Temperature field real-time monitoring method and system based on special ceramic sintering

Abstract

The invention discloses a temperature field real-time monitoring method and system based on special ceramic sintering, and relates to the field of temperature monitoring of special ceramic sintering, and the method specifically comprises the following steps: S1, processing a sensor array; s2, construction of multi-parameter data acquisition; s3, a temperature field reconstruction and prediction model based on multi-physics field coupling; and S4, temperature field real-time monitoring and abnormity early warning are carried out. According to the temperature field real-time monitoring method and system based on special ceramic sintering, the temperature sensor is preset in the green body, a three-dimensional temperature field reconstruction function is combined, discrete sensor node real-time temperature data can be converted into a continuous three-dimensional temperature field cloud picture, temperature distribution of the surface of the green body can be clearly presented, and the temperature field can be monitored in real time. And the temperature gradient change from the surface layer to the core can be more accurately restored, so that an operator can intuitively master the temperature state of the whole green body, the situation that the whole process is misjudged due to local data is avoided, and the comprehensiveness and accuracy of temperature field monitoring are fundamentally improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of temperature monitoring of special ceramic sintering, and particularly relates to a temperature field real-time monitoring method and system based on special ceramic sintering. BACKGROUND

[0002] Special ceramics are widely used in high-end fields such as aerospace, electronic information and new energy due to their excellent properties such as high strength, high temperature resistance and corrosion resistance.

[0003] The existing special ceramic sintering temperature field monitoring technology mainly includes contact type and non-contact type. The contact type monitoring takes a thermocouple as the core, directly inserts the thermocouple probe into the ceramic body or places it at a specific position in the sintering furnace, and realizes temperature measurement through the thermoelectric effect. However, this technology has obvious defects: on the one hand, the thermocouple itself is prone to oxidation and embrittlement at high temperatures above 1600℃, and has a short service life; on the other hand, the spatial arrangement of the thermocouple is limited in number, and cannot realize the full coverage of the temperature field, only discrete point temperature data can be obtained, and it is difficult to reflect the continuous distribution characteristics of the temperature field. In addition, the existing monitoring method generally lacks dynamic response capability, cannot real-time track the rapid changes of the temperature field in the sintering process (such as the stage with a heating rate of 50℃ / min or above), and the data processing method is simple, only single temperature data can be recorded, and the spatial reconstruction and trend prediction of the temperature field cannot be realized. Therefore, it is an urgent technical problem in the field of special ceramic preparation to develop a real-time monitoring method with high temperature adaptability, high spatial resolution, fast dynamic response and the ability to monitor the surface and internal temperature field of the body at the same time.

[0004] Therefore, it is necessary to propose a temperature field real-time monitoring method and system based on special ceramic sintering to solve the above problems. SUMMARY

[0005] The main purpose of the present application is to provide a temperature field real-time monitoring method and system based on special ceramic sintering, which can effectively solve the problems in the background art.

[0006] To achieve the above purpose, the technical scheme adopted by the present application is as follows: A temperature field real-time monitoring method based on special ceramic sintering, comprising the following operation steps: S1: processing of the sensor array, preparing a temperature sensor doped with conductive particles based on special ceramic as the matrix, embedding the sensor array in the ceramic body according to a three-dimensional grid layout, forming an array of 216 nodes, and realizing full-domain temperature acquisition; S2: construction of multi-parameter data acquisition, building a system including signal conditioning, acquisition card, environmental sensor and upper computer, synchronously collecting sensor temperature signal and furnace pressure, airflow and humidity parameters, and displaying in real time after pretreatment; S3: Temperature field reconstruction and prediction model based on multi-physical field coupling, combining discrete temperature data and environmental parameters, reconstructing the global temperature field by interpolation; using the LSTM model to predict the future temperature trend based on historical data, achieving dynamic prediction; S4: Real-time temperature field monitoring and abnormal warning, integrating data to display the temperature field distribution and trend in real time, identifying abnormalities through temperature gradient, deviation and predicted value, and grading warning.

[0007] Preferably, in S1, the following steps are specifically included: S101: Selection of temperature sensor, prepared from special ceramic powder as base body and doped with nano-sized conductive particles, whose working principle is based on the contact resistance change rule of special ceramic powder base body and nano-sized conductive particles at high temperature: when the temperature rises, the lattice vibration of special ceramic powder base body intensifies, and the contact resistance between nano-sized conductive particles changes linearly with temperature rise, so the temperature can be inverted by measuring the resistance value; S102: Presetting of sensor array, according to the shape and size of the special ceramic body to be sintered, the spatial layout of the sensor array is designed in a three-dimensional grid layout, and the diameter and height of the special ceramic body are equally divided into six parts as the basis for the layout of the sensor array, as follows: Radial direction: along the radius direction of the body, 1 sensor layer is set at each division, a total of 6 layers, respectively denoted as R0, R1, R2, R3, R4, R5, wherein R0 is the center layer and R5 is the surface layer; Axial direction: along the height direction of the body, 1 sensor row is set at each division, a total of 6 rows, respectively denoted as H0, H1, H2, H3, H4, H5; Circumferential direction: along the circumference of the special ceramic body, 1 sensor column is set every 60°, a total of 6 columns, respectively denoted as C0, C1, C2, C3, C4, C5; Through the above layout, 216 sensor nodes are formed inside the body, each node corresponds to a unique spatial coordinate, achieving full coverage of the internal and surface temperature field of the special ceramic body.

