Infrared radiation temperature measurement system based on point detector

Abstract

The invention relates to the technical field of infrared temperature measurement, and particularly discloses an infrared radiation temperature measurement system based on a point detector, which is used for solving the problems that in a 200-1200 DEG C heating furnace, infrared temperature measurement of an existing point detector cannot dynamically select a point detector and a filtering wave band along with the state of a hearth, accurate measurement, quick response and long-term calibration of a wide temperature region are difficult to meet at the same time, and the accuracy of temperature measurement is low. The device comprises an optical platform, an optical path switching unit is arranged on the optical platform, the optical path switching unit is connected with a plurality of optical inlets, the optical inlets are converged to a main optical axis, and a dynamic narrow-band filter and a point detector selection module are sequentially arranged on the main optical axis; through adaptive selection of the dynamic filtering wave band and the point detector, multi-spectrum fusion inversion and an online self-calibration mechanism, the temperature measurement error and drift are reduced under the conditions of wide temperature range, high flue gas interference and long-term continuous operation, and the precision, response speed and reliability of temperature measurement of the heating furnace are improved.

Description

Technical Field

[0001] This invention relates to the field of infrared temperature measurement technology, and in particular to an infrared radiation temperature measurement system based on a point detector. Background Technology

[0002] Infrared radiation thermometry has significant advantages such as fast response speed, non-interference with the measured target, and adaptability to harsh environments. However, current infrared thermometry technology based on point detectors still faces problems such as insufficient temperature measurement accuracy, poor environmental adaptability, and slow dynamic response speed. Especially in intelligent manufacturing, semiconductor processing, and online monitoring of power equipment, accurate temperature measurement of small targets is crucial for process control and equipment safety. For industrial heating furnaces with operating temperatures ranging from 200 to 1200°C, the wide temperature fluctuation range of the furnace makes infrared temperature measurement by a single infrared point detector difficult. Moreover, the temperature measurement accuracy and response time cannot meet the needs of real-time industrial monitoring. It is impossible to dynamically select the type of infrared point detector and the parameters of the filter according to the furnace conditions. At the same time, it is difficult to suppress temperature drift and calibration failure during long-term operation, which affects the temperature measurement reliability of the heating furnace. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides an infrared radiation temperature measurement system based on a point detector. By adaptively selecting a narrow-band filter channel based on signal-to-noise ratio and temperature uncertainty, and combining deep learning for multi-spectral temperature-emissivity joint inversion, the system maximizes the utilization rate of multi-spectral information within a limited measurement time. It selects appropriate infrared point detectors based on different temperature ranges and combines blackbody calibration to solve the calibration failure problem, thereby reducing temperature measurement uncertainty and system error under complex furnace conditions.

[0004] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: An infrared radiation temperature measurement system based on a point detector includes an optical platform with an optical path switching unit connected to an optical inlet. The optical inlet converges to a main optical axis. A dynamic narrowband filter and a point detector selection module are sequentially arranged on the main optical axis. The point detector selection module is connected to a signal conditioning module, which is further connected to a multispectral fusion analysis module, a spatial uniformity correction module, a blackbody calibration module, and a real-time temperature analysis module. Calibration information and uniformity correction coefficients output from the blackbody calibration module and the spatial uniformity correction module are input to a temperature inversion module. The outputs of the temperature inversion module and the multispectral fusion analysis module are both sent to the real-time temperature analysis module, which is connected to a multispectral channel output terminal. The dynamic narrowband filter is connected to a filter control unit, which controls the dynamic narrowband filter to switch between multiple narrowband channels within a preset wavelength range. The system acquires the signal-to-noise ratio (SNR) of each narrowband channel signal and receives the temperature estimation results and corresponding uncertainty indices from the temperature inversion module. Based on measurement time constraints and the temperature uncertainty cost function, it adaptively selects the target center wavelength set, passband width, channel scanning order, and dwell time parameters within the current measurement cycle to minimize the temperature estimation uncertainty of the measured area within the measurement cycle. This allows for the adjustment of the narrowband channels of the dynamic narrowband filter according to the specific furnace temperature. The temperature inversion module includes a deep learning-based temperature inversion model. The model takes the spatially corrected multi-band radiation characteristics output by the spatial uniformity correction module, the multi-point calibration data output by the blackbody calibration module, and the multi-band radiation intensity, ratio, and difference characteristics output by the multispectral fusion analysis module as inputs. It then jointly outputs the measured target temperature estimate, the equivalent emissivity estimate, and the temperature estimation uncertainty index.

[0005] As a further aspect of this invention, the optical entry point is used to: connect with the temperature measurement window of the furnace under test to obtain a first radiation signal; connect with the radiation outlet of the blackbody furnace to obtain a second radiation signal; connect with a medium-sized uniform light source system to obtain a third radiation signal; and connect with a tunable laser source to obtain a fourth radiation signal. This addresses the problems of existing industrial furnace temperature measurement systems, which rely on separate optical paths or frequent disassembly and reassembly of optical components for online temperature measurement, blackbody temperature scale traceability, spatial uniformity and spectral response calibration, and tunable laser testing. These issues lead to inconsistencies in the incident angles, spot sizes, and optical path structures of various radiation sources, making it difficult to maintain coaxiality and field of view between calibration conditions and actual temperature measurement conditions. Furthermore, the calibration results are not traceable. Additionally, the coupling of the medium-sized uniform light source and the tunable laser source with the main optical axis lacks standardized interfaces and switching mechanisms. During on-site operation, issues such as misinsertion, crosstalk, or incomplete switching can easily occur, increasing the difficulty of system debugging and introducing new sources of error. This makes it difficult to simultaneously meet the practical needs of multi-source sharing, optical path isolation, and rapid switching. By setting up optical entry points and optical path switching units, the system can quickly switch between multiple operating modes without replacing optical components or adjusting probe positions. All radiation sources complete temperature measurement, calibration, and testing under a unified principal optical axis and a unified field of view, ensuring consistency in optical path geometry and incident conditions. This reduces system errors introduced by optical path reconstruction, improves the consistency and traceability of calibration results and online temperature measurement results, enhances the safety and reliability of on-site operation, shortens the time for mode switching and realignment, and improves the engineering applicability and long-term stable operation capability of the entire temperature measurement system.

