Temperature drift cooperative compensation control system of photoelectric detector

Abstract

The invention discloses a temperature drift cooperative compensation control system of a photoelectric detector. A sensing module collects temperature distribution data of a sensitive element array and state data of a surface temperature-sensitive phase change coating; the modeling module receives multi-source data, processes the multi-source data through a digital twin model, and outputs a coating compensation contribution value, an electronic compensation demand coefficient and a performance drift prediction parameter; the decision module calculates cooperative execution parameters of the coating and electronic compensation based on a dynamic weight distribution algorithm; the compensation module comprises a temperature-sensitive phase change coating unit and an electronic compensation unit which are respectively used for carrying out physical compensation from a source and carrying out rear-end electronic compensation according to the collaborative parameters; and the feedback module compares a performance measured value with a predicted value, and drives self-optimization updating of the model and the parameters. According to the invention, dual-channel cooperative temperature compensation from a physical level to a circuit level is realized, and the gain stability, the time sequence precision and the energy resolution of the photoelectric detector in a wide temperature range are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and in particular to a temperature drift collaborative compensation control system for a photoelectric detector. Background Technology

[0002] High-precision photodetectors play a crucial role in numerous fields, including environmental monitoring, spectral analysis, industrial imaging, and scientific research. The core of these detectors typically contains temperature-sensitive sensing elements (such as scintillation crystals and semiconductor photosensitive devices), whose photoelectric conversion characteristics, including light output intensity, response time, and gain, are highly susceptible to changes in operating temperature. In practical applications, detectors often suffer from uneven temperature distribution, rapid localized temperature rise, and material aging due to long-term operation, resulting in temperature-induced performance drift exhibiting complex, nonlinear, and multi-parameter coupled characteristics.

[0003] To address temperature drift, traditional methods primarily rely on back-end electronic compensation, such as adjusting the bias voltage of photoelectric conversion devices or signal processing thresholds. However, this purely electronic compensation approach has inherent limitations: First, it is a passive, end-point compensation method that cannot physically suppress changes in the light output of the sensing element itself; second, it struggles to cope with drastic temperature changes or significant spatial temperature gradients, limiting response speed and compensation accuracy; and third, it lacks the ability to perceive and integrate the surface conditions of the sensing element (such as the aging of the protective coating), making it unable to adapt to changes in compensation requirements caused by gradual changes in material properties during long-term operation.

[0004] Therefore, existing technologies struggle to achieve high-precision, adaptive, and stable control of photodetector performance under varying temperature conditions, complex internal physical states of the detector, and long-term aging. There is an urgent need for an intelligent temperature drift compensation system that can integrate multi-source information, possess forward-looking predictive capabilities, and coordinate regulation from the physical source to the electronic link. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a temperature drift collaborative compensation control system for photodetectors, which can adaptively maintain the gain stability, timing accuracy, and energy resolution of photodetectors under wide temperature ranges and complex operating conditions.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a temperature drift collaborative compensation control system for a photodetector, comprising: The sensing module includes an implantable temperature sensing network and a coating state monitoring unit. The implantable temperature sensing network is used to collect temperature distribution data of the sensitive element array in the photodetector, and the coating state monitoring unit is used to collect coating state data of the temperature-sensitive phase change coating applied to the surface of the sensitive element. The modeling module, connected to the sensing module, is used to receive the temperature distribution data and coating state data, and, in combination with the detector operating parameters, process them through a digital twin model to output the coating compensation contribution value, electronic compensation demand coefficient, and performance drift prediction parameters. The decision module, connected to the modeling module, is used to receive the coating compensation contribution value, electronic compensation demand coefficient and performance drift prediction parameters, and process them through a dynamic weight allocation algorithm. The dynamic weight allocation algorithm dynamically calculates the collaborative execution parameters of coating compensation and electronic compensation based on the temperature change rate, coating aging state and individual characteristics of sensitive components. The compensation module, connected to the decision module, includes a temperature-sensitive phase change coating unit and an electronic compensation unit. The temperature-sensitive phase change coating unit is coated on the surface of the sensitive element, and its optical transmission characteristics are adaptively adjusted with temperature changes to compensate for the light output drift of the sensitive element from the source. The electronic compensation unit is used to perform back-end electronic compensation according to the cooperative execution parameters. The feedback module, connected to the compensation module and the modeling module, is used to collect the measured performance values ​​of the photodetector and compare the measured values ​​with the output values ​​of the modeling module to drive the self-optimization update of the digital twin model and coating state parameters.

