A real-time monitoring method for temperature field of a forging process based on infrared images

Abstract

The present application belongs to the technical field of image processing, and particularly relates to a forging process temperature field real-time monitoring method based on infrared images, which comprises the following steps: firstly, a temperature prediction model is established based on Newton's cooling law; secondly, a spectral selectivity distortion factor is constructed to quantify environmental interference by comparing the measured and expected radiation intensity ratios of multi-band infrared images; at the same time, a surface emissivity mutation index is constructed to identify surface changes by calculating the space-time gradient of radiation intensity; finally, the reliability score of colorimetric temperature is calculated by comprehensively considering the two indexes, and the predicted temperature and the colorimetric temperature are dynamically weighted and fused according to the score, so as to output an accurate, continuous and strong anti-interference optimal temperature estimation value. The present application can realize accurate temperature measurement by constructing environmental interference and surface state change indexes, evaluating the reliability of colorimetric temperature, and dynamically weighting and fusing the colorimetric temperature and the predicted temperature based on a physical model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing. More particularly, the present application relates to a real-time monitoring method for temperature field of forging process based on infrared image. BACKGROUND

[0002] In modernized forging production line, accurate and real-time monitoring of temperature of forgings during heating, transferring and forging forming process is a key link for controlling metallographic structure, mechanical properties of forgings and optimizing forging process. Infrared thermal imaging technology has become a mainstream monitoring method in this field due to its non-contact, fast response and providing full-field temperature distribution.

[0003] In order to achieve the above monitoring target, colorimetric temperature measurement algorithm is a core technical means which is widely researched and applied. The algorithm calculates the temperature by collecting radiation intensity information of forgings at least at two different infrared bands and calculating the ratio. In theory, if the attenuation effects of the environment medium on the two bands are the same, and the emissivity of the forgings at the two bands is also the same or the ratio is constant, then the error caused by the environment attenuation and the uncertainty of the emissivity can be eliminated to a great extent by the ratio calculation.

[0004] However, in the actual forging workshop environment, the above ideal premise is often difficult to meet, thereby constituting the limitation of the prior art: the existing colorimetric temperature measurement algorithm is isolated, which only calculates the instantaneous independent temperature according to the ratio of radiation intensity at two bands at the moment, and cannot effectively distinguish the physical root of the change of the observed radiation intensity ratio. Whether the change is the real temperature fluctuation of the forgings, or the transmission attenuation ratio of the infrared signal is changed due to the complex and changeable environment medium in the forging field, such as sudden passing of water vapor group, increased oil smoke concentration, or the emissivity ratio of the forgings is changed due to the generation or peeling of the oxidation skin on the surface of the forgings; the three kinds of physical events with completely different properties may show similar temperature reading jumps at the output end of the traditional colorimetric algorithm, which leads to the monitoring system being unable to accurately attribute, thereby causing serious misjudgment of the real thermal working condition of the forgings and affecting the subsequent process control decision. SUMMARY

[0005] To solve the technical problem that the existing colorimetric temperature measurement algorithm cannot accurately attribute the source of radiation signal change, leading to inaccurate temperature measurement results, the present application provides a real-time monitoring method for temperature field in forging process based on infrared image, comprising: continuously collecting radiation intensity image sequences of forgings under multiple different infrared wavebands; based on the optimal temperature estimation value at the previous moment and Newton's cooling law, calculating the predicted temperature value of each pixel point of the forgings at the current moment; selecting a waveband insensitive to environmental absorption as a reference waveband, comparing the measured radiation intensity ratio of other wavebands with the reference waveband with the expected radiation intensity ratio calculated based on the predicted temperature value, to obtain a spectral selectivity distortion factor for representing the degree of non-gray attenuation of the environmental medium; calculating the normalized spatial gradient and normalized time gradient of the radiation intensity value at the reference wavelength, and taking their sum as a surface emissivity mutation index for representing the surface emissivity mutation of the forgings; obtaining the colorimetric temperature value of each pixel point of the forgings at the current moment through the colorimetric temperature measurement algorithm; inputting the spectral selectivity distortion factor and surface state change index into a preset exponential decay function to obtain a credibility score of the colorimetric temperature value; and according to the credibility score, dynamically weighting and fusing the predicted temperature value and the colorimetric temperature value to obtain the optimal temperature estimation value at the current moment.