[0006] As a further aspect of the present invention, the system's operating stages include a spatial uniformity and spectral response stage, a blackbody calibration stage, and an online temperature measurement stage for the heating furnace. During the online temperature measurement stage, an online self-calibration sub-stage is inserted according to a preset self-calibration cycle. In this sub-stage, the optical path switching unit switches the main optical axis from the heating furnace under test to a blackbody furnace or a medium-sized uniform light source system without altering the overall system operating state. Within a time period not exceeding a preset self-calibration duration threshold, a calibration data frame sequence is acquired. This sequence is then combined with the accumulated temperature deviation statistics within a preset drift monitoring time window to correct parameter drift generated by the uniformity correction coefficients in the temperature inversion module and the spatial uniformity correction module within the drift monitoring time window. In existing temperature measurement systems, the detector's field-of-view spatial non-uniformity error and spectral response mismatch error are typically mixed with the temperature calibration process. The blackbody calibration results cannot distinguish between various error sources from the field-of-view response, spectral response, and temperature scale, resulting in an untraceable calibration link. Infrared measurement... Temperature measurement systems often rely on one-time factory calibration or periodic manual shutdown calibration, making it difficult to correct slow parameter drifts caused by detector gain and bias drift, optical window contamination, narrowband filter transmittance changes, and environmental conditions. During long-term continuous operation, absolute temperature deviations accumulate over time and exceed the allowable range. Traditional solutions often require temporary optical path reconfiguration for compensation calibration, which interrupts online temperature measurement of the furnace, impacting continuous production and introducing new optical path errors during frequent disassembly and realignment. By setting up the three working stages and embedding a short-term online self-calibration sub-stage within the online temperature measurement stage, spatial uniformity and spectral response errors are first modeled separately in independent stages. The temperature scale is established based on the corrected response. Subsequently, without changing the overall system operating state, calibration data frames from a blackbody or uniform light source are periodically used to correct the time drift of the aforementioned model. This ensures continuous online temperature measurement of the furnace while achieving decomposable error sources, traceable temperature calibration relationships, and controllable parameter drift during the temperature measurement process.

[0007] As a further aspect of the present invention, in the spatial uniformity and spectral response stage, the optical path switching unit introduces the third radiation signal emitted by the medium-sized uniform light source system into the main optical axis through the optical entrance, and the optical path switching unit introduces the fourth radiation signal emitted by the tunable laser light source into the main optical axis through the optical entrance. The third and fourth radiation signals are sent to the spatial uniformity correction module and the multispectral fusion analysis module respectively through the infrared point detector selected by the dynamic narrowband filter and the point detector selection module, the signal conditioning module, and the ADC conversion module. The spatial uniformity correction module is used to analyze the response changes of the optical platform and the selected infrared point detector at different incident angles and different incident positions in the field of view, establish a spatial uniformity correction coefficient field, and map the coefficient field to the pixel coordinates of the target area of ​​the tested heating furnace. The multispectral fusion analysis module obtains the response curves at each wavelength in the target band and the relative sensitivity of each spectral channel by scanning the fourth radiation signal sent by the tunable laser light source point by point in the target band according to a preset wavelength sequence, and generates spectral response weights and band correction factors for multispectral fusion.

[0008] As a further aspect of the present invention, during the blackbody calibration stage, the radiation outlet of the blackbody furnace transmits the second radiation signal to the optical path switching unit. The optical path switching unit couples the second radiation signal to the optical entrance and cuts off the first radiation signal. The second radiation signal is then sent to the blackbody calibration module and the temperature inversion module via a dynamic narrowband filter, a selected infrared point detector, a signal conditioning module, and an ADC conversion module. The blackbody calibration module is used to collect the output signal of the infrared point detector under the multispectral channel after intermediate frequency chopping when the temperature of the blackbody furnace is within a preset calibration range at multiple temperature points. It establishes a multi-channel voltage-temperature correspondence table and constructs an analytical relationship between radiation energy and temperature based on Planck's law and Stefan-Boltzmann's law. During the blackbody calibration stage, the temperature inversion module performs offline pre-training of the deep learning temperature inversion model based on the calibration relationship provided by the blackbody calibration module and the spectral response weights provided by the multispectral fusion analysis module, obtaining the initial temperature estimation function for different band combinations.

[0009] As a further aspect of the present invention, during the online temperature measurement stage of the heating furnace, the temperature measurement window of the heating furnace under test transmits the first radiation signal to the optical entrance. The optical path switching unit keeps the optical entrance connected to the main optical axis to receive the first radiation signal and cuts off the second radiation signal. The first radiation signal is sent to the temperature inversion module and the multispectral fusion analysis module through the dynamic narrowband filter, the selected infrared point detector, and the signal conditioning module. The real-time temperature analysis module outputs the real-time temperature field and the corresponding uncertainty field of the target area of ​​the heating furnace under test under the collaborative analysis of the spatial uniformity correction coefficient output by the spatial uniformity correction module, the multi-band radiation characteristics output by the multispectral fusion analysis module, and the multi-point calibration relationship provided by the blackbody calibration module. The real-time temperature analysis module returns the temperature uncertainty field to the filter control unit to update the channel configuration strategy for the next measurement cycle.

[0010] As a further aspect of the present invention, the filter control unit includes an uncertainty assessment subunit and a channel optimization decision subunit. The uncertainty assessment subunit is used to calculate the temperature estimation cost function of the current channel combination based on the temperature estimation result and its uncertainty index output by the temperature inversion module, the signal-to-noise ratio index of each narrowband channel, and the temperature change rate output by the real-time temperature analysis module. The channel optimization decision subunit adopts a pre-trained offline deep learning decision model or reinforcement learning decision model, taking the working stage command, channel signal-to-noise ratio index, temperature change rate, and current cost function as input, and outputs the target center wavelength set, passband width, channel scanning order, and dwell time parameters that meet the measurement time budget constraint, so as to minimize the expected value of the predicted temperature estimation uncertainty for the next measurement cycle, and transmits the parameters to the dynamic narrowband filter to achieve active adaptive band selection.