[0007] Furthermore, the input parameters of the digital twin model of the modeling module include the temperature distribution data, the coating state data, and the detector operating parameters; The temperature distribution data includes the temperature between sensitive elements, the temperature around optoelectronic devices, and their location labels. The coating status data includes the real-time transmittance, refractive index, and aging factor calculated based on the cumulative working time of the temperature-sensitive phase change coating. The detector operating parameters include real-time count rate, detector operating time, and photoelectric device dark count rate. The digital twin model is a hybrid neural network model with an integrated attention mechanism layer. The attention mechanism layer dynamically weights the coating state parameters to establish a mapping relationship between the temperature field, coating state, and detector performance.

[0008] Furthermore, the digital twin model is trained and optimized through the following process: A training set was constructed by collecting historical data covering different temperature points, coating aging states, and count rate conditions. A multi-objective loss function is constructed based on the prediction errors of the coating compensation contribution value, electronic compensation demand coefficient, and performance drift prediction parameters for model training; The feedback module incrementally trains the digital twin model by periodically using quality control data obtained from standard tests.

[0009] Furthermore, the aging factor in the coating state parameters is calculated using the following formula: ; in, As an aging factor, This refers to the cumulative operating time of the temperature-sensitive phase change coating. For predefined maintenance thresholds, The time scaling parameter, the time scaling parameter Related to the material degradation rate of the coating, The shape parameter is a shape parameter. Related to the nonlinear acceleration characteristics of the coating aging process, The position parameter is a location parameter. Related to the characteristic time point when the coating undergoes significant aging, The scale parameter is the scale parameter. The consistency of coating aging rate varies across different batches of sensitive components, and The range of values ​​is ,when ≤ At that time, the feedback module triggers a maintenance reminder.

[0010] Furthermore, the attention mechanism layer is a coating characteristic attention mechanism layer, used to calculate the attention weights of the coating state parameters, wherein for the Each coating state parameter has an attention weight. Calculated using the following formula: ; in For the first Attention weights for each coating state parameter, This refers to the temperature value in the temperature field parameters. For the first The transmittance or refractive index of each coating state parameter The number of coating state parameters, The shape parameter of the Gamma function. For the first The scaling parameter of the error function corresponding to each coating state parameter. For the first The scaling parameter of the logarithmic function corresponding to each coating state parameter. For the first The parameters of the Riemann Zeta function for each coating state parameter, For the summation index, Let j be the current value of the j-th coating state parameter. The shape adjustment parameter of the Gamma function corresponding to the j-th coating state parameter is... The temperature coupling scaling parameter corresponds to the j-th coating state parameter. Let be the numerical scaling parameter for the j-th coating state parameter. The Riemann Zeta function adjustment parameter is the parameter corresponding to the j-th coating state parameter, and The range of values ​​is (0,1) and all The sum is 1.

[0011] Furthermore, the performance drift prediction parameters include gain drift, timing performance change, and energy resolution change rate; the cooperative execution parameters include optoelectronic device bias adjustment, signal discrimination threshold adjustment, and data weighting coefficient.

[0012] Furthermore, the dynamic weight allocation algorithm of the decision module is specifically used for: The electronic compensation requirement coefficient is dynamically calculated based on the temperature change rate, the coating aging state, and the pre-stored individual sensitivity coefficients of the sensitive elements. The bias adjustment of the optoelectronic device is calculated based on the electronic compensation demand coefficient, the gain drift, and the temperature change rate. The signal discrimination threshold adjustment amount is calculated based on the electronic compensation demand coefficient and the time-series performance change amount, combined with the temperature change rate factor. The data weighting coefficient is generated using the energy resolution change rate, the difference between the current temperature and the reference temperature, and the electronic compensation requirement coefficient.

[0013] Furthermore, the electronic compensation unit includes: A bias voltage adjustment circuit is used to receive the bias voltage adjustment amount of the optoelectronic device and output a corresponding bias voltage adjustment signal to the optoelectronic device. A threshold adjustment circuit is used to receive the signal, determine the threshold adjustment amount, and output the threshold adjustment signal to the discriminator; A data correction circuit is used to receive the data weighting coefficients and perform real-time weighted correction on the detector output data.

