concentration multiplied by one minus the product of the shielding coefficient and the surface coverage rate. The original value of moisture detection is equal to the actual moisture dew point minus the product of the hydrolysis reaction coefficient and the surface coverage rate of change. The short-term trend of oxygen concentration is predicted based on historical data through a forward prediction filter and fused with the Kalman filter estimate according to dynamic weights, with the weights allocated according to the inverse of the variance of the historical residuals.

[0045] Specifically, the state vector is , the state equation is: , the nonlinear observation equation is , the Kalman gain is updated to , where the Jacobian matrix , where x is the state vector, is the true concentration of oxygen, is the actual water concentration, is the sensor surface temperature, is the surface coverage, and is the attenuation coefficient of oxygen and moisture, which indicates the natural attenuation rate of oxygen and moisture during the detection process. is the input heat, which means the heat absorbed by the sensor surface from the outside world. h is the convection heat transfer coefficient, which means the convection heat transfer capacity between the sensor surface and the surrounding environment. A is the convection heat transfer area of ​​the sensor surface. is the ambient temperature, is the specific heat capacity, z is the observation vector containing the corrected oxygen concentration , moisture concentration 、 , sensor surface temperature , is the prior state error covariance matrix, which represents the statistical characteristics of the state estimation error before the observation update, is the observation noise covariance matrix, which represents the statistical characteristics of the observation noise.

[0046] Through the above method steps, the state vector design of the Kalman filter comprehensively covers the factors of adsorption masking, oxygen release interference, and thermodynamic changes, and can estimate the true gas concentration in real time, overcoming the nonlinear bias caused by sensor surface coverage. The fusion mechanism of forward prediction and Kalman filtering takes into account both historical trends and real-time status, and dynamic weight allocation optimizes the stability and response speed of the estimation results. This method is particularly suitable for scenarios with slow drift caused by nanoparticle adsorption and periodic fluctuations caused by oxygen release from oxides, making the oxygen and moisture detection values ​​closer to the actual environmental levels and reducing detection errors.

[0047] The Kalman filter's state vector includes key variables such as true oxygen concentration, moisture dew point, sensor surface coverage, and coverage change rate. Using nonlinear observation equations, the sensor's raw data is linked to the true state. For example, the oxygen detection value is obscured by surface coverage. The rate of change of true oxygen concentration is negatively correlated with the current concentration, the rate of change of moisture dew point is negatively correlated with the current dew point, and the rate of change of surface coverage is determined by the adsorption-desorption rate balance. Simultaneously, a forward prediction filter predicts the short-term trend of oxygen concentration based on historical data and fuses it with the Kalman filter estimate using dynamic weights assigned by the inverse of the historical residual variance, achieving coordinated optimization of short-term prediction and real-time correction.

[0048] Preferably, the model parameters are updated online and adaptively, and the system accuracy is maintained by combining periodic self-checking and surface cleanliness monitoring, including the following steps: The recursive least squares method is used to update the adsorption rate constant, desorption rate constant and shielding coefficient online. The forgetting factor is set to 0.95 to 0.99. The regression vector contains particle concentration, temperature and surface coverage. Specifically, the recursive least squares estimate , where the parameter vector , the regressor Including gas concentration , temperature T, surface coverage , is the parameter estimation error covariance matrix at the k-1th step, which is used to measure the uncertainty of parameter estimation. For the forgetting factor, is the observation value, which represents the observation data at the kth step; Switch to the built-in high-purity argon gas source every 2 hours to perform zero point calibration and span calibration. The zero point offset is equal to the original detection value minus the standard reference value; The deposition quality on the sensor surface is monitored in real time by a quartz crystal microbalance. When the deposition exceeds 100 nanograms per square centimeter, ultrasonic cleaning or high-pressure argon reverse purge is triggered. After cleaning, it is verified whether the zero drift has returned to the initial value.

[0049] In the above method steps, the recursive least squares method is used to update the adsorption rate constant, desorption rate constant and shielding coefficient online. The forgetting factor balances the weights of historical data and new data to ensure that the model parameters track the changes in the system state in real time. The built-in high-purity argon source is switched to for zero point and span calibration every 2 hours. The sensor drift is dynamically corrected by subtracting the standard value from the detection value. The quartz crystal microbalance monitors the deposition quality on the sensor surface in real time. When the deposition exceeds the threshold, ultrasonic cleaning or high-pressure argon reverse purge is triggered. After cleaning, the zero point recovery is verified to ensure long-term detection accuracy. In this way, the recursive least squares method realizes online optimization of model parameters, avoiding system shutdown caused by traditional offline calibration. The periodic self-test and calibration mechanism effectively suppresses sensor drift, extending the calibration cycle from the traditional 8 hours to more than 24 hours. The baseline drift problem caused by nanoparticle accumulation can also be solved through surface deposition monitoring and automatic cleaning technology.