[0008] Preferably, in S2, the following steps are specifically included: S201: Signal conditioning, the resistance signal of the temperature sensor is amplified and filtered by the signal conditioning module, which specifically includes: using a high-precision programmable constant current source to provide a constant current of 100μA to the temperature sensor; differential amplification: using an instrument amplifier of model AD620 to differentially amplify the voltage signal between the two ends of the temperature sensor, with an amplification factor of 1000; filtering: using a second-order low-pass filter with a cutoff frequency of 10Hz; S202: Data acquisition, select the model of high-speed synchronous data acquisition card NIPCIe-6363, used to collect the voltage signals of all temperature sensors at the same time; S203: Obtain the environmental parameters in the special ceramic sintering furnace through the environmental data parameter sensor, set three environmental parameter monitoring points in the furnace, located at the top, middle and bottom of the furnace respectively, and install the following sensors at each monitoring point: Atmosphere pressure sensor: piezoelectric pressure sensor is used to monitor the pressure change of the protective atmosphere in the furnace; airflow velocity sensor: hot-wire anemometer is used to monitor the flow velocity of the airflow in the furnace; humidity sensor: ceramic humidity sensor is used to monitor the residual moisture content in the furnace; S204: Host computer, through the host computer: control the data acquisition card to collect temperature signals and environmental parameter signals at the set sampling interval, and store the raw data to the hard disk; data preprocessing: denoising processing for the collected temperature signals, and smoothing processing for the environmental parameter signals; real-time display: real-time display of the temperature value and environmental parameter value of each temperature sensor node, and display in the form of numerical table and curve.

[0009] Preferably, in S3, the following steps are specifically included: S301: Temperature field reconstruction model, formula one is: ; Wherein is the temperature value of any spatial point inside the special ceramic body; is the three-dimensional rectangular coordinate of the spatial point; is the weight coefficient of the temperature sensor node, used to measure the contribution degree of the node to the temperature; is the index of the radial direction R0-R5 of the temperature sensor; ; is the index of the axial direction H0-H5 of the temperature sensor; is the index of the circumferential direction C0-C5 of the temperature sensor; is the Euclidean distance between the target spatial point and the sensor node , which is the core independent variable of the radial function ; is the three-dimensional rectangular coordinate of the sensor node; is the environmental parameter correction factor, used to correct the influence of the furnace atmosphere pressure and airflow velocity on the temperature field distribution; According to the formula, the temperature value of any spatial point inside the special ceramic body can be calculated, and then the three-dimensional visualization reconstruction of the temperature field is realized through the Matlab software, and the uniformity of the distribution of the temperature field is intuitively displayed. S302: Based on the LSTM model, the historical temperature data and the environmental parameter data are combined to construct a temperature field dynamic prediction model, and the input parameters of the LSTM model include: The historical temperature data: the temperature values of each temperature sensor node in the past 10 sampling times, a total of 100ms, a total of 216x10=2160 features; The historical environmental parameter data: the atmospheric pressure , airflow speed , humidity in the past 10 sampling times, a total of 3x10=30 features; The current sintering stage parameters: the current heating rate , the current holding time , a total of 2 features; The output parameters of the model are: the temperature prediction values of each temperature sensor node in the future 5 sampling times, a total of 50ms, a total of 216x5=1080 features.

[0010] Preferably, the training loss function of the LSTM model adopts the root mean square error, and formula two is: ; Wherein is the average deviation degree of all predicted data and real data; is the number of training samples; is the number of temperature data points of each sample; is the index of the th training sample ; is the index of the th temperature data point ; is the predicted temperature value of the th temperature data point in the th training sample ; is the real temperature value of the th temperature data point in the th training sample ; The LSTM model is trained through the Adam optimizer, the number of training iterations is 500 times, the learning rate is set to 0.001, when the RMSE is less than 0.3℃, the LSTM model training is completed, and the trained model can be embedded in the host computer, real-time receive current data and output the temperature field prediction result in the future 50ms, realize the dynamic warning of the temperature field.

[0011] Preferably, in S4, the following steps are specifically included: S401: Real-time monitoring, specifically including: Data synchronization acquisition: Collect the original resistance signals of 216 temperature sensor nodes at a fixed sampling interval of 20 ms, and convert them into real-time temperature values through the signal conditioning module Here, represents the current sampling time, represents the temperature represent the atmosphere pressure of three monitoring points in the furnace , airflow speed , humidity , and all collected data is transmitted to the upper computer for temporary caching; Temperature field real-time reconstruction: Use Formula One to take the current time as the measured temperature of the , node, and substitute it into Formula One to calculate the temperature value of any point in the three-dimensional space of the special ceramic body ; Convert to a three-dimensional temperature field cloud map through the LabVIEW Matlab engine, with a cloud map update frequency of 10 Hz; Temperature dynamic prediction: Input the past 10 time , , , and current process parameters into the LSTM model to output the predicted temperature for the next 5 times ; display the real-time and predicted trends in the upper computer with virtual-real curves, where is the time; S402: Real-time anomaly judgment based on the of Formula One reconstructed through S401, if the temperature gradient of any adjacent point > 5℃ / mm, the mean deviation of a node and adjacent nodes exceeds 8℃, trigger a red alarm, and display the abnormal node and its , where is the temperature difference between the space point and its adjacent points, is the straight-line distance between the two points; Predictive anomaly judgment: if exceeds the sintering process set temperature ±10℃, and the prediction rate > 60℃ / min, trigger a yellow warning, and mark the warning time and its .