[0011] As a further embodiment of the present invention, the system further includes a path medium parameter inversion module. The signal conditioning module is connected to the multispectral fusion analysis module, spatial uniformity correction module, blackbody calibration module, and real-time temperature analysis module via the ADC conversion module. The path medium parameter inversion module is connected to the tunable laser source, the signal conditioning module, and the ADC conversion module. It is used to control the tunable laser source to select several narrow wavelengths containing the absorption lines of combustion gas and water vapor on the field of view path of the heated furnace under test for scanning during the online temperature measurement stage or the online self-calibration sub-stage of the heated furnace, collect the corresponding laser transmission signals, and obtain the path average gas temperature, gas concentration, and path transmittance parameters based on the gas absorption spectrum model. The parameters are then input into the temperature inversion module. While outputting the surface temperature and equivalent emissivity of the heated furnace under test, the temperature inversion module substitutes the path medium parameters into the radiative transfer equation to separate and compensate for the self-radiation and absorption effects of the combustion gas and dust in the furnace, thereby reducing the temperature measurement error under high gas absorption and high dust conditions.

[0012] As a further embodiment of the present invention, the system also includes an environmental parameter acquisition module and an online self-calibration management module. The environmental parameter acquisition module is used to collect ambient temperature, furnace pressure, combustion air distribution parameters, flue gas composition parameters, and optical window pollution level indicators around and inside the tested heating furnace. The online self-calibration management module is used to construct calibration drift indicators and spatial uniformity drift indicators based on the environmental parameters output by the environmental parameter acquisition module, the path transmittance change output by the path medium parameter inversion module, and the temperature deviation statistics output by the real-time temperature analysis module. When any indicator exceeds a preset threshold, the optical path switching unit is triggered to enter the online self-calibration sub-stage, and the calibration frames of the blackbody and uniform light source collected in the online self-calibration sub-stage are called to incrementally update the uniformity correction coefficient of the spatial uniformity correction module and the deep learning temperature inversion model parameters in the temperature inversion module, so that the absolute temperature measurement error of the system during long-term operation remains within the preset error limit.

[0013] As a further aspect of the present invention, the deep learning temperature inversion model in the temperature inversion module is obtained through offline training, which includes: During the blackbody calibration stage, a blackbody furnace is used as the test object. Multi-band radiation characteristics and corresponding reference temperatures are collected at multiple temperature set points and processed by the spatial uniformity correction module and the multispectral fusion analysis module to form a blackbody training sample set. During the material calibration stage, a variety of test materials and surface states are selected and placed in a temperature environment equivalent to a blackbody furnace. Multi-band radiation characteristics and their corresponding reference temperatures are collected, and spectral emission priors from the material emissivity database are introduced to form an extended training sample set. A composite loss function is constructed that simultaneously constrains the temperature inversion error, emissivity estimation error, and residuals of Planck's radiation law and radiative transfer equation. The network parameters of the deep learning temperature inversion model are iteratively optimized so that the model can output temperature estimates, equivalent emissivity estimates, and corresponding temperature estimation uncertainty indices that satisfy physical constraints under multiple material, surface, and operating conditions.

[0014] Compared with the prior art, the present invention has the following technical effects: This invention integrates a main optical axis for optical entry switching, a dynamic narrowband filter, and an optional infrared point detector on an optical platform. It also introduces multispectral fusion analysis, spatial uniformity correction, blackbody calibration, and deep learning temperature inversion modules into the signal link. Combined with path medium parameter inversion, environmental parameter acquisition, and an uncertainty-based adaptive optimization and online self-calibration management mechanism, the system maintains high accuracy, high sensitivity, and rapid response in temperature field measurement under complex conditions such as a wide temperature range of 200–1200℃, high flue gas content, window contamination, and slow operational drift. This reduces systematic errors and long-term drift caused by emissivity uncertainty, field-of-view non-uniformity, aging of optical and electronic devices, and gas absorption. It improves the stability and traceability of temperature measurement data, as well as the safety and energy efficiency of the heating furnace operation. Simultaneously, it considers adaptability to different furnace types, fuel ratios, and various typical workpiece material surface conditions, effectively avoiding product scrapping, energy waste, and the risk of accumulating hidden faults in key equipment and safety accidents caused by inaccurate temperature measurement. Attached Figure Description

[0015] Figure 1 This is a system block diagram of the present invention; Figure 2 This is a schematic diagram showing the composition and connection of multiple optical ports and optical path switching units of the present invention; Figure 3 This is a schematic diagram of the composition of the filter control unit of the present invention; Figure 4 This is the path medium parameter inversion module of the present invention. Detailed Implementation

[0016] The principles and features of the present invention are described below with reference to the accompanying drawings. The description is for illustrative purposes only and is not intended to limit the scope of the invention.

[0017] like Figure 1As shown, the present invention proposes an infrared radiation temperature measurement system based on a point detector, comprising an optical platform with an optical path switching unit connected to an optical entrance. The optical entrance converges to a main optical axis. A dynamic narrowband filter and a point detector selection module are sequentially arranged on the main optical axis. The point detector selection module is connected to a signal conditioning module, which is connected to a multispectral fusion analysis module, a spatial uniformity correction module, a blackbody calibration module, and a real-time temperature analysis module. The blackbody calibration module and the spatial uniformity correction module are connected to a temperature inversion module, and both the temperature inversion module and the multispectral fusion analysis module are connected to the real-time temperature analysis module. The dynamic narrowband filter is connected to a filter control unit. The filter control unit switches the channels of the dynamic narrowband filter within a preset band, obtains the signal-to-noise ratio, and receives the output of the temperature inversion module. Based on the measurement time constraint and the temperature uncertainty cost function, it selects the target center wavelength set, passband width, channel scanning order, and dwell time of the measurement period to minimize the uncertainty index. The temperature inversion module takes the spatially corrected multi-band radiation characteristics, multi-point calibration data, and multi-band radiation intensity, ratio, and difference characteristics as inputs. Based on the deep learning temperature inversion model, it outputs the temperature estimate, the equivalent emissivity estimate, and the temperature estimation uncertainty index.