[0006] The present application establishes a multi-dimensional comprehensive analysis framework, introduces a physical prediction model based on Newton's cooling law as a continuous reference for temperature change; at the same time, constructs a spectral selectivity distortion factor and a surface emissivity mutation index, realizes the quantitative identification and separation of the two main interference sources, i.e. environmental medium interference and surface state mutation of forgings; finally, through a dynamic weighting fusion strategy based on credibility score, the stable physical prediction value and the real-time colorimetric measurement value are intelligently combined, the intelligent attribution of radiation signal change is realized, the ambiguity problem of traditional methods is fundamentally solved, and more accurate, robust and continuous temperature field data can be output in complex and variable forging site environment.

[0007] Preferably, the multiple different infrared wavebands include 1.6 µm, 2.3 µm, 3.9 µm and 4.5 µm, wherein the reference wavelength is 3.9 µm.

[0008] Preferably, the calculation of the predicted temperature value of each pixel point of the forgings at the current moment based on the optimal temperature estimation value at the previous moment and Newton's cooling law comprises: ; wherein, is the predicted temperature value of the pixel point at the current moment ; is the optimal temperature estimation value of the pixel point at the previous moment ; The time interval for image acquisition; This represents the overall cooling coefficient at the pixel location in the previous time step, and , For pixels At any moment The optimal temperature estimate; For a moment Ambient temperature, For the previous moment Ambient temperature.

[0009] Preferably, obtaining the spectral selectivity distortion factor for characterizing the degree of non-gray attenuation of the environmental medium includes: In the formula, For pixels At any moment Spectral selectivity distortion factor; This represents the total number of bands; This indicates that the summation is performed over all non-reference bands. for At all times in wavelength Pixels in the image below The radiation intensity value at that location; for At all times at the reference wavelength Pixels in the image below The radiation intensity value at that location; for At all times in wavelength The expected radiation intensity; for At all times at the reference wavelength The expected radiation intensity; This indicates taking the absolute value.

[0010] This invention compares the measured radiation intensity ratio with the expected radiation intensity ratio calculated based on the physical predicted temperature, thereby canceling out the effects caused by changes in actual temperature and emissivity. This accurately separates and quantifies the spectral distortion caused by environmental media such as water vapor and oil fumes, enabling the system to accurately identify the intensity of external environmental interference and providing a direct basis for subsequent adaptive fusion.

[0011] Preferably, the expected radiation intensity is calculated based on the predicted temperature using Planck's law under ideal conditions with no environmental attenuation, for each wavelength. The corresponding expected radiation intensity.

[0012] Preferably, the formula for calculating the surface emissivity abrupt change index is: In the formula, For pixels At any moment Surface emissivity abrupt change index; , for Time and At all times at the reference wavelength Pixels in the image below The radiation intensity value at that location; Based on pixels The central window All pixels within the reference wavelength The average radiation intensity below; This represents the normalized spatial gradient of the radiation intensity value at the reference wavelength. This represents the normalized time gradient of the radiation intensity value at the reference wavelength.

[0013] This invention, by simultaneously calculating the normalized spatial gradient and normalized temporal gradient of the radiation intensity at the reference wavelength, can effectively capture local and instantaneous signal anomalies caused by abrupt changes in emissivity, i.e., inconsistencies in space with the surroundings and inconsistencies in time with the previous moment. This enables the system to accurately identify changes in the surface state of the forging itself and achieves effective monitoring of internal interference sources.

[0014] Preferably, the step of inputting the spectral selectivity distortion factor and surface state change index into a preset exponential decay function to obtain the confidence score of the colorimetric temperature value includes: In the formula, For pixels At the present moment The confidence score of the colorimetric temperature; For pixels At any moment Spectral selectivity distortion factor; For pixels At any moment Surface emissivity abrupt change index; and It is a dimensionless sensitivity coefficient used to control the response speed of the weights to each indicator; This represents the natural exponential function.

[0015] This invention transforms the two types of interference indicators identified—distortion factor and mutation index—into a comprehensive confidence score through an exponential decay function. Then, it dynamically adjusts the weights of the predicted temperature and colorimetric temperature through linear mapping, which greatly enhances the stability and robustness of the algorithm under extreme conditions while ensuring measurement accuracy.