[0012] The application discloses a temperature field real-time monitoring system based on special ceramic sintering, which comprises a multi-parameter data acquisition module, a temperature field reconstruction module, a temperature field dynamic prediction module and a real-time monitoring and abnormal early warning module. The signal conditioning sub-module is used for converting resistance change of the temperature sensor into a voltage signal, differentially amplifying the millivolt-level voltage signal by 1000 times to a volt-level through an instrument amplifier, filtering out high-frequency electromagnetic interference in the furnace through a second-order 10Hz low-pass filter and ensuring that the signal-to-noise ratio of the temperature sensor is greater than 60dB; The temperature data acquisition sub-module is used for collecting the voltage signal through a high-speed synchronous data acquisition card, converting the voltage signal into real-time temperature of the corresponding node and transmitting the real-time temperature to an upper computer through a PCIe interface at high speed. The environment parameter acquisition sub-module is used for synchronously collecting furnace atmosphere pressure, air flow speed and humidity by arranging one monitoring point at the top, middle and bottom of the sintering furnace and installing an atmosphere pressure sensor, an air flow speed sensor and a humidity sensor respectively. The upper computer data processing sub-module is used for running a program developed by LabVIEW on an industrial control computer, pre-processing the collected real-time temperature value by using a wavelet threshold denoising algorithm and simultaneously displaying the real-time temperature value of each node, the values of the environment parameters and the change curves of the environment parameters on an interface in real time.

[0013] Preferably, the temperature field reconstruction module is based on discrete real-time temperature values and environment parameters and realizes continuous spatial distribution display of the temperature field in the green body.

[0014] Preferably, the temperature field dynamic prediction module is used for constructing a complete input feature system, automatically extracting historical key data, specifically, real-time temperature data of 216 sensor nodes in the past 10 sampling moments, atmosphere pressure, air flow speed and humidity in the sintering furnace at the same period and simultaneously integrating key process parameters in the current sintering stage to form a multi-dimensional input feature set covering temperature, environment and process.

[0015] Preferably, the real-time monitoring and abnormal early warning module specifically comprises: The real-time monitoring sub-module is used for receiving global continuous temperature distribution data of the temperature field reconstruction module, generating a three-dimensional temperature field cloud picture, simultaneously integrating real-time temperature data and environment parameters of the multi-parameter data acquisition module and predicted temperature data of the temperature field dynamic prediction module and displaying the real-time temperature data, the environment parameters and the predicted temperature data in the form of subareas in the same upper computer interface, including a real-time temperature value list of each node, a comparison picture of a real-time temperature change curve and a predicted curve and a real-time value and change trend picture of the environment parameters. Further comprising an abnormal early warning submodule for automatically calculating the temperature gradient of any adjacent spatial point in the three-dimensional temperature field, if the gradient exceeds 5℃ / mm, and the deviation of the real-time temperature value of a node from the average temperature of the surrounding six adjacent nodes exceeds 8℃, a red first-level alarm is triggered immediately, and the abnormal node number, corresponding spatial coordinates and real-time temperature value are marked; The predicted temperature of each node is automatically compared with the process set temperature range, if the predicted temperature exceeds the range, and the heating rate in the predicted period exceeds the upper limit of the safe heating rate, a yellow second-level early warning is triggered: the warning time, corresponding node number and predicted temperature are marked, reminding the operator to adjust the process parameters in advance to avoid abnormalities.

[0016] Compared with the prior art, the present application provides a temperature field real-time monitoring method and system based on special ceramic sintering, which has the following beneficial effects: The temperature field real-time monitoring method and system based on special ceramic sintering realizes the globalization and stereoscopic of temperature field monitoring, and completely solves the pain point that traditional single-point or local monitoring cannot reflect the overall temperature distribution of the blank. Traditional methods rely on a few external or surface temperature sensors, and can only obtain discrete local temperature data, which is difficult to capture the temperature change inside the blank, and is easy to cause sintering defects due to normal surface temperature and abnormal internal temperature. The present method can convert the real-time temperature data of discrete sensor nodes into continuous three-dimensional temperature field cloud images by pre-setting distributed temperature sensors inside the blank and combining three-dimensional temperature field reconstruction function, which can not only clearly present the temperature distribution on the surface of the blank, but also accurately restore the temperature gradient change from the surface to the core, so that the operator can intuitively master the temperature state of the whole blank, avoid misjudgment of the overall process due to local data, and fundamentally improve the comprehensiveness and accuracy of temperature field monitoring.

[0017] The temperature field real-time monitoring method and system based on special ceramic sintering can output the temperature field change trend in the future short time based on historical monitoring data, current environmental parameters and process parameters through temperature field dynamic prediction. This prediction ability enables the operator to adjust the process parameters such as heating rate and atmosphere pressure in advance according to the predicted temperature trend, so as to eliminate potential temperature risks and effectively reduce quality problems such as special ceramic cracking and insufficient density caused by temperature mutation, thereby significantly improving the stability of sintering process and product qualification rate.

[0018] The temperature field real-time monitoring method and system based on special ceramic sintering can integrate global temperature field cloud image, real-time temperature of each sensor node, environmental parameters, predicted temperature curve and other data into the same upper computer interface, and display in the form of sub-regions and strong correlation, so as to quickly judge whether the process is normal, significantly reduce the operation difficulty, and improve the control efficiency of the sintering process.

[0019] The temperature field real-time monitoring method and system based on special ceramic sintering have two types of abnormal early warning functions: real-time abnormality and predicted abnormality. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is a flowchart of the present application. DETAILED DESCRIPTION

[0021] To make the technical means, creative features, purposes and effects of the present application easy to understand, the present application is further described below in conjunction with specific embodiments.