[0018] It should be noted that, as Figure 2As shown, the optical inlet is fixed to the optical platform using an optical tube seat with a standard flange or threaded interface. An infrared window plate operating in the 3-5μm and / or 8-12μm band, such as an infrared window made of ZnSe, CaF2, or chalcogenide glass, is installed at the flange. Behind the window, a collimating optical head consisting of an infrared collimating lens and an adjustable bracket is installed, or an infrared fiber optic flange with an SMA905 interface and a fiber optic collimator are installed. This allows the collimated input of the temperature measuring window of the heated furnace under test, the radiation outlet of the blackbody furnace, the integrating sphere outlet of the medium-sized uniform light source system, and the tunable laser source to be collimated. The output end reliably docks with the optical entrance, thus forming a fixed and repeatable optical entrance mechanically and optically. The optical path switching unit is realized through an electric reflector assembly. The first working mode of the optical entrance is to use an electric tilting frame or electric linear slide with a high-reflectivity plane mirror installed. The tilting frame or slide has two repeatable positions. In the first position, the first radiation signal beam of the optical entrance is deflected and coupled to the principal optical axis. In the second position, the second radiation signal beam of the optical entrance is deflected and coupled to the principal optical axis, thereby realizing the first radiation signal of the optical entrance. The first and second radiation signals are mutually exclusive and switched. The second working mode is to drive the reflector to enter or exit the main optical axis after receiving the control signal. When the reflector of the optical path switching unit is in the working position, it guides the third radiation signal from the optical entrance to be coaxial with the main optical axis. The third working mode is that when the reflector is in the working position, it guides the fourth radiation signal from the optical entrance to be coaxial with the main optical axis. Each electric flip-up mirror frame, electric slide or electric rotary table is driven by a stepper motor or servo motor and connected to the system control unit through a motion controller to realize automatic, repeatable, and manual reinstallation-free switching of multiple entrance optical paths on a unified main optical axis. The optical entrance is used to connect with the temperature measuring window of the heating furnace under test to obtain the first radiation signal, to connect with the radiation outlet of the blackbody furnace to obtain the second radiation signal, to connect with the medium-sized uniform light source system to obtain the third radiation signal, and to connect with the tunable laser source to obtain the fourth radiation signal. The optical path switching unit is used to mutually exclusive switch the first and second radiation signals at the optical entrance, to control the third radiation signal to be introduced into the main optical axis through the optical entrance, and to control the fourth radiation signal to be introduced into the main optical axis through the optical entrance.

[0019] It should be noted that the system's operating stages include a spatial uniformity and spectral response stage, a blackbody calibration stage, and an online temperature measurement stage for the heating furnace. During the online temperature measurement stage, an online self-calibration sub-stage is inserted according to a preset self-calibration cycle. In this sub-stage, the optical path switching unit switches the main optical axis from the heating furnace under test to a blackbody furnace or a medium-sized uniform light source system without changing the overall operating state of the system. Within a time period not exceeding a preset self-calibration duration threshold, a calibration data frame sequence is collected. This calibration data frame sequence is then combined with the accumulated temperature deviation statistics within a preset drift monitoring time window to correct the parameter drift generated by the uniformity correction coefficients in the temperature inversion module and the spatial uniformity correction module within the drift monitoring time window.

[0020] The three working stages and the online self-calibration sub-stage are specifically implemented by the system control unit in collaboration with the optical path switching unit through a software state machine. Specifically, the system control unit is an industrial computer, embedded controller, or PLC, which has a preset working stage configuration table and state machine logic. In the configuration table, the spatial uniformity and spectral response stage, the blackbody calibration stage, and the online temperature measurement stage of the heating furnace are respectively bound to the corresponding optical path state, light source working mode, and data acquisition process. In the spatial uniformity and spectral response stage, the control unit sends a stage command to the optical path switching unit, causing the optical path switching unit to connect the medium-sized uniform light source system and the tunable laser light source to the main optical axis, and at the same time issue the integrating sphere brightness. The system collects laser scanning wavelength sequences and sampling timings, acquires corresponding data frames, and stores them in the spatial uniformity correction module and multispectral fusion analysis module. During the blackbody calibration stage, the control unit switches the optical path switching unit to the optical entrance to receive the second radiation signal, allowing the blackbody furnace radiation to access the main optical axis. Simultaneously, it drives the blackbody furnace temperature controller to gradually raise and lower the temperature according to the calibration temperature gauge, triggering data acquisition at each temperature point and writing it into the blackbody calibration module and temperature inversion module. During the online temperature measurement stage of the heating furnace, the control unit switches the optical path switching unit to the optical entrance to receive the first radiation signal and maintains the online temperature measurement state in the state machine, continuously triggering the data acquisition and real-time temperature analysis modules according to the set frame rate. Operation; To achieve the online self-calibration sub-stage, the control unit is equipped with a self-calibration cycle timer and a drift monitoring module. When the self-calibration cycle timer expires according to the preset self-calibration cycle, without changing the overall operating state of the temperature inversion module and the real-time temperature analysis module, it temporarily switches the state machine from online temperature measurement to online self-calibration sub-state, sends a switching command to the optical path switching unit, switches the main optical axis to receive signals at the optical entrance between the first radiation signal and the second or third radiation signal, and triggers the acquisition of calibration data frame sequences within a time period not exceeding the preset self-calibration duration threshold; after the self-calibration is completed, the control unit automatically switches the optical path back to the optical entrance. The system receives the first radiation signal and resumes online temperature measurement. The background system calculates the temperature deviation statistics within the preset drift monitoring time window and sends them together with the calibration data frame sequence obtained in the online self-calibration sub-stage to the temperature inversion module and the spatial uniformity correction module. The corresponding parameter update unit performs incremental updates of the uniformity correction coefficient and the weight of the inversion model. When the drift index is below the threshold, the current self-calibration cycle is maintained. When the drift index is close to or exceeds the threshold, the preset self-calibration cycle can be appropriately shortened or the number of frames acquired in a single self-calibration can be extended. Thus, the automatic setting and switching of the three working stages and the online self-calibration sub-stage are completed without interfering with the continuous production of the heating furnace.