[0014] Furthermore, the temperature-sensitive phase change coating unit is a composite coating applied to the surface of the sensitive element, and its components include a temperature-sensitive polymer substrate, inorganic nanoparticle functional fillers, and interface modifiers. The coating has a thickness below a predefined thickness threshold, its transmittance at a reference temperature is higher than a first transmittance threshold, and the rate of increase of its transmittance with increasing temperature is configured to be inversely complementary to the rate of decrease of the light output of the sensitive element with increasing temperature.

[0015] Furthermore, the model self-optimization process of the feedback module is specifically as follows: The measured values ​​of the detector performance are compared with the output values ​​of the modeling module. When the deviation between the two exceeds a predefined deviation threshold, the incremental training process is automatically started. The parameters of the digital twin model are optimized and updated using the currently collected temperature distribution data, coating state data, and detector operating condition data.

[0016] The beneficial effects of this invention are: (1) Multi-dimensional and forward-looking temperature drift sensing and prediction are realized. The system synchronously collects temperature field distribution and coating physical state data through the sensing module, and performs fusion processing using an integrated digital twin model. It can forward-lookingly predict performance drift trends, breaking through the limitations of traditional single temperature detection or passive response compensation.

[0017] (2) An adaptive and dynamic collaborative compensation decision-making mechanism was constructed. The decision-making module adjusts the compensation strategy in real time through a dynamic weight allocation algorithm based on multi-dimensional parameters such as temperature change rate and coating aging state, and intelligently allocates the strength of physical compensation and back-end electronic compensation for the coating, so that the system can adapt to complex and time-varying working conditions.

[0018] (3) A dual-path compensation capability from the physical source to the signal link has been formed. The system combines the physical compensation at the source of the temperature-sensitive phase change coating with the precise adjustment at the back end of the electronic compensation unit to construct a dual-path collaborative compensation architecture, which effectively suppresses the cumulative effect of temperature disturbance in the detection link and improves the comprehensiveness and response speed of the compensation.

[0019] (4) It has the long-term stability capability of closed-loop feedback and continuous self-optimization. The system compares the measured and predicted performance through the feedback module, drives the online update of model parameters and compensation strategies, so that the system can adapt to the aging and performance changes of the detector during long-term operation and ensure stable and reliable operation throughout the entire life cycle. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the temperature drift collaborative compensation control system in this invention; Figure 2 This is a schematic diagram of the electronic compensation unit in this invention.

[0021] Reference numerals: 1. Sensing module; 11. Embedded temperature sensing network; 12. Coating state monitoring unit; 2. Modeling module; 3. Decision module; 4. Compensation module; 41. Temperature-sensitive phase change coating unit; 42. Electronic compensation unit; 421. Bias voltage adjustment circuit; 422. Threshold adjustment circuit; 423. Data correction circuit; 5. Feedback module. Detailed Implementation

[0022] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Identical components are denoted by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "upper," and "lower" used in the following description refer to directions in the accompanying drawings, and the terms "bottom surface," "top surface," "inner," and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0023] like Figure 1 As shown, Embodiment 1 discloses a temperature drift collaborative compensation control system for a photodetector. This embodiment is based on a high-precision photodetector experimental platform, and constructs a temperature drift collaborative compensation control system. The core detection unit of this platform is a 3×3 scintillation crystal array (as a sensitive element array), with each crystal measuring 4mm×4mm×20mm, coupled to a corresponding silicon photomultiplier tube (SiPM) array (as a photoelectric device array) for signal readout.

[0024] 1. System Construction and Hardware Configuration: (1) Preparation and coating of temperature-sensitive phase change coating unit 41: First, the surface of the scintillation crystal array was cleaned with argon plasma (parameters: power 100W, time 3 minutes, vacuum degree 10). -3 To enhance adhesion, an atomic layer deposition (ALD) process was used to form a uniform PNIPAM-TiO2-fluorinated ester composite thermosensitive phase change coating on the crystal surface at a deposition temperature of 80°C and a deposition rate of 0.5 nm / s. The coating thickness was approximately 80 nm. Subsequently, annealing was performed in a vacuum environment at 120°C for 30 minutes to improve coating stability. UV-Vis spectrophotometry showed that the transmittance of the coating at 25°C (reference temperature) was greater than 98% (meeting the first transmittance threshold), and its transmittance increased approximately linearly with increasing temperature. The rate of change was designed to be inversely complementary to the light output temperature decay rate of the scintillation crystal used (approximately -1.2% / °C).