[0016] Preferably, the step of dynamically weighting and fusing the predicted temperature value and the colorimetric temperature value based on the confidence score to obtain the optimal temperature estimate for the current moment includes: ; wherein, is the pixel point at the time of the optimal temperature estimation value; is the pixel point at the time of the predicted temperature value; is the pixel point at the time of the colorimetric temperature value; , are the weight coefficients of the predicted temperature value and the colorimetric temperature value respectively, and .

[0017] According to the confidence score, the present application dynamically weights and fuses the predicted temperature value and the colorimetric temperature value, so that the final temperature estimation value can smoothly switch between physical prediction and optical measurement, while ensuring measurement accuracy, greatly enhancing the stability and robustness of the algorithm under extreme working conditions.

[0018] Preferably, the calculation formula of the weight coefficients of the predicted temperature value and the colorimetric temperature value is: ; ; wherein, is the maximum weight that the colorimetric temperature value can obtain; is the minimum weight that the colorimetric temperature value can obtain; is the pixel point at the current time of the confidence score of the colorimetric temperature.

[0019] Preferably, when the colorimetric temperature value of each pixel point of the forging at the current time is obtained by the colorimetric temperature measurement algorithm, the wavelength = 1.6 µm and the wavelength = 2.3 µm are selected as the wavelength combination for colorimetric temperature measurement.

[0020] The present application has the following beneficial effects:

[0021] The present application establishes a multi-dimensional comprehensive analysis framework, introduces a physical prediction model based on Newton's cooling law as a continuous reference for temperature change; at the same time, constructs a spectral selective distortion factor and a surface emissivity mutation index, realizes the quantitative identification and separation of the two main interference sources, environmental medium interference and forging surface state mutation; finally, through a dynamic weighting fusion strategy based on the confidence score, the stable physical prediction value and the real-time colorimetric measurement value are intelligently combined, the intelligent attribution of the radiation signal change is realized, the ambiguity problem of the traditional method is fundamentally solved, and more accurate, robust and continuous temperature field data can be output in the complex and variable forging site environment. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating a method for real-time monitoring of the temperature field in a forging process based on infrared images, according to the present invention.

[0023] Figure 2 This is a schematic diagram illustrating the results of feature analysis;

[0024] Figure 3 This is a schematic diagram illustrating the temperature monitoring results. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0027] This invention discloses a method for real-time monitoring of the temperature field during forging based on infrared images, referring to... Figure 1 This includes steps S1 to S5:

[0028] S1: Continuously acquire a sequence of radiation intensity images of the forging in multiple different infrared bands.

[0029] It should be noted that in order to dynamically analyze the temperature change process of the forging and distinguish the influence of different physical events on the temperature measurement results, basic data including time, space and spectrum were first acquired. Therefore, image sequences of the forging were continuously acquired in multiple different infrared bands.

[0030] Specifically, a multi-band infrared thermal imager is used during the forging process at a preset frame rate. Continuous collection of forgings at Different center wavelengths The following is a sequence of radiation intensity images, in which, =30Hz (Hertz).

[0031] In order to achieve multi-band analysis, especially to quantify the spectral distortion caused by water vapor and oil fumes, the following settings are configured: =4, four different center wavelengths including 1.6µm, 2.3µm, 3.9µm and 4.5µm. Among them, 1.6µm is short-wave infrared. For high-temperature targets such as forgings, the radiation signal of short wavelengths is very strong, with a high signal-to-noise ratio. Moreover, short wavelengths are relatively insensitive to changes in the emissivity of the object surface. As one of the main temperature measurement channels, it has a good ability to capture high-temperature details and can provide relatively stable temperature readings. 2.3µm is short-wave infrared, located in the strong radiation region of high-temperature signals and close to a relatively weak water vapor absorption band. Combined with 1.6µm, it can form an effective colorimetric measurement. Temperature comparison, and by comparing it with longer wavelengths, the influence of water vapor can be initially perceived; 3.9µm is the mid-wave infrared, which is the optimal reference wavelength. It is located in the clean region of the mid-wave infrared atmospheric window and can most effectively avoid the strong absorption of water vapor and carbon dioxide; 4.5µm is the mid-wave infrared, which is at the edge of the strong carbon dioxide absorption band (4.2-4.4µm). It is a detection channel. When the concentration of oil fumes in the forging workshop increases, leading to an increase in carbon dioxide content, the signal of this channel will be significantly and characteristically attenuated, thus capturing and quantifying the smoke interference.