[0021] Specifically, in the spatial uniformity and spectral response stage, the optical path switching unit introduces the third radiation signal emitted by the medium-sized uniform light source system into the main optical axis through optics, and the optical path switching unit introduces the fourth radiation signal emitted by the tunable laser source into the main optical axis through the optical entrance. The third and fourth radiation signals are then sent to the spatial uniformity correction module and the multispectral fusion analysis module, respectively, via a dynamic narrowband filter, an infrared point detector selected by the point detector selection module, a signal conditioning module, and an ADC conversion module. The spatial uniformity correction module analyzes the response changes of the optical platform and the selected infrared point detector at different incident angles and positions in the field of view, establishes a spatial uniformity correction coefficient field, and maps the coefficient field to the pixel coordinates of the target area of ​​the tested heating furnace. The multispectral fusion analysis module obtains the response curves at each wavelength in the target band and the relative sensitivity of each spectral channel by scanning the fourth radiation signal sent in by the tunable laser source point by point in the target band according to a preset wavelength sequence, and generates spectral response weights and band correction factors for multispectral fusion.

[0022] It should be specifically explained that during the blackbody calibration stage, the radiation outlet of the blackbody furnace transmits the second radiation signal to the optical path switching unit. The optical path switching unit couples the second radiation signal to the optical entrance and cuts off the reception of the first radiation signal. The second radiation signal is then sent to the blackbody calibration module and the temperature inversion module via a dynamic narrowband filter, a selected infrared point detector, a signal conditioning module, and an ADC conversion module. The blackbody calibration module is used to collect the output signals of the infrared point detector under the multispectral channels after intermediate frequency chopping when the blackbody furnace temperature is within a preset calibration range at multiple temperature points. It establishes a multi-channel voltage-temperature correspondence table and constructs an analytical relationship between radiation energy and temperature based on Planck's law and Stefan-Boltzmann's law. During the blackbody calibration stage, the temperature inversion module performs offline pre-training of the deep learning temperature inversion model based on the calibration relationship provided by the blackbody calibration module and the spectral response weights provided by the multispectral fusion analysis module, obtaining the initial temperature estimation function for different band combinations.

[0023] It should be specifically noted that during the online temperature measurement phase of the heating furnace, the temperature measurement window of the heating furnace under test transmits the first radiation signal to the optical entrance. The optical path switching unit keeps the optical entrance connected to the main optical axis while cutting off the access of the second radiation signal. The first radiation signal is sent to the temperature inversion module and the multispectral fusion analysis module through the dynamic narrowband filter, the selected infrared point detector, and the signal conditioning module. The real-time temperature analysis module outputs the real-time temperature field and the corresponding uncertainty field of the target area of ​​the heating furnace under test under the collaborative analysis of the spatial uniformity correction coefficient output by the spatial uniformity correction module, the multi-band radiation characteristics output by the multispectral fusion analysis module, and the multi-point calibration relationship provided by the blackbody calibration module. The real-time temperature analysis module returns the temperature uncertainty field to the filter control unit to update the channel configuration strategy for the next measurement cycle.

[0024] It should be noted that, as Figure 3 As shown, the filter control unit includes an uncertainty assessment subunit and a channel optimization decision subunit. The uncertainty assessment subunit is used to calculate the temperature estimation cost function of the current channel combination based on the temperature estimation result and its uncertainty index output by the temperature inversion module, the signal-to-noise ratio index of each narrowband channel, and the temperature change rate output by the real-time temperature analysis module. The channel optimization decision subunit adopts a pre-trained offline deep learning decision model or reinforcement learning decision model, taking the working stage command, channel signal-to-noise ratio index, temperature change rate, and current cost function as input, and outputs the target center wavelength set, passband width, channel scanning order, and dwell time parameters that meet the measurement time budget constraint, so as to minimize the expected value of the predicted temperature estimation uncertainty for the next measurement cycle, and transmits the parameters to the dynamic narrowband filter to achieve active adaptive band selection.

[0025] The filter control unit includes a filter control processor and a memory. The filter control processor is a CPU or embedded microprocessor, DSP (digital signal processor) chip or FPGA (field programmable gate array) with floating-point arithmetic unit on an industrial computer.

[0026] The uncertainty assessment subunit is implemented by software running on the filter control processor. It reads temperature estimates, temperature uncertainty indices, and temperature change rates from the temperature inversion module and real-time temperature analysis module via the system's internal bus, Ethernet, or high-speed serial interface. It also reads the signal-to-noise ratio (SNR) indices of each narrowband channel from the multispectral fusion analysis module or acquisition board. In RAM, it calculates the temperature estimation cost function for the current channel combination in real-time according to a preset formula. The channel optimization decision subunit also runs on the filter control processor or its coprocessor (e.g., embedded GPU, FPGA, or DSP core). A pre-trained deep learning or reinforcement learning decision model is stored as a weight file in non-volatile memory (Flash or SSD). During runtime, the control software loads this model into memory for inference calculations. It takes the operating phase instructions, channel SNR indices, temperature change rates, and the current cost function as input, and outputs the target center wavelength set, passband width, channel scanning order, and dwell time parameters.

[0027] It should be noted that, as Figure 4 As shown, the system also includes a path medium parameter inversion module. The signal conditioning module is connected to the multispectral fusion analysis module, spatial uniformity correction module, blackbody calibration module, and real-time temperature analysis module, respectively. The path medium parameter inversion module is connected to the tunable laser source, signal conditioning module, and ADC conversion module. It is used to control the tunable laser source to select several narrow wavelengths containing the absorption lines of combustion gas and water vapor on the field of view path of the tested heating furnace during the online temperature measurement stage or online self-calibration sub-stage, collect the corresponding laser transmission signals, and obtain the path average gas temperature, gas concentration, and path transmittance parameters based on the gas absorption spectrum model. The parameters are then input into the temperature inversion module. While outputting the surface temperature and equivalent emissivity of the tested heating furnace, the temperature inversion module substitutes the path medium parameters into the radiative transfer equation to separate and compensate for the self-radiation and absorption effects of the combustion gas and dust in the furnace. The system also includes an environmental parameter acquisition module and an online self-calibration management module. The environmental parameter acquisition module is used to collect ambient temperature, furnace pressure, combustion air distribution parameters, flue gas composition parameters, and optical window pollution level indicators around and inside the tested heating furnace. The online self-calibration management module is used to construct calibration drift indicators and spatial uniformity drift indicators based on the environmental parameters output by the environmental parameter acquisition module, the path transmittance change output by the path medium parameter inversion module, and the temperature deviation statistics output by the real-time temperature analysis module. When any indicator exceeds a preset threshold, the optical path switching unit is triggered to enter the online self-calibration sub-stage, and the calibration frames of the blackbody and uniform light source collected in the online self-calibration sub-stage are called to incrementally update the uniformity correction coefficient of the spatial uniformity correction module and the parameters of the deep learning temperature inversion model in the temperature inversion module.