[0022] Embodiment one: As shown in the figure, a temperature field real-time monitoring method based on special ceramic sintering includes the following operation steps: Figure 1 S1: processing of the sensor array, preparing a temperature sensor with special ceramic as the base body and doped with conductive particles, embedding the ceramic body in a three-dimensional grid layout to form an array of 216 nodes, realizing global temperature collection, specifically including the following steps: S101: selection of the temperature sensor, using special ceramic powder such as alumina and silicon carbide as the base body, and doped with nanoscale conductive particles such as graphene and titanium carbide, the working principle is as follows: based on the contact resistance change rule of the special ceramic powder base body and the nanoscale conductive particles at high temperature: when the temperature rises, the lattice vibration of the special ceramic powder base body intensifies, and the contact resistance between the nanoscale conductive particles changes linearly with the temperature rise, the temperature can be inverted by measuring the resistance value, the specific preparation process of the temperature sensor is as follows: Raw material mixing: take alumina powder with a particle size of 500 nm and graphene powder with a particle size of 20 nm in a mass ratio of 95:5, add anhydrous ethanol as a dispersant, and ball mill in a planetary ball mill at a speed of 300 r / min for 4 h to form a uniform slurry; Molding: pour the slurry into a mold, and use a cold isostatic pressing process: pressure 200 MPa, holding time 10 min, to press into a circular sensor body with a diameter of 5 mm and a thickness of 1 mm; Pre-sintering: place the sensor body in a sintering furnace, heat to 1200℃ at a rate of 5℃ / min in an air atmosphere, and hold for 2h to complete pre-sintering, forming a sensor base body with certain strength; ​Electrode preparation: platinum electrodes with a thickness of 100 nm were deposited on the upper and lower surfaces of the sensor substrate using magnetron sputtering technology as the lead-out end for resistance measurement; Packaging: a 500 nm thick silicon carbide coating was deposited on the surface of the sensor using chemical vapor deposition to improve the high-temperature corrosion resistance and insulation of the sensor; S102: Sensor array presetting, according to the shape and size of the special ceramic body to be sintered, a three-dimensional grid layout method was used to design the spatial layout of the sensor array, and the diameter and height of the special ceramic body were equally divided into six parts as the basis for the layout of the sensor array, as follows: Radial direction: along the radius direction of the body, 1 sensor layer was set at each division, a total of 6 layers, respectively denoted as R0, R1, R2, R3, R4, R5, wherein R0 is the center layer and R5 is the surface layer; Axial direction: along the height direction of the body, 1 sensor row was set at each division, a total of 6 rows, respectively denoted as H0, H1, H2, H3, H4, H5; Circumferential direction: along the circumferential direction of the special ceramic body, 1 sensor column was set at every 60°, a total of 6 columns, respectively denoted as C0, C1, C2, C3, C4, C5; Through the above layout, 6x6x6=216 sensor nodes were formed inside the body, each node corresponding to a unique spatial coordinate, achieving full coverage of the internal and surface temperature field of the special ceramic body. The presetting process of the sensor is as follows: A layer of ceramic slurry with a thickness of 5 mm was laid at the bottom of the special ceramic body mold, and the 6x6 sensors in the H0 row were fixed on the surface of the slurry according to the circumferential and radial coordinates, ensuring that the sensor electrode lead-out end faces outward from the mold; The mold was filled with ceramic slurry until the H0 row of sensors was completely covered, then the H1 row of sensors was laid, and the above steps were repeated until all 6 rows of sensors were laid; The ceramic slurry was filled to the top of the mold, and a cold isostatic pressing process was used: pressure 180 MPa, holding pressure for 15 min to integrate the body and sensors into a single piece; After molding, the electrode lead-out end of the sensor was connected to the high-temperature resistant wire preset on the outside of the mold for data acquisition.

[0023] S2: Construction of multi-parameter data acquisition, a system containing signal conditioning, acquisition card, environmental sensor and host computer was built to synchronously collect sensor temperature signals and furnace pressure, airflow and humidity parameters, and real-time display after preprocessing, including the following steps: S201: Signal conditioning, the resistance signal of the temperature sensor is amplified and filtered by the signal conditioning module, the signal conditioning module is specifically: a high-precision programmable constant current source with model AD590 is used to provide a constant current of 100 mu A to the temperature sensor, which is used to ensure the measurement stability of the resistance signal; differential amplification: an instrument amplifier with model AD620 is used to differentially amplify the voltage signal between the two ends of the temperature sensor, and the amplification factor is set to 1000 times, which is used to amplify the millivolt level signal to the volt level signal; filtering: a second-order low-pass filter is used, and the cutoff frequency is set to 10 Hz, which is used to filter out the high-frequency noise generated by the electromagnetic interference in the special ceramic sintering furnace; S202: Data acquisition, a high-speed synchronous data acquisition card with model NIPCIe-6363 is selected to simultaneously acquire the voltage signals of all temperature sensors, and the high-speed synchronous data acquisition card is connected with the host computer through a PCIe interface to realize high-speed data transmission; S203: The environmental parameters in the special ceramic sintering furnace are obtained by the environmental data parameter sensor, and three environmental parameter monitoring points are set in the furnace, which are located at the top, middle and bottom of the furnace, and the following sensors are installed at each monitoring point: Atmosphere pressure sensor: a piezoelectric pressure sensor is used, with a measurement range of 0-1 MPa and an accuracy of ±0.1% FS, which is used to monitor the pressure change of the protective atmosphere such as nitrogen and argon in the furnace; air flow velocity sensor: a hot-wire anemometer is used, with a measurement range of 0-10 m / s and an accuracy of ±0.05 m / s, which is used to monitor the flow velocity of the air flow in the furnace; humidity sensor: a ceramic humidity sensor is used, with a measurement range of 0-100% RH and an accuracy of ±2% RH, which is used to monitor the residual moisture content in the furnace; The output signals of the above environmental parameter sensors are connected to the remaining analog input channels of the data acquisition card after signal conditioning, which are used to realize synchronous acquisition with the temperature signal, and the sampling interval is also 10 ms; S204: Host computer, through the host computer: control the data acquisition card to collect temperature signals and environmental parameter signals at the set sampling interval, and store the original data to the hard disk; data preprocessing: the collected temperature signals are denoised by wavelet threshold denoising algorithm, and the environmental parameter signals are smoothed by moving average method; real-time display: real-time display of the temperature value and environmental parameter value of each temperature sensor node, and display in the form of numerical table and curve.