[0050] Preferably, triggering a hierarchical alarm based on a multi-factor dynamic threshold and executing a fault tracing and redundancy control strategy includes the following steps: The oxygen dynamic alarm threshold is the historical mean plus three standard deviations, with temperature compensation and pressure compensation added. The temperature compensation coefficient is 0.02 ppm per degree Celsius, and the pressure compensation coefficient is 0.05 ppm per MPa. The dynamic moisture alarm threshold is the historical mean plus three standard deviations, with a particle concentration change rate correction factor of 0.1 degrees Celsius per cubic centimeter per second. Fault tracing uses a Bayesian network to construct a causal relationship diagram, linking abnormal oxygen increase events with potential fault sources, including pipeline leakage, filter element blockage, or sensor failure; When the main sensor exceeds the threshold value alarm three times in a row, it switches to the backup sensor and triggers the high-pressure argon back-purge process. The purge pressure is 1.2 to 1.5 times the system working pressure.

[0051] Through the above method steps, the oxygen alarm threshold is composed of the historical mean plus three times the standard deviation, with temperature and pressure compensation terms superimposed to dynamically adapt to environmental changes. The moisture alarm threshold introduces a particle concentration change rate correction term to prevent false alarms caused by particle adsorption. Fault tracing uses a Bayesian network to construct a causal relationship diagram to associate abnormal oxygen increase events with potential fault sources, such as pipeline leakage and filter element blockage. After the main sensor exceeds the threshold value for three consecutive alarms, it automatically switches to the backup sensor and triggers high-pressure argon backflushing to remove pipeline deposits and restore the detection function, thereby ensuring the continuity and safety of inert gas monitoring in the metal powder making process.

[0052] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A dynamic monitoring and control system for inert gas used in metal powder making, characterized in that: It includes a parameter set operation module (2), an information output module (3) communicatively connected to the output end thereof, and a core detection module (101), a nanoparticle detection module (102), an environment detection module (103), and an auxiliary detection module (104) communicatively connected to the input end of the parameter set operation module (2); The core detection module (101) collects oxygen concentration and water dew point, the nanoparticle detection module (102) collects particle concentration, particle composition, and surface deposition amount, the environmental detection module (103) collects multi-point temperature, pressure, and argon flow rate, the auxiliary detection module (104) collects metal vapor concentration and oxide decomposition rate, the parameter set operation module (2) performs adsorption-release oxygen modeling, state estimation and correction, parameter adaptation and safety control based on the feedback acquisition parameters, and the information output module (3) feeds back the output result of the parameter set operation module (2) to the outside.

2. The inert gas dynamic monitoring and control system for metal powder production according to claim 1, characterized in that: The core detection module (101) includes a laser absorption spectrometer and a cold mirror dew point meter, the nanoparticle detection module (102) includes an electromigration particle size spectrometer, an aerosol mass spectrometer, and a quartz crystal microbalance, the environmental detection module (103) includes a platinum resistance temperature sensor PT1000 array, a piezoresistive pressure sensor, and a vortex flowmeter, the auxiliary detection module (104) includes an atomic emission spectrometer and a high-temperature in-situ mass spectrometer, the parameter set operation module (2) includes a processor, and the information output module (3) includes a display.

3. The inert gas dynamic monitoring and control system for metal powder production according to claim 1, characterized in that: The parameter set operation module (2) performs adsorption-oxygen release modeling, state estimation and correction, parameter adaptation and safety control according to the feedback collection parameters, including the following steps: Simultaneously collect oxygen concentration, water dew point, particle concentration, particle composition, surface deposition, temperature, pressure, flow rate, metal vapor concentration and oxide decomposition rate parameters; Based on the adsorption kinetics and thermal decomposition mechanism, a coupled interference model of nanoparticle adsorption and oxide oxygen release was established; The real gas concentration is dynamically estimated through extended Kalman filtering, and the detection error is corrected by fusing forward prediction and residual feedback; Online adaptive update of model parameters, combined with periodic self-checking and surface cleanliness monitoring to maintain system accuracy; Trigger hierarchical alarms based on multi-factor dynamic thresholds, and execute fault tracing and redundancy control strategies.