[0028] The path medium parameter inversion module is preferably implemented by a processing unit, a storage unit, and an interface circuit and algorithm program that cooperate with the tunable laser source and signal acquisition link. Specifically, the path medium parameter inversion module uses an industrial computer, embedded microprocessor, or digital signal processor as the path medium inversion processor to perform the computation task. Its memory stores pre-set absorption spectrum parameter tables for combustion gases and water vapor, as well as corresponding gas absorption spectrum models. This processor is connected to the drive power supply and scanning controller of the tunable laser source via a serial bus or dedicated control interface. It sends control parameters such as scanning center wavelength, scanning step, scanning frequency, and modulation depth to the tunable laser source. It also receives the laser transmission signal sampling data converted by the photodetector in real time through a digital interface connected to the signal conditioning module and ADC conversion module. Internally, it sequentially performs baseline correction, normalization, modulation and demodulation or digital lock-in amplification, and spectral analysis. Signal processing algorithms, such as fitting, convert the transmittance intensity that varies with wavelength into an absorption rate curve. Based on the Lambert-Beer law and a preset absorption spectrum model, least squares fitting or nonlinear optimization methods are used to invert the path average gas temperature, gas component concentration, and path transmittance parameters. These parameters are then provided to the temperature inversion module in structured data form by the path medium parameter output unit. The path medium parameter inversion module can also have a built-in quality control and confidence assessment program. When the fitting residual or signal-to-noise ratio is lower than a preset threshold, the corresponding result is marked as low confidence and the online self-calibration management module is notified. Thus, through the cooperation of an existing industrial control computer or embedded control board, laser drive interface circuit, and digital signal processing algorithms, the automatic inversion and reliable output of path medium parameters are completed.

[0029] The environmental parameter acquisition module specifically uses armored thermocouples or platinum resistance temperature sensors, equipped with temperature transmitters, to convert temperature signals into standard 4-20mA current signals or industrial bus signals around and inside the furnace. Differential pressure transmitters or micro-pressure sensors are installed at the top of the furnace, in the flue, or at suitable locations to measure furnace and flue pressure, outputting standard current or digital values. Orifice flow meters, vortex flow meters, or thermal mass flow meters are installed in the primary and secondary air ducts, equipped with flow transmitters, to acquire air distribution parameters such as air volume and velocity. It also acquires operating parameters such as fan speed and damper opening through communication with the fan frequency converter or motor controller, serving as supplementary characteristics for combustion air distribution. An online flue gas analyzer is installed at a suitable location in the flue or furnace to measure parameters such as O2, CO, CO2, and NO. xAn extractive or in-situ analyzer with equal volume fraction outputs corresponding flue gas composition data. A low-power indicator light source and photodiode or photoelectric sensor are placed near the temperature measurement window to periodically measure the relative change in light intensity transmitted through the window, thereby constructing an optical window pollution level index. Alternatively, a small industrial camera is installed outside the window to image the window area, and image processing algorithms assess the proportion of attached material area, grayscale attenuation, etc., converting them into a quantified pollution index. The online self-calibration management module runs as a software function module on an industrial computer, embedded controller, or PLC. It triggers and manages the online self-calibration process by performing statistical analysis and decision-making on multi-source data, specifically including: The system periodically reads environmental data such as ambient temperature, furnace pressure, air distribution parameters, flue gas composition, and window pollution index uploaded by the environmental parameter acquisition module via Ethernet or fieldbus. Simultaneously, it receives the path transmittance, path average gas temperature and concentration changes output by the path medium parameter inversion module, as well as the temperature deviation statistics (such as the mean, variance or moving average error of the deviation from the blackbody calibration curve) output by the real-time temperature analysis module. Establish a timestamp cache queue in the controller memory or industrial computer memory to form a data sequence within a preset drift monitoring time window; The software pre-defines the calculation formulas for calibration drift index and spatial uniformity drift index, such as: the sliding mean, sliding standard deviation, relative change rate of path transmittance, and change rate of window contamination index based on temperature deviation, to construct a comprehensive drift metric. The cached data is processed in each monitoring cycle to obtain the current calibration drift index and spatial uniformity drift index, and then compared with the preset threshold. When any drift index exceeds the corresponding threshold, the online self-calibration management module sends an online self-calibration sub-stage trigger command to the system control unit and the optical path switching unit, instructing the optical path switching unit to temporarily switch the main optical axis to the blackbody furnace or the medium-sized uniform light source system. At the same time, the acquisition duration and number of calibration frames for this self-calibration are issued to coordinate the data acquisition unit to acquire blackbody calibration frames and uniform light source calibration frames within the preset self-calibration duration threshold. After self-calibration is completed, the online self-calibration management module packages the newly acquired calibration frame with the historical deviation data in the drift monitoring window and sends it to the spatial uniformity correction module and the temperature inversion module according to the predetermined interface protocol. The module calls the incremental update interface for uniformity correction coefficients and the incremental update interface for parameters of the deep learning temperature inversion model to perform online parameter correction. Based on changes in the updated drift metrics, the subsequent self-calibration cycle is automatically adjusted (e.g., appropriately shortened or extended) to optimize system maintenance frequency.