[0032] Furthermore, obtain At all times in wavelength Pixels in the image below The radiation intensity value at that location is recorded as follows: .

[0033] S2: Based on the optimal temperature estimate from the previous moment and Newton's law of cooling, calculate the predicted temperature value of each pixel of the forging at the current moment.

[0034] It should be noted that in order to distinguish the actual temperature fluctuations of the forging from the measurement jumps caused by changes in the environment or surface condition, it is necessary to establish a temperature change benchmark that conforms to physical laws. When there is no external sudden event interference, the temperature change of the forging mainly follows the natural cooling law. Therefore, constructing a prediction model based on thermodynamic laws can provide an expected evolution trajectory of the temperature of each pixel point as a basis for subsequent judgment of whether the measurement value is abnormal. Based on the physical laws of the temperature change of the forging during the forging process, a local thermodynamic trend prediction model is established for each pixel point. During the forging interval, the forging mainly dissipates heat to the surrounding environment through thermal radiation and thermal convection, and its temperature change can be approximately described by Newton's law of cooling.

[0035] Specifically, in order to build this model, it is first necessary to obtain the pixel data at the previous time step. Optimal temperature estimate It should be noted that, since this invention obtains the optimal temperature estimate in real time at each moment using the method of this invention, therefore, at the current moment, i.e., time... It has already obtained its previous moment, i.e., the moment. The optimal temperature estimate.

[0036] Furthermore, based on the optimal temperature estimate from the previous moment, the predicted moment is... The physical trend of temperature is used to determine the predicted temperature value at the current moment. The specific calculation formula is as follows:

[0037]

[0038] In the formula, For pixels At any moment The predicted temperature value; For pixels In the previous moment The optimal temperature estimate; The time interval for image acquisition is denoted as , and , Preset frame rate; This represents the overall cooling coefficient at the pixel location at the previous time step. This coefficient characterizes the rate of local heat dissipation, and , For pixels At any moment The optimal temperature estimate; For a moment Ambient temperature, For the previous moment The ambient temperature can be obtained by measuring it using an independent sensor.

[0039] The calculation formula provides a physically-compliant expected trajectory for the temperature change of each pixel. Under normal circumstances, without external events such as forging or strong environmental interference, the actual temperature should roughly follow this trajectory. The temperature at the previous moment... The higher the temperature, the greater the temperature difference, and the greater the predicted temperature drop, which is consistent with the basic principle of natural cooling.

[0040] It should be noted that this model provides a smooth and continuous physical benchmark for temperature monitoring. Even in the event of brief optical signal distortion, the system can still provide a reasonable temperature estimate based on physical laws, which greatly improves the continuity and reliability of the temperature data stream and provides a key criterion for distinguishing between real temperature changes and measurement artifacts.

[0041] S3: By comparing the ratio of measured radiation intensity of other bands to that of the reference band with the ratio of expected radiation intensity calculated based on the predicted temperature value, a spectral selective distortion factor is obtained to characterize the degree of non-grey attenuation of the environmental medium.

[0042] It should be noted that the environmental media such as water vapor and oil fume in the forging site has selective absorption or scattering characteristics for infrared radiation of different wavelengths, that is, non-gray attenuation, which will directly distort the proportional relationship of multi-band radiation intensity, leading to the failure of colorimetric temperature measurement, therefore, an index is needed to evaluate the deviation degree of the current radiation signal spectrum from the normal spectrum, which is only affected by the temperature and emissivity of the forging, so as to identify and quantify the intensity of this spectral selective interference.

[0043] Specifically, based on the predicted temperature , the expected radiation intensity of each wavelength in the ideal case without environmental attenuation is calculated by using Planck's law, which is a known technology and will not be described here.

[0044] It should be noted that based on the comprehensive consideration of the actual environmental interference in the forging workshop and the infrared physical characteristics, the main interference comes from the sudden passing of water vapor group and the increased concentration of oil fume, and the products of the combustion process such as oil fume must contain a large amount of carbon dioxide.