[0025] (2) Deployment of Sensing Module 1: Implantable Temperature Sensing Network 11: The network consists of 24 CMOS miniature temperature sensors (model TMP117, accuracy ±0.1℃). Nine sensors are precisely embedded in the gaps of a 3×3 crystal array for direct measurement of the temperature between sensitive elements; the remaining sensors are arranged around the SiPM substrate and readout circuitry to measure the temperature around optoelectronic devices. All sensor data are tagged with location information.

[0026] Coating condition monitoring unit 12: Employs a miniature fiber optic spectral sensor (Ocean Insight FX series), with a probe diameter of 1.5 mm, a sampling frequency of 1 kHz, and a transmittance measurement accuracy of ±0.1%. This sensor is fixed to the inner wall of the detection module housing, aligned with the optical window on the side of the crystal array, and is used for real-time non-contact acquisition of the coating's real-time transmittance and refractive index.

[0027] (3) Deployment of Modeling and Decision Module 3: The digital twin model (CNN-LSTM hybrid architecture) and dynamic weight allocation algorithm are deployed on a heterogeneous computing platform. The model's forward inference (prediction) and decision-making algorithms are compiled and burned into a Xilinx Kintex-7 FPGA chip for real-time processing (latency <1ms). The model's training and optimization (backend) processes run on a server equipped with an Intel Xeon processor and an NVIDIA Tesla T4 GPU, interacting with the FPGA via a PCIe interface for high-speed data exchange and parameter synchronization.

[0028] (4) Configuration of Electronic Compensation Module 4: Bias regulation circuit 421: Its core is a 16-bit high-precision digital-to-analog converter (DAC) chip (ADI AD5060). Its regulation accuracy is ±0.01mV, which is the first predetermined threshold. The output of this DAC is directly connected to the bias input port of the SiPM array.

[0029] Threshold adjustment circuit 422: It consists of a high-speed comparator and a programmable voltage reference source, and outputs an adjustment signal to a constant ratio discriminator (as a discriminator).

[0030] Data correction circuit 423: Based on the digital multiplier implemented inside the FPGA, it performs real-time weighted calculation on the digitized pulse amplitude data.

[0031] (5) Configuration of Feedback Module 5: It includes a standard reference source (in this embodiment, a weak radiation source emitting monoenergetic gamma rays) and data quality analysis software. The system automatically performs standard tests periodically (e.g., every 24 hours) to collect measured performance values ​​of the detector's energy resolution, gain, and coincidence time resolution (as a measured benchmark for time-series performance variations).

[0032] 2. Working principle of this embodiment: The workflow of this system follows the closed-loop control principle of "perception-prediction-decision-execution-optimization", as detailed below: (1) Front-end physical self-compensation principle: The thermosensitive phase change coating serves as the first compensation barrier in this system. As ambient temperature rises, the light output of the scintillation crystal decreases (approximately -1.2% / ℃). Simultaneously, the thermosensitive polymer in the coating undergoes a phase change, with chain segment contraction and increased microporosity, leading to an increase in transmittance with increasing temperature (e.g., from 98% at 25℃ to 99.5% at 40℃). This physical change allows more photons emitted from the crystal to pass through the coating and be detected by the downstream SiPM, thus partially offsetting the attenuation of the crystal's own light output at the source. This process is entirely driven by material properties, with a response time in the millisecond range, achieving rapid initial compensation without the need for external energy or control.