[0030] It should be specifically noted that the deep learning temperature inversion model in the temperature inversion module is obtained through offline training, which includes: During the blackbody calibration stage, a blackbody furnace is used as the test object. Multi-band radiation characteristics and corresponding reference temperatures are collected at multiple temperature set points and processed by the spatial uniformity correction module and the multispectral fusion analysis module to form a blackbody training sample set. During the material calibration stage, a variety of test materials and surface states are selected and placed in a temperature environment equivalent to a blackbody furnace. Multi-band radiation characteristics and their corresponding reference temperatures are collected, and spectral emission priors from the material emissivity database are introduced to form an extended training sample set. A composite loss function is constructed that simultaneously constrains the temperature inversion error, emissivity estimation error, and residuals of Planck's radiation law and radiative transfer equation. The network parameters of the deep learning temperature inversion model are iteratively optimized so that the model can output temperature estimates, equivalent emissivity estimates, and corresponding temperature estimation uncertainty indices that satisfy physical constraints under multiple material, surface, and operating conditions.

[0031] It should be noted that the offline training of the deep learning temperature inversion model in the temperature inversion module is preferably completed on an independent training computing platform. This platform includes a training server or industrial computer configured with a central processing unit (CPU) and a graphics processing unit (GPU) or a field-programmable gate array (FPGA), as well as a database and large-capacity storage units for storing the training sample set. The multi-band radiation characteristics, corresponding reference temperatures, and material emission priors collected during the blackbody calibration and material calibration stages, after processing by the spatial homogeneity correction module and the multispectral fusion analysis module, are uploaded to the training server via a local area network or industrial Ethernet. Within the training server, the deep learning training software module performs data preprocessing, sample partitioning, and batch loading. Based on the constructed network structure and composite loss function, the deep learning training software module uses stochastic gradient descent, Adam optimization algorithm, or other iterative optimization algorithms to iteratively update the network parameters multiple times. It also monitors the temperature inversion error, emissivity estimation error, and physical constraint residuals on the validation set. When the preset convergence condition or early stopping condition is met, the obtained optimal network weight parameters and normalized coefficients are stored in a non-volatile storage medium as a model file. After training, the model file is downloaded to the temperature inversion module of the temperature measurement system via wired network, mobile storage medium or remote update. The inference program in the temperature inversion module loads the file and uses it for online temperature estimation and temperature uncertainty output, thereby separating the offline training and online deployment of the deep learning temperature inversion model and ensuring the real-time performance and stability of the on-site temperature measurement system.

[0032] The above provides a detailed description of an infrared radiation temperature measurement system based on a point detector provided by the present invention. Specific examples have been used to illustrate the principle and implementation of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation and application scope based on the idea of ​​the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. An infrared radiation temperature measurement system based on a point detector, characterized in that, The system includes an optical platform with an optical path switching unit connected to an optical entrance. These entrances converge onto a main optical axis. A dynamic narrowband filter and a point detector selection module are sequentially arranged on the main optical axis. The point detector selection module is connected to a signal conditioning module, which in turn is connected to a multispectral fusion analysis module, a spatial uniformity correction module, a blackbody calibration module, and a real-time temperature analysis module. The blackbody calibration module and the spatial uniformity correction module are connected to a temperature inversion module, and both the temperature inversion module and the multispectral fusion analysis module are connected to the real-time temperature analysis module. The dynamic narrowband filter is connected to a filter... The control unit and filter control unit switch the channels of the dynamic narrowband filter within a preset band to obtain the signal-to-noise ratio and receive the output of the temperature inversion module. Based on the measurement time constraint and the temperature uncertainty cost function, the control unit selects the target center wavelength set, passband width, channel scanning order and dwell time of the measurement period to minimize the uncertainty index. The temperature inversion module takes the spatially corrected multi-band radiation characteristics, multi-point calibration data and multi-band radiation intensity, ratio and difference characteristics as input, and outputs the temperature estimate, equivalent emissivity estimate and temperature estimation uncertainty index based on the deep learning temperature inversion model.

2. The infrared radiation temperature measurement system based on a point detector according to claim 1, characterized in that, The optical entrance is used to: connect to the temperature measuring window of the furnace under test to obtain a first radiation signal; connect to the radiation outlet of the blackbody furnace to obtain a second radiation signal; and connect to a medium-sized uniform light source system to obtain a third radiation signal. It is connected to a tunable laser source to obtain a fourth radiation signal.

3. The infrared radiation temperature measurement system based on a point detector according to claim 2, characterized in that, The system's operation includes a spatial uniformity and spectral response stage, a blackbody calibration stage, and an online temperature measurement stage for the heating furnace. During the online temperature measurement stage, an online self-calibration sub-stage is inserted according to a preset self-calibration cycle. This sub-stage uses an optical path switching unit to switch the main optical axis from the heating furnace under test to a blackbody furnace or a medium-sized uniform light source system without altering the overall system operation. Within a time period not exceeding a preset self-calibration duration threshold, a calibration data frame sequence is acquired. This sequence is then combined with the accumulated temperature deviation statistics within a preset drift monitoring time window to correct parameter drift in the temperature inversion module and the uniformity correction coefficients in the spatial uniformity correction module during the drift monitoring time window.

4. The infrared radiation temperature measurement system based on a point detector according to claim 3, characterized in that, In the spatial uniformity and spectral response stage, the optical path switching unit introduces the third radiation signal emitted by the medium-sized uniform light source system into the main optical axis through the optical entrance. The optical path switching unit also introduces the fourth radiation signal emitted by the tunable laser source into the main optical axis through the optical entrance. The third and fourth radiation signals are then sent to the spatial uniformity correction module and the multispectral fusion analysis module, respectively, via a dynamic narrowband filter, an infrared point detector selected by the point detector selection module, a signal conditioning module, and an ADC conversion module. The spatial uniformity correction module analyzes the response changes of the optical platform and the selected infrared point detector at different incident angles and positions in the field of view, establishes a spatial uniformity correction coefficient field, and maps the coefficient field to the pixel coordinates of the target area of ​​the tested heating furnace. The multispectral fusion analysis module obtains the response curves at each wavelength in the target band and the relative sensitivity of each spectral channel by scanning the fourth radiation signal sent in by the tunable laser source point by point in the target band according to a preset wavelength sequence, and generates spectral response weights and band correction factors for multispectral fusion.