[0024] S3: Temperature field reconstruction and prediction model based on multi-physical field coupling, combining discrete temperature data and environmental parameters, reconstructing the global temperature field by interpolation. Use LSTM model to predict future temperature trends based on historical data, realize dynamic prediction, including the following steps: S301: Temperature field reconstruction model, formula one is: ; in is the temperature value of any spatial point inside the special ceramic body; is the three-dimensional rectangular coordinate of the space point; For the The weight coefficient of each temperature sensor node is used to measure the effect of the node on the The contribution of temperature is solved by the least square method; Index of the temperature sensor in the radial direction R0-R5 ; Index of the temperature sensor in the axial direction H0-H5 ; Index of the temperature sensor in the circumferential direction C0-C5 ; The target space point With the sensor nodes The Euclidean distance between is the radial function The core independent variable, Use Gaussian function; For the The three-dimensional rectangular coordinates of the sensor nodes; It is the environmental parameter correction factor, used to correct the atmosphere pressure in the furnace , air flow velocity Impact on temperature field distribution; air flow velocity When the airflow increases, the convective heat transfer is enhanced and the local temperature gradient is reduced, so the airflow correction coefficient is positive, and the influence of the two on the temperature field is reflected through linear correction, in which the atmosphere pressure : High pressure will increase the density of gas molecules and improve the efficiency of heat conduction; low pressure will inhibit heat transfer and have the following effects on the temperature field: the temperature gradient in the high pressure area will decrease and the uniformity will be improved; the low pressure area will easily form local hot spots / cold spots; the air flow speed : Forced convection accelerates heat migration, reduces boundary layer thickness, and promotes temperature uniformity; it has the following effects on the temperature field: temperature fluctuations in high-speed airflow areas are reduced, but turbulence may cause instantaneous unevenness; humidity : Water vapor phase change (vaporization / condensation) absorbs or releases latent heat, which will change the local heat capacity. The temperature change in the high humidity area is delayed, and the dry area responds faster; in: ; in The reference value of standard atmospheric pressure is 0.1MPa; The reference air velocity is 0.5m / s; The pressure correction factor is 0.03; The airflow correction factor is 0.07, which is obtained by 、 The temperature deviation is measured and obtained by least square fitting; Its logic is atmospheric pressure When the pressure increases, the frequency of gas molecule collisions in the furnace increases, the heat conduction efficiency improves, and the temperature field distribution becomes more uniform, so the pressure correction coefficient is positive; The pressure term is in logarithmic form because the heat transfer coefficient is logarithmically related to the pressure; the airflow term is in exponential decay form because the convective heat transfer is It tends to be saturated after ≥2m / s; both corrections are constrained by the correction factor of 1 at the baseline value to ensure that no additional deviation is introduced under normal working conditions; According to the formula, the temperature value of any spatial point inside the special ceramic body can be calculated, and then the three-dimensional visualization reconstruction of the temperature field can be realized through Matlab software to intuitively display the distribution uniformity of the temperature field; S302: Based on the LSTM model, a temperature field dynamic prediction model is constructed by combining historical temperature data and environmental parameter data. The input parameters of the LSTM model include: Historical temperature data: The temperature values ​​of each temperature sensor node at the past 10 sampling moments, totaling 100ms, with a total of 216 × 10 = 2160 features; Historical environmental parameter data: atmospheric pressure at the past 10 sampling moments , air flow velocity ,humidity , a total of 3×10=30 features; Current sintering stage parameters: current heating rate , unit is ℃ / min, current holding time , the unit is min, a total of 2 features; The output parameters of the model are: the temperature prediction value of each temperature sensor node for the next five sampling moments, a total of 50ms, a total of 216 × 5 = 1080 features; The structure of the LSTM model includes: Input layer: 2160+30+2=2192 neurons, corresponding to the input parameters; hidden layer: 3-layer LSTM network, each layer contains 512 neurons, using ReLU activation function; dropout layer: dropout probability is set to 0.2 to prevent model overfitting; fully connected layer: 2-layer fully connected network, the first layer contains 2048 neurons, the second layer contains 1080 neurons; output layer: 1080 neurons, corresponding to the output temperature prediction value; The training loss function of the LSTM model uses the root mean square error, and the formula 2 is: ; in is the average deviation between all predicted data and true data; is the number of training samples; is the number of temperature data points for each sample; For the The index of the training samples ; For the The index of the temperature data point ; For the Of the training samples The predicted temperature value for each temperature data point; For the Of the training samples The actual temperature value of each temperature data point; The LSTM model was trained using the Adam optimizer with 500 training iterations and a learning rate of 0.001. When the RMSE was less than 0.3°C, the LSTM model training was completed. The trained model can be embedded in a host computer to receive current data in real time and output temperature field prediction results for the next 50ms, thus achieving dynamic early warning of the temperature field.