[0045] Further, a reference wavelength which is least sensitive to environmental absorption is selected In the field of infrared thermal imaging, there is an important mid-wave infrared atmospheric window (MWIR) about 3-5µm; this window is very ideal for measuring high-temperature targets such as forgings, with high signal strength; however, this window is not completely clean, and the 4.2µm to 4.4µm band is a strong absorption band of carbon dioxide, while the 2.6µm to 2.8µm and 5µm above bands are strong absorption bands of water vapor; therefore, 3.9µm is finally selected as the reference wavelength, which is located in a relatively clean position within this mid-wave window, avoiding the main absorption peaks of water vapor and carbon dioxide, so when smoke or water vapor passes through the path, the attenuation of the infrared signal of this wavelength is much smaller than that of other wavelengths.

[0046] Specifically, the ratio of the measured radiation intensity of each band to the measured radiation intensity of the reference band is calculated, and compared with the expected radiation intensity ratio, and the spectral selective distortion factor at the current time is calculated according to the difference between the measured radiation intensity ratio and the expected radiation intensity ratio, and the specific calculation formula is:

[0047]

[0048] In the formula, is the spectral selective distortion factor of the pixel point at time . is the total number of wavebands; represents the sum of all non-reference wavebands; is the reference waveband; is the radiation intensity value at the pixel point in the image at the wavelength at the time t; is the reference waveband; is the radiation intensity value at the pixel point in the image at the reference wavelength at the time t; is the reference waveband; is the expected radiation intensity at the wavelength at the time t; is the reference waveband; is the expected radiation intensity at the reference wavelength at the time t; represents the absolute value.

[0049] wherein, is the measured radiation intensity ratio; is the expected radiation intensity ratio calculated based on the physical prediction temperature; if there is a medium with spectral selective absorption in the environment, such as water vapor, it will cause different degrees of attenuation to different wavelengths, causing the measured radiation intensity ratio sequence to deviate from the ratio sequence predicted based on the single-temperature physical model, and the spectral selective distortion factor quantifies the non-gray degree of environmental interference at the time t by accumulating the absolute value of this deviation; when the value of the spectral selective distortion factor is small, it indicates that the attenuation of the environmental medium is close to gray, or the interference is weak; when its value significantly increases, it strongly indicates that there is a spectral selective environmental interference at present.

[0050] It should be noted that this factor can identify and quantify the non-gray attenuation interference caused by water vapor, oil fume, etc. in real time, accurately capture external environmental interference events, provide a direct basis for subsequent adaptive adjustment of temperature measurement strategy, and avoid the dramatic fluctuation and systematic low of temperature measurement values caused by environmental mutations.

[0051] S4: Calculate the normalized spatial gradient and normalized time gradient of the radiation intensity value at the reference wavelength, and take the sum of the two as the surface emissivity mutation index for representing the surface emissivity mutation of the forging surface.

[0052] It should be noted that the generation or peeling of the oxide skin on the surface of the forging will cause a mutation of its own emissivity, which will also cause a dramatic change in radiation intensity, thereby affecting the accuracy of temperature measurement. Unlike the diffuse environmental interference, such events are usually locally sudden, therefore, an index capable of capturing the abnormal changes of the signal in the neighborhood is needed to identify such surface state change events, and finally the spatial inconsistency and temporal mutation of the radiation intensity signal in a small neighborhood are analyzed to determine whether there is a surface state change.

[0053] Specifically, the pixel point at the reference wavelength at the reference wavelength

[0054]

[0055] In the formula, is the pixel point at the reference wavelength at the reference wavelength at the reference wavelength at the reference wavelength at the reference wavelength at the reference wavelength is the average radiation intensity of all pixel points in the window centered on the pixel point at the reference wavelength , and the size of the window is 7x7; at the reference wavelength at the reference wavelength at the reference wavelength at the reference wavelength represents taking the absolute value.

[0056] wherein the first term of the calculation formula represents the normalized spatial gradient, measuring the out-of-group degree of the point and the surrounding area; the second term represents the normalized time gradient, measuring the mutation degree of the radiation intensity of the point; the peeling or new generation of the oxide skin is usually a locally sudden event, which will cause the emissivity of the point to change instantaneously, thereby causing the radiation intensity of the point to be obviously different from the surrounding area which has not changed, i.e. high spatial gradient, and in time, it also has a huge jump from the state of the previous moment, i.e. high time gradient, the surface emissivity mutation index combines the two features to measure; when the surface emissivity mutation index A low value indicates a stable surface condition; a sudden spike in emissivity strongly suggests an abrupt change in emissivity.