5. The infrared radiation temperature measurement system based on a point detector according to claim 4, characterized in that, During the blackbody calibration stage, the radiation outlet of the blackbody furnace transmits the second radiation signal to the optical path switching unit. The optical path switching unit couples the second radiation signal to the optical entrance and cuts off the first radiation signal. The second radiation signal is then sent to the blackbody calibration module and the temperature inversion module via a dynamic narrowband filter, a selected infrared point detector, a signal conditioning module, and an ADC conversion module. The blackbody calibration module is used to collect the output signal of the infrared point detector under the multispectral channel after intermediate frequency chopping when the blackbody furnace temperature is within a preset calibration range at multiple temperature points. It establishes a multi-channel voltage-temperature correspondence table and constructs an analytical relationship between radiation energy and temperature based on Planck's law and Stefan-Boltzmann's law. During the blackbody calibration stage, the temperature inversion module performs offline pre-training of the deep learning temperature inversion model based on the calibration relationship provided by the blackbody calibration module and the spectral response weights provided by the multispectral fusion analysis module, obtaining the initial temperature estimation function for different band combinations.

6. The infrared radiation temperature measurement system based on a point detector according to claim 5, characterized in that, During the online temperature measurement phase of the heating furnace, the temperature measurement window of the heating furnace under test transmits the first radiation signal to the optical entrance. The optical path switching unit keeps the optical entrance connected to the main optical axis to receive the first radiation signal and cuts off the second radiation signal. The first radiation signal is sent to the temperature inversion module and the multispectral fusion analysis module through the dynamic narrowband filter, the selected infrared point detector, and the signal conditioning module. The real-time temperature analysis module outputs the real-time temperature field and the corresponding uncertainty field of the target area of ​​the heating furnace under test under the collaborative analysis of the spatial uniformity correction coefficient output by the spatial uniformity correction module, the multi-band radiation characteristics output by the multispectral fusion analysis module, and the multi-point calibration relationship provided by the blackbody calibration module. The real-time temperature analysis module returns the temperature uncertainty field to the filter control unit to update the channel configuration strategy for the next measurement cycle.

7. The infrared radiation temperature measurement system based on a point detector according to claim 6, characterized in that, The filter control unit includes an uncertainty assessment subunit and a channel optimization decision subunit. The uncertainty assessment subunit calculates the temperature estimation cost function for the current channel combination based on the temperature estimation results and uncertainty index output by the temperature inversion module, the signal-to-noise ratio index of each narrowband channel, and the temperature change rate output by the real-time temperature analysis module. The channel optimization decision subunit uses a pre-trained offline deep learning decision model or reinforcement learning decision model, taking the working phase command, channel signal-to-noise ratio index, temperature change rate, and current cost function as input, and outputs the target center wavelength set, passband width, channel scanning order, and dwell time parameters that meet the measurement time budget constraint, so as to minimize the expected value of the predicted temperature estimation uncertainty for the next measurement cycle, and transmits the parameters to the dynamic narrowband filter to achieve active adaptive band selection.

8. The infrared radiation temperature measurement system based on a point detector according to claim 7, characterized in that, The system also includes a path medium parameter inversion module. The signal conditioning module is connected to the multispectral fusion analysis module, spatial uniformity correction module, blackbody calibration module, and real-time temperature analysis module, respectively. The path medium parameter inversion module is connected to the tunable laser source, the signal conditioning module, and the ADC conversion module. It is used to control the tunable laser source to select several narrow wavelengths containing the absorption lines of combustion gas and water vapor on the field of view path of the heated furnace during the online temperature measurement stage or the online self-calibration sub-stage, collect the corresponding laser transmission signal, invert the path average gas temperature, gas concentration, and path transmittance parameters based on the gas absorption spectrum model, and input the parameters into the temperature inversion module. While outputting the surface temperature and equivalent emissivity of the furnace under test, the temperature inversion module substitutes the path medium parameters into the radiative transfer equation to separate and compensate for the self-radiation and absorption effects of the combustion gases and dust in the furnace.

9. The infrared radiation temperature measurement system based on a point detector according to claim 8, characterized in that, The system also includes an environmental parameter acquisition module and an online self-calibration management module. The environmental parameter acquisition module is used to collect ambient temperature, furnace pressure, combustion air distribution parameters, flue gas composition parameters, and optical window pollution level indicators around and inside the tested heating furnace. The online self-calibration management module is used to construct calibration drift indicators and spatial uniformity drift indicators based on the environmental parameters output by the environmental parameter acquisition module, the path transmittance change output by the path medium parameter inversion module, and the temperature deviation statistics output by the real-time temperature analysis module. When any indicator exceeds a preset threshold, the optical path switching unit is triggered to enter the online self-calibration sub-stage, and the calibration frames of the blackbody and uniform light source collected in the online self-calibration sub-stage are called to incrementally update the uniformity correction coefficient of the spatial uniformity correction module and the parameters of the deep learning temperature inversion model in the temperature inversion module.

10. The infrared radiation temperature measurement system based on a point detector according to claim 2, characterized in that, The deep learning temperature inversion model in the temperature inversion module is obtained through offline training, which includes: During the blackbody calibration stage, a blackbody furnace is used as the test object. Multi-band radiation characteristics and corresponding reference temperatures are collected at multiple temperature set points and processed by the spatial uniformity correction module and the multispectral fusion analysis module to form a blackbody training sample set. During the material calibration stage, a variety of test materials and surface states are selected and placed in a temperature environment equivalent to a blackbody furnace. Multi-band radiation characteristics and their corresponding reference temperatures are collected, and spectral emission priors from the material emissivity database are introduced to form an extended training sample set. A composite loss function is constructed that simultaneously constrains the temperature inversion error, emissivity estimation error, and residuals of Planck's radiation law and radiative transfer equation. The network parameters of the deep learning temperature inversion model are iteratively optimized so that the model can output temperature estimates, equivalent emissivity estimates, and corresponding temperature estimation uncertainty indices that satisfy physical constraints under multiple material, surface, and operating conditions.

Images