[0025] S4: Real-time temperature field monitoring and abnormal warning. The integrated data displays the temperature field distribution and trend in real time. The abnormalities are identified through temperature gradient, deviation and predicted value, and graded warning is issued. The specific steps include: S401: Real-time monitoring, specifically including: Synchronous data acquisition: The original resistance signals of 216 temperature sensor nodes are collected at a fixed sampling interval of 20ms and converted into real-time temperature values ​​through the signal conditioning module. , here Represents the current sampling moment, Represents the temperature and represents the sensor node number; synchronously collects the atmosphere pressure of 3 monitoring points in the furnace , air flow velocity ,humidity ,All the collected data are transmitted to the host computer for temporary ,buffering; Real-time reconstruction of temperature field: Call formula 1 to convert the current moment As The measured temperature of the node is compared with 、 Substitute into formula 1 to calculate any point in the three-dimensional space of the special ceramic body Temperature value ; Call the Matlab engine through LabVIEW and Convert to a three-dimensional temperature field cloud map with a cloud map update frequency of 10Hz; Temperature dynamic prediction: input the past 10 time points of , , and the current process parameters into the LSTM model, and output the predicted temperature of the next 5 time points , and display the real-time and predicted trends in the host computer with virtual-real curves, wherein is the time point; S402: Real-time anomaly judgment based on the reconstructed formula one of S401 , if the temperature gradient of any adjacent point > 5℃ / mm, the mean deviation of a node and its adjacent nodes exceeds 8℃, trigger a red alarm, and display the abnormal node and its , wherein is the temperature difference between the space point and its adjacent points, is the straight-line distance between the two points; Predictive anomaly judgment, if exceeds the sintering process set temperature ±10℃, the predicted rate > 60℃ / min, trigger a yellow warning, and mark the warning time and its .

[0026] Example two: A temperature field real-time monitoring system based on special ceramic sintering, comprising a multi-parameter data acquisition module, a temperature field reconstruction module, a temperature field dynamic prediction module, a real-time monitoring and anomaly warning module, the multi-parameter data acquisition module specifically comprising: Signal conditioning submodule: used for converting the resistance change of the temperature sensor into a voltage signal, then differentially amplifying the millivolt-level voltage signal by 1000 times to the volt-level through an instrument amplifier, and then filtering out the high-frequency electromagnetic interference in the furnace through a second-order 10Hz low-pass filter to ensure that the signal-to-noise ratio of the temperature sensor is greater than 60dB; Temperature data acquisition submodule: collect voltage signals through a high-speed synchronous data acquisition card, convert them into real-time temperatures of corresponding nodes, and transmit them to the host computer through a PCIe interface at high speed; Environment parameter acquisition submodule: set up 1 monitoring point at the top, middle and bottom of the sintering furnace respectively, and install atmosphere pressure sensor, airflow speed sensor and humidity sensor respectively for synchronous collection of furnace atmosphere pressure, airflow speed and humidity; Host computer data processing submodule: run the program developed by LabVIEW on the industrial control computer, pre-process the collected real-time temperature values using wavelet threshold denoising algorithm, and simultaneously display the real-time temperature values of each node and the numerical values and change curves of the environmental parameters on the interface.

[0027] The temperature field reconstruction module realizes continuous spatial distribution display of the temperature field inside the blank based on the discrete real-time temperature values and the environmental parameters, specifically: Convert the numbers of 216 sensor nodes into corresponding three-dimensional rectangular coordinates , to determine the spatial position of each node; then, according to the real-time temperature values of each node, the weight coefficients required for spatial interpolation are calculated by the least squares method to ensure that the interpolation results accurately reflect the measured temperature distribution; then, the collected 、 environmental parameter correction factors are calculated to compensate for the influence of atmospheric pressure and airflow speed on heat conduction efficiency; then, the temperature of any spatial point inside the blank is interpolated and calculated by a Gaussian function to obtain global continuous temperature distribution data; finally, the temperature distribution data is converted into a three-dimensional temperature field cloud chart by calling the Matlab engine, which is updated in real time on the host computer at a frequency of 10 Hz to intuitively display the uniformity of the temperature field inside the blank.

[0028] The temperature field dynamic prediction module is used to build a complete input feature system and automatically extract historical key data, specifically the real-time temperature data of 216 sensor nodes, the atmospheric pressure, airflow speed and humidity in the sintering furnace at the same period within the past 10 sampling times, and the key process parameters of the current sintering stage, forming a multi-dimensional input feature set covering temperature, environment and process.

[0029] The real-time monitoring and abnormal warning module specifically includes: The real-time monitoring submodule is used to receive the global continuous temperature distribution data from the temperature field reconstruction module, generate a three-dimensional temperature field cloud chart, and clearly present the temperature gradient and distribution uniformity inside the blank from the surface to the core; at the same time, the real-time temperature data, environmental parameters from the multi-parameter data acquisition module, and the predicted temperature data from the temperature field dynamic prediction module are integrated and displayed in the same host computer interface in the form of subareas: including the list of real-time temperature values of each node, the comparison chart of real-time temperature change curve and prediction curve, and the real-time value and trend chart of environmental parameters; It also includes an abnormal warning submodule, which is used to automatically calculate the temperature gradient of any adjacent spatial point in the three-dimensional temperature field. If the gradient exceeds 5℃ / mm and the deviation of the real-time temperature value of a node from the average temperature of the surrounding six adjacent nodes exceeds 8℃, a red first-level alarm is triggered immediately, and the abnormal node number, corresponding spatial coordinates and real-time temperature value are marked; The predicted temperature of each node is automatically compared with the set temperature range, and if the predicted temperature exceeds the range and the heating rate in the predicted period exceeds the upper limit of the safe heating rate, a yellow second-level warning is triggered: the warning time, corresponding node number and predicted temperature are marked to remind the operator to adjust the process parameters in advance to avoid abnormalities.

[0030] The above shows and describes the basic principles and main features of the present application and the advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for real-time monitoring of temperature field based on sintering of special ceramics, characterized in that: The method comprises the following operation steps: S1: processing of the sensor array, preparing a temperature sensor with a special ceramic as a base body and doped with conductive particles, embedding the ceramic body in a three-dimensional grid layout to form an array of 216 nodes, and realizing global temperature collection; S2: construction of multi-parameter data acquisition, building a system containing signal conditioning, acquisition card, environmental sensor and host computer, synchronously collecting sensor temperature signal and furnace pressure, airflow and humidity parameters, and displaying in real time after preprocessing; S3: temperature field reconstruction and prediction model based on multi-physical field coupling, combining discrete temperature data and environmental parameters, reconstructing the global temperature field through interpolation; using the LSTM model to predict the future temperature trend based on historical data, and realizing dynamic prediction; S4: real-time monitoring and abnormal warning of temperature field, integrating data to display temperature field distribution and trend in real time, identifying abnormalities through temperature gradient, deviation and predicted value, and grading warning.