[0057] It should be noted that the surface emissivity mutation index enables the effective identification of the internal physical event of a sudden change in the surface emissivity of forgings. This allows the system to clearly distinguish whether a temperature reading jump is caused by external environmental interference or by changes in the internal surface state, laying the foundation for attributing measurement anomalies.

[0058] S5: Input the spectral selectivity distortion factor and surface state change index into the preset exponential decay function to obtain the confidence score of the colorimetric temperature value; based on the confidence score, dynamically weight and fuse the predicted temperature value and the colorimetric temperature value to obtain the optimal temperature estimate at the current moment.

[0059] It should be noted that traditional colorimetric temperature measurement, temperature prediction based on physical models, and event judgment based on feature indicators each provide information about the temperature of forgings in different dimensions. In order to obtain the temperature value that is closest to the actual situation, information from multiple different dimensions is integrated. Based on the reliability of the working conditions diagnosed in real time, the weight of each information source in the final temperature decision is dynamically adjusted to achieve adaptive optimal temperature estimation.

[0060] Specifically, select wavelength =1.6µm and wavelength =2.3µm was used as the wavelength combination for colorimetric temperature measurement. The colorimetric temperature value was obtained through a colorimetric temperature measurement algorithm. Among them, the colorimetric temperature measurement algorithm is a well-known technology and will not be elaborated here.

[0061] Furthermore, specifically, the time is calculated based on the predicted temperature value and the colorimetric temperature value. The optimal temperature estimate is calculated using the following formula:

[0062]

[0063] In the formula, For pixels At any moment The optimal temperature estimate; For pixels At any moment The predicted temperature value; For pixels At any moment The colorimetric temperature value; , These are the weighting coefficients for the predicted temperature value and the colorimetric temperature value, respectively. .

[0064] It should be noted that the weight coefficient and are dynamically adjusted according to the spectral selectivity distortion factor and the surface emissivity mutation index , which are risk indicators representing external environmental interference and internal surface changes, respectively. By fusing them, a comprehensive score measuring the reliability of the current colorimetric temperature result is obtained.

[0065] Specifically, the calculation formula of the reliability score of the colorimetric temperature is as follows:

[0066]

[0067] In the formula, is the reliability score of the colorimetric temperature of the pixel point at the current time , and its value range is (0, 1]; is the spectral selectivity distortion factor of the pixel point at time ; is the surface emissivity mutation index of the pixel point at time ; and are dimensionless sensitivity coefficients for controlling the response speed of the weight to each indicator; represents a natural exponential function.

[0068] wherein, and take values between 5 and 20, and a higher value means that the system reacts more sharply to a small change in the indicator and the reliability decreases faster. In the present embodiment, and are both set to 8.

[0069] It should be noted that the exponential decay function can well simulate the process of the reliability smoothly decreasing with the increase of the abnormal indicators, and multiplying the two exponential terms ensures that the sharp deterioration of any one risk factor will have a decisive one-vote veto effect on the final reliability: when the spectral selectivity distortion factor and the surface emissivity mutation index are both close to 0, the working condition is stable, the exponential term is close to 1, and therefore the reliability score is close to 1, indicating high reliability; when the spectral selectivity distortion factor significantly increases, becomes a larger negative number, resulting in a reliability score rapidly approaches 0, indicating that the trustworthiness is reduced due to environmental disturbance; similarly, when the surface emissivity mutation index significantly increases, the trustworthiness score rapidly approaches 0, indicating that the trustworthiness is reduced due to surface state mutation.

[0070] Further, according to the trustworthiness score, the weight coefficients of the predicted temperature value and the colorimetric temperature value are set in a linear mapping manner, and are smoothly adjusted within a preset safe weight range, so that even in an extreme case, the system does not completely discard any information source, thereby enhancing the stability and robustness of the algorithm; the calculation formula of the weight coefficients of the predicted temperature value and the colorimetric temperature value is:

[0071]

[0072]

[0073] In the formula, is the maximum weight that the colorimetric temperature value can obtain, and is recommended to be set to 0.9, indicating that even in the most ideal working condition, the physical model is still given a basic weight of 10%, so as to maintain the continuity of the temperature; is the minimum weight that the colorimetric temperature value can obtain, and is recommended to be set to 0.1, indicating that even in the worst working condition, it is still believed that the actual measured value contains some information and cannot be completely discarded.