4. The inert gas dynamic monitoring and control system for metal powder production according to claim 3, characterized in that: The synchronous collection of oxygen concentration, water dew point, particle concentration, particle composition, surface deposition amount, temperature, pressure, flow rate, metal vapor concentration and oxide decomposition rate parameters includes the following steps: All sensor data is hardware-triggered and synchronized through the precision time protocol to ensure data timestamp alignment, and cubic spline interpolation is used to fill time gaps in asynchronous data. The oxygen and moisture signals are decomposed into five layers of wavelet decomposition to filter out high-frequency noise layer by layer, where the noise threshold of each layer is dynamically adjusted according to the signal standard deviation; A reference noise model is constructed based on the vibration signal collected by the triaxial accelerometer, and the coupling interference of mechanical vibration on the detection signal is eliminated through the least mean square adaptive filter.

5. The inert gas dynamic monitoring and control system for metal powder production according to claim 4, characterized in that: The nanoparticle adsorption-oxide oxygen release coupling interference model is established based on adsorption kinetics and thermal decomposition mechanism, comprising the following steps: In the nanoparticle adsorption kinetic model, the rate of change of surface coverage over time is equal to the particle concentration multiplied by the product of the adsorption rate constant and the uncovered surface area, minus the product of the desorption rate constant and the current coverage, where the adsorption rate constant decreases exponentially with increasing temperature and the desorption rate constant increases exponentially with increasing temperature; In the oxide thermal decomposition model, the oxygen release of aluminum oxide and titanium oxide is determined by activation energy, real-time temperature and gas residence time. The oxygen release rate is exponentially related to temperature and is proportional to the oxide concentration and the square root of the residence time.

6. The inert gas dynamic monitoring and control system for metal powder production according to claim 5, characterized in that: The method of dynamically estimating the true gas concentration by using an extended Kalman filter and fusing forward prediction with residual feedback to correct the detection error includes the following steps: The state vector of the extended Kalman filter includes the true oxygen concentration, the true water dew point, the sensor surface coverage, the coverage change rate, the sensor surface temperature and the amount of oxygen released by the oxide; The rate of change of the true oxygen concentration is negatively correlated with the current concentration. The rate of change of the water dew point is negatively correlated with the current dew point. The rate of change of the surface coverage is determined by the dynamic balance of adsorption and desorption rates. The rate of change of the surface temperature is determined by the difference between the heat input and the ambient heat loss. In the nonlinear observation equation, the original value of oxygen detection is equal to the actual oxygen concentration multiplied by one minus the product of the shielding coefficient and the surface coverage rate. The original value of moisture detection is equal to the actual moisture dew point minus the product of the hydrolysis reaction coefficient and the surface coverage rate of change. The short-term trend of oxygen concentration is predicted based on historical data through a forward prediction filter and fused with the Kalman filter estimate according to dynamic weights, with the weights allocated according to the inverse of the variance of the historical residuals.

7. The inert gas dynamic monitoring and control system for metal powder production according to claim 6, characterized in that: The online adaptive updating of model parameters, combined with periodic self-checking and surface cleanliness monitoring to maintain system accuracy, includes the following steps: The recursive least squares method was used to update the adsorption rate constant, desorption rate constant, and shielding coefficient online, with the forgetting factor set to 0.95 to 0.

99. The regression vector included particle concentration, temperature, and surface coverage. Switch to the built-in high-purity argon gas source every 2 hours to perform zero point calibration and span calibration. The zero point offset is equal to the original detection value minus the standard reference value; The deposition quality on the sensor surface is monitored in real time by a quartz crystal microbalance. When the deposition exceeds 100 nanograms per square centimeter, ultrasonic cleaning or high-pressure argon reverse purge is triggered. After cleaning, it is verified whether the zero drift has returned to the initial value.

8. The inert gas dynamic monitoring and control system for metal powder production according to claim 7, characterized in that: The method of triggering a hierarchical alarm based on a multi-factor dynamic threshold and executing a fault tracing and redundancy control strategy includes the following steps: The oxygen dynamic alarm threshold is the historical mean plus three standard deviations, with temperature compensation and pressure compensation added. The temperature compensation coefficient is 0.02 ppm per degree Celsius, and the pressure compensation coefficient is 0.05 ppm per MPa. The dynamic moisture alarm threshold is the historical mean plus three standard deviations, with a particle concentration change rate correction factor of 0.1 degrees Celsius per cubic centimeter per second. Fault tracing uses a Bayesian network to construct a causal relationship diagram, linking abnormal oxygen increase events with potential fault sources, including pipeline leakage, filter element blockage, or sensor failure; When the main sensor exceeds the threshold value alarm three times in a row, it switches to the backup sensor and triggers the high-pressure argon back-purge process. The purge pressure is 1.2 to 1.5 times the system working pressure.

Images