2. The method for real-time monitoring of temperature field based on special ceramic sintering according to claim 1, characterized in that: In S1, the following steps are specifically included: S101: selection of temperature sensor, prepared with special ceramic powder as base body and doped with nano-sized conductive particles, its working principle is: based on the contact resistance change rule of special ceramic powder base body and nano-sized conductive particles at high temperature: when the temperature rises, the lattice vibration of the special ceramic powder base body intensifies, the contact resistance between the nano-sized conductive particles changes linearly with the temperature rise, and the temperature can be inversely calculated by measuring the resistance value; S102: presetting of sensor array, according to the shape and size of the special ceramic body to be sintered, the spatial layout of the sensor array is designed in a three-dimensional grid layout, the diameter and height of the special ceramic body are equally divided into six parts as the basis for the layout of the sensor array, specifically as follows: Radial direction: along the radius direction of the body, 1 sensor layer is set at each division, a total of 6 layers, respectively marked as R0, R1, R2, R3, R4, R5, wherein R0 is the center layer and R5 is the surface layer; Axial direction: along the height direction of the body, 1 sensor row is set at each division, a total of 6 rows, respectively marked as H0, H1, H2, H3, H4, H5; Circumferential direction: along the circumference of the special ceramic body, 1 sensor column is set every 60°, a total of 6 columns, respectively marked as C0, C1, C2, C3, C4, C5; Through the above layout, 216 sensor nodes are formed in the body, each node corresponds to a unique spatial coordinate, realizing global coverage of the internal and surface temperature field of the special ceramic body.

3. The method according to claim 1, wherein the method is characterized by: In S2, the following steps are specifically included: S201: signal conditioning, the resistance signal of the temperature sensor is amplified and filtered through the signal conditioning module, the signal conditioning module specifically includes: a high-precision programmable constant current source is used to provide a constant current of 100 mu A to the temperature sensor; differential amplification: an instrument amplifier with model AD620 is used to differentially amplify the voltage signal between the two ends of the temperature sensor, and the amplification multiple is set to 1000 times; filtering: a second-order low-pass filter is used, and the cutoff frequency is set to 10 Hz; S202: Data acquisition, select the model of high-speed synchronous data acquisition card NIPCIe-6363, used to collect voltage signals of all temperature sensors at the same time; S203: Obtain the environmental parameters in the special ceramic sintering furnace through the environmental data parameter sensor, set three environmental parameter monitoring points in the furnace, located at the top, middle and bottom of the furnace respectively, and install the following sensors at each monitoring point: Atmosphere pressure sensor: piezoelectric pressure sensor is used to monitor the pressure change of the protective atmosphere in the furnace; airflow velocity sensor: hot-wire anemometer is used to monitor the flow velocity of the airflow in the furnace; humidity sensor: ceramic humidity sensor is used to monitor the residual moisture content in the furnace; S204: Host computer, through the host computer: control the data acquisition card to collect temperature signals and environmental parameter signals at the set sampling interval, and store the raw data to the hard disk; data preprocessing: denoising processing for the collected temperature signals, and smoothing processing for the environmental parameter signals; real-time display: real-time display of the temperature value and environmental parameter value of each temperature sensor node, and display in the form of numerical table and curve.

4. The method according to claim 2, wherein the method is characterized by: In the S3, the following steps are specifically included: S301: Temperature field reconstruction model, formula one is: ； wherein T is the temperature value of an arbitrary spatial point inside the special ceramic body; is the three-dimensional Cartesian coordinate of the spatial point; is the weight coefficient of the th temperature sensor node, which measures the contribution degree of the node to the temperature; is the index of the radial direction R0-R5 of the temperature sensor; ; is the index of the axial direction H0-H5 of the temperature sensor; ; is the index of the circumferential direction C0-C5 of the temperature sensor; is the Euclidean distance between the target spatial point and the th sensor node, which is the core independent variable of the radial function ; is the three-dimensional Cartesian coordinate of the th sensor node; is the environmental parameter correction factor, which is used to correct the influence of the furnace atmosphere pressure and the gas flow speed on the temperature field distribution; According to the formula, the temperature value of any spatial point inside the special ceramic body can be calculated, and the three-dimensional visualization reconstruction of the temperature field is realized through Matlab software, which intuitively displays the uniformity of the temperature field distribution; S302: Based on the LSTM model, combine historical temperature data and environmental parameter data to build a temperature field dynamic prediction model, the input parameters of the LSTM model include: Historical temperature data: temperature values of each temperature sensor node at the past 10 sampling times, a total of 100ms, a total of 216x10=2160 features; Historical environmental parameter data: atmospheric pressure at the last 10 sampling times , air flow speed , humidity , a total of 3x10=30 features; Current sintering stage parameters: current heating rate , current soaking time , total 2 features; The output parameters of the model are: temperature prediction values of each temperature sensor node at the future 5 sampling times, a total of 50ms, a total of 216x5=1080 features.

5. The method according to claim 4, wherein the method is based on real-time monitoring of the temperature field of the sintering of the special ceramic. The training loss function of the LSTM model adopts root mean square error, formula two is: ； wherein is the average deviation of all predicted data from real data; is the number of training samples; is the number of temperature data points per sample; is the index of the th training sample ; is the index of the th temperature data point ; is the predicted temperature value of the th temperature data point in the th training sample; is the real temperature value of the th temperature data point in the th training sample; Through the Adam optimizer, the LSTM model is trained for 500 iterations, and the learning rate is set to 0.