[0074] When the trustworthiness score approaches 1, the working condition is stable, approaches , at this time, the final temperature mainly depends on the colorimetric method; when the trustworthiness score approaches 0, strong disturbance or surface mutation occurs, approaches , at this time, the final temperature will mainly depend on the more stable physical prediction model .

[0075] It should be noted that in this way, the present application no longer independently solves the temperature, but couples the physical continuity of temperature change, the spectral characteristics of environmental disturbance and the mutation characteristics of surface state, realizes intelligent attribution of the change of radiation signal ratio, and thus obtains more accurate and robust temperature field monitoring results in a complex dynamic forging environment.

[0076] Exemplarily, for a pixel point, by analyzing the radiation intensity value of the pixel point in a radiation intensity image sequence of a continuously collected forging, a characteristic analysis result diagram as shown in Figure 2 and a temperature monitoring result diagram as shown in Figure 3 are obtained.

[0077] (1) The first 9 frames of radiation intensity images correspond to the stable cooling stage, at this stage, the environmental disturbance is weak, the surface state of the forging is stable, therefore, the spectral selectivity distortion factor and the surface emissivity mutation index are kept at a very low level, which in turn leads to the credibility score of the colorimetric temperature close to 1, and the weight of the colorimetric temperature close to its maximum value; At this time, the physical prediction temperature and the colorimetric temperature both follow a smooth cooling trend, and the values are very close. Since the weight of the colorimetric temperature is dominant, the final optimal temperature is mainly determined by the colorimetric temperature, which accurately reflects the natural cooling process of the forging.

[0078] (2) At the 10th frame of radiation intensity image, an interference event occurs, at this time, water vapor / smoke and oxide scale peeling occur simultaneously, the spectral selectivity distortion factor soars due to the severe distortion of the spectral signal, and the surface emissivity mutation index is caused by the dramatic change of the surface radiation intensity in space and time; The sharp increase of the two risk indicators at the same time, through the exponential decay function, leads to the sharp decline of the credibility score of the colorimetric temperature, and then the weight of the colorimetric temperature is dynamically adjusted to the vicinity of its minimum value set, at this time, the colorimetric temperature is severely disturbed, the reading is abnormally high, and it completely deviates from the normal cooling trend; However, since its weight is very low at this time, its influence on the final result is greatly weakened, at this time, the physical prediction temperature gives a predicted value according to the smooth data of the previous two frames, which conforms to the cooling law, and the final optimal temperature mainly adopts the physical prediction value with higher weight, successfully suppresses the large error of the colorimetric temperature, and maintains the continuity of the temperature sequence.

[0079] (3) The last 9 frames of radiation intensity images are disturbed, the spectral selectivity distortion factor and the surface emissivity mutation index gradually fall to normal level, the credibility score and the weight of the colorimetric temperature gradually recover to a higher level, and the reading of the colorimetric temperature gradually returns to the normal cooling track, with the weight of the colorimetric temperature rising, the optimal temperature starts to be mainly guided by the colorimetric temperature again, and continues to smoothly decline.

[0080] In summary, the spectral selectivity distortion factor and the surface emissivity mutation index are used to accurately identify two different types of abnormal events, external environmental disturbance and internal surface mutation, based on the diagnosis results, the weight of the colorimetric measurement value and the physical prediction value is dynamically adjusted through the credibility score of the colorimetric temperature, the influence of the unreliable measurement source, i.e. the colorimetric method, is reduced when the disturbance occurs, and the influence of the stable physical model is increased, when the colorimetric temperature deviates seriously due to the disturbance, the adaptive weighting mechanism effectively suppresses the error transmission to the final result, so that the optimal temperature estimation maintains relative continuity and reasonableness, after the disturbance disappears, the system can automatically restore the trust of the colorimetric measurement, so that it dominates the temperature estimation again.