001. When the RMSE is less than 0.3℃, the LSTM model training is completed. The trained model can be embedded into the host computer, which can receive current data and output temperature field prediction results for the next 50ms in real time, realizing dynamic warning of the temperature field.

6. The method according to claim 5, wherein: In the S4, the following steps are specifically included: S401: Real-time monitoring, specifically including: Data synchronization acquisition: collect the original resistance signal of 216 temperature sensor nodes at a fixed sampling interval of 20 ms, and convert it into real-time temperature value through the signal conditioning module , where represents the current sampling time, represents the temperature of the sensor node number; Synchronously collect the atmosphere pressure of three monitoring points in the furnace , airflow speed , humidity , all collected data is transmitted to the upper computer temporary buffer; Real-time reconstruction of temperature field: Call formula 1 to convert the current moment As The measured temperature of the node is compared with 、 Substitute into formula 1 to calculate any point in the three-dimensional space of the special ceramic body Temperature value ; Call the Matlab engine through LabVIEW and Convert to a three-dimensional temperature field cloud map with a cloud map update frequency of 10Hz; Temperature dynamic prediction: input the past 10 time points of , , and the current process parameters into the LSTM model, output the predicted temperature of the next 5 time points , and display the real-time and predicted trends in the host computer with virtual-real curve comparison, where is the time point; S402: Real-time anomaly judgment, based on the reconstructed formula one in S401 If the temperature gradient of any adjacent points > 5℃ / mm, the mean deviation of a node and its adjacent nodes exceeds 8℃, trigger red alarm, and display the abnormal node and its , where is the temperature difference between the space point and its adjacent points, is the straight-line distance between the two points; Predictive anomaly judgment, if Exceeding sintering process set temperature ±10℃, predicted rate >60℃ / min, triggering yellow warning, marking the warning time and its .

7. A temperature field real-time monitoring system based on special ceramic sintering, using a temperature field real-time monitoring method based on special ceramic sintering according to any one of claims 1-6, comprising a multi-parameter data acquisition module, a temperature field reconstruction module, a temperature field dynamic prediction module, a real-time monitoring and abnormal early warning module, characterized in that: The multi-parameter data acquisition module specifically includes: Signal conditioning sub-module: used to convert the resistance change of the temperature sensor into a voltage signal, and then amplify the millivolt-level voltage signal by 1000 times to the volt-level through an instrument amplifier, and then filter out the high-frequency electromagnetic interference in the furnace through a second-order 10Hz low-pass filter, to ensure that the signal-to-noise ratio of the temperature sensor is greater than 60dB; Temperature data acquisition sub-module: collect voltage signals through high-speed synchronous data acquisition card, convert them into real-time temperature of corresponding nodes, and transmit them to the host computer through PCIe interface at high speed; The environmental parameter acquisition submodule: one monitoring point is arranged at each of the top, middle and bottom of the sintering furnace, and an atmosphere pressure sensor, an airflow velocity sensor and a humidity sensor are installed respectively, for synchronously collecting the atmosphere pressure, airflow velocity and humidity in the furnace; The host computer data processing submodule: the program developed by LabVIEW is run on the industrial control computer, the real-time temperature values collected are preprocessed by using the wavelet threshold denoising algorithm, and the real-time temperature values of each node, the values and change curves of the environmental parameters are displayed on the interface in real time.

8. The temperature field real-time monitoring system based on special ceramic sintering according to claim 7, characterized in that: The temperature field reconstruction module is based on the discrete real-time temperature values and environmental parameters, and realizes continuous spatial distribution display of the temperature field inside the green body.

9. The temperature field real-time monitoring system based on special ceramic sintering according to claim 7, characterized in that: The temperature field dynamic prediction module is used for constructing a complete input feature system, automatically extracting historical key data, specifically, the real-time temperature data of 216 sensor nodes in the past 10 sampling time points, the atmosphere pressure, airflow velocity and humidity in the sintering furnace at the same period, and the key process parameters of the current sintering stage are integrated to form a multi-dimensional input feature set covering temperature, environment and process.

10. The temperature field real-time monitoring system based on special ceramic sintering according to claim 7, characterized in that: The real-time monitoring and abnormal early warning module specifically includes: The real-time monitoring submodule: used for receiving the global continuous temperature distribution data of the temperature field reconstruction module, generating a three-dimensional temperature field cloud picture; meanwhile, the real-time temperature data and environmental parameters of the multi-parameter data acquisition module and the predicted temperature data of the temperature field dynamic prediction module are integrated, and the same host computer interface is displayed in the form of subareas: including the real-time temperature value list of each node, the comparison chart of real-time temperature change curve and prediction curve, and the real-time value and change trend chart of environmental parameters; The abnormal early warning submodule is further included, which is used for automatically calculating the temperature gradient of any adjacent spatial points in the three-dimensional temperature field, if the gradient exceeds 5℃ / mm, and the deviation of the real-time temperature value of a node from the average temperature of the surrounding six adjacent nodes exceeds 8℃, a red first-level alarm is triggered immediately, and the abnormal node number, corresponding spatial coordinates and real-time temperature value are labeled; The predicted temperature of each node is automatically compared with the process set temperature range, if the predicted temperature exceeds the range, and the heating rate in the predicted period exceeds the upper limit of the safe heating rate, a yellow second-level early warning is triggered: the warning time, corresponding node number and predicted temperature are labeled, reminding the operator to adjust the process parameters in advance to avoid abnormality.

Images