Claims

1. A method for real-time monitoring of temperature field in a forging process based on infrared images, characterized in that, The method comprises: continuously collecting a sequence of radiation intensity images of the forging at multiple different infrared wavebands; calculating a predicted temperature value of each pixel point of the forging at a current time based on an optimal temperature estimation value at a previous time and Newton's cooling law; A waveband insensitive to the environment absorption is selected as a reference waveband, and a spectral selection distortion factor for characterizing the non-gray attenuation degree of the environmental medium is obtained by comparing the measured radiation intensity ratio of other wavebands with the reference waveband and the expected radiation intensity ratio calculated based on the predicted temperature value, comprising: ; wherein, is a pixel point At time , the spectral selection distortion factor is: is the total number of wavebands; indicates that all non-reference wavebands are summed up; is the expected radiation intensity at wavelength at time ; is the radiation intensity value of pixel point in the image at reference wavelength at time ; is the expected radiation intensity at wavelength at time ; is the expected radiation intensity at reference wavelength at time ; indicates taking the absolute value; The normalized spatial gradient and the normalized time gradient of the radiation intensity value at the reference wavelength are calculated, and the sum of the two is taken as a surface emissivity mutation index for characterizing the surface emissivity mutation of the forging surface, comprising: ; in the formula, is a pixel point At time The surface emissivity mutation index of the forging surface is obtained. , is The radiation intensity value of the pixel point at time at the reference wavelength in the image; is the average radiation intensity of all pixel points in the window centered on the pixel point at the reference wavelength ; The normalized spatial gradient of the radiation intensity value at the reference wavelength is represented by The normalized time gradient of the radiation intensity value at the reference wavelength is represented by obtaining a colorimetric temperature value of each pixel point of the forging at the current time through a colorimetric temperature measurement algorithm; inputting the spectral selectivity distortion factor and the surface state change index into a preset exponential decay function to obtain a reliability score of the colorimetric temperature value; and dynamically weighting and fusing the predicted temperature value and the colorimetric temperature value according to the reliability score to obtain an optimal temperature estimation value at the current time.

2. The method of claim 1, wherein, The multiple different infrared wavebands include 1.6 µm, 2.3 µm, 3.9 µm and 4.5 µm, wherein the reference wavelength is 3.9 µm.

3. The method of claim 1, wherein, The calculation of the predicted temperature value of each pixel point of the forging at the current time based on the optimal temperature estimation value at the previous time and Newton's cooling law comprises: ； In the formula, For pixels At any moment The predicted temperature value; For pixels In the previous moment The optimal temperature estimate; The time interval for image acquisition; This represents the overall cooling coefficient at the pixel location in the previous time step, and , For pixels At any moment The optimal temperature estimate; For a moment Ambient temperature, For the previous moment Ambient temperature.

4. The method of claim 1, wherein, The expected radiance is calculated using Planck's law for each wavelength based on the predicted temperature in an ideal situation without environmental attenuation The corresponding expected radiance.

5. The method of claim 1, wherein, The inputting of the spectral selectivity distortion factor and the surface state change index into the preset exponential decay function to obtain the reliability score of the colorimetric temperature value comprises: ； wherein is the pixel point at the current time of the colorimetric temperature; is the pixel point at the time of the spectral selectivity distortion factor; is the pixel point at the time of the surface emissivity mutation index; and is a dimensionless sensitivity coefficient for controlling the response speed of the weight to each index; denotes the natural exponential function.

6. The method of real-time monitoring of temperature field in a forging process based on infrared images according to claim 1, characterized in that, The dynamic weighting and fusing of the predicted temperature value and the colorimetric temperature value according to the reliability score to obtain the optimal temperature estimation value at the current time comprises: ； In the formula, is a pixel point at time is an optimal temperature estimation value; is a pixel point at time is a predicted temperature value; is a pixel point at time is a colorimetric temperature value; , are weight coefficients of the predicted temperature value and the colorimetric temperature value respectively, and .

7. The method of real-time monitoring of temperature field in a forging process based on infrared images according to claim 6, characterized in that, The calculation formula of the weight coefficients of the predicted temperature value and the colorimetric temperature value is: ； ； In the formula, It is the maximum weight that can be obtained from the colorimetric temperature value; It is the minimum weight that can be obtained from the colorimetric temperature value; For pixels At the present moment The confidence score of the colorimetric temperature.

8. The method of real-time monitoring of temperature field in a forging process based on infrared images according to claim 2, characterized in that, When the colorimetric temperature value of each pixel point of the forging at the current moment is obtained through the colorimetric temperature measurement algorithm, the wavelength = 1.6 pm and the wavelength = 2.3 pm are selected as the wavelength combination for colorimetric temperature measurement.

Images