[0033] (2) Principle of multi-source prediction for digital twins: The temperature distribution data (spatial temperature field), coating state data (transmittance, refractive index), and detector operating parameters (such as count rate) uploaded in real time by sensing module 1 are synchronously input into the digital twin model in the FPGA. This model is a trained CNN-LSTM hybrid network: the CNN part extracts the spatial distribution features of the temperature field; the LSTM part analyzes the time-series changes in coating state and temperature; and the integrated attention mechanism dynamically evaluates the weights of different coating state parameters (such as transmittance and aging factor) on current performance. The model ultimately outputs a quantitative coating compensation contribution value (predicting the amount of drift already compensated by the coating), an electronic compensation requirement coefficient (quantifying the remaining drift proportion that needs electronic compensation), and accurate predictions of gain drift, temporal performance changes, and energy resolution change rates. This model is equivalent to a "virtual image" of the detector, enabling early and accurate perception of future performance drift.

[0034] (3) Dynamic weight allocation and collaborative decision-making principle: Decision module 3 receives the output of modeling module 2 and executes a dynamic weight allocation algorithm. This algorithm is based not only on the current temperature but, more importantly, incorporates the temperature change rate (dT / dt), coating aging factor (A), and the individual sensitivity coefficient (factory calibrated) of each crystal-SIPM unit. For example, when the temperature changes rapidly, the algorithm significantly increases the electronic compensation demand coefficient, giving the backend electronic compensation a higher weight and faster response command to compensate for the response delay of the coating physical compensation. When the coating aging factor A is lower than a predefined aging threshold (set to 0.7 in this embodiment), the algorithm systematically increases the baseline weight of electronic compensation to address the degradation of the coating compensation capability. Through this multi-parameter, adaptive decision-making, the system calculates the optimal collaborative execution parameters in real time: optoelectronic device bias adjustment, signal discrimination threshold adjustment, and data weighting coefficients.

[0035] (4) Dual-path collaborative execution principle: Compensation module 4 receives the collaborative execution parameters and executes them in two separate streams: Physical pathway: The temperature-sensitive coating works continuously based on its own temperature response characteristics, providing basic and rapid compensation.

[0036] Electronic Path: The bias adjustment circuit 421 fine-tunes the SiPM operating bias voltage using a 16-bit DAC based on the bias adjustment amount of the optoelectronic device, precisely correcting its gain. The threshold adjustment circuit 422 adjusts the discriminator threshold based on the signal discrimination threshold adjustment amount to stabilize the measurement time (meeting time resolution). The data correction circuit 423 performs real-time digital correction on the pulse amplitude of each detection event based on the data weighting coefficients, optimizing the energy spectrum peak position and resolution. The total response time from receiving parameters to completing all electronic compensation actions is less than 10 μs (this value is better than the second predetermined threshold), ensuring the timeliness of compensation.

[0037] (5) Closed-loop feedback self-optimization principle: Feedback module 5 periodically obtains measured values ​​of key performance indicators such as energy resolution and gain through standard tests and compares them with the predicted values ​​of the digital twin model. When the deviation of any indicator continues to exceed a predefined deviation threshold (set to 5% in this embodiment), the system automatically triggers an incremental training process: the server retrains the digital twin model using all recently collected operational data, optimizes its parameters, and synchronizes the updated model to the FPGA. Simultaneously, the aging factor is recalculated based on the monitored coating data. When A ≤ 0.7, in addition to increasing the decision weight, a "coating maintenance reminder" signal is generated on the user interface. This allows the system to continuously improve itself over time and adapt to long-term aging.

[0038] 3. Technical effects of this embodiment: Under the experimental platform and the aforementioned working conditions, this system exhibits the following significant technical effects: (1) Improved gain stability across the entire temperature range: In the operating temperature range of 10℃ to 40℃, the system gain drift is suppressed to within ±0.3% / ℃, which is a fundamental improvement compared to the state without compensation or with single electronic compensation (drift can reach -1.2% / ℃).

[0039] (2) Enhanced stability of timing accuracy: During temperature cycling (ΔT=15℃), the system maintains a fluctuation range within ±12ps that meets the time resolution, effectively ensuring the accuracy of time measurement.

[0040] (3) Long-term stable energy resolution: During the continuous operation test of up to 1000 hours, the fluctuation of the system energy resolution remained at a relatively stable level of ±1.2%, which proved the effectiveness of dual-path compensation and closed-loop optimization.

[0041] (4) Rapid temperature disturbance response capability: Faced with rapid temperature changes of 0.5℃ / s, the system has millisecond-level physical response and microsecond-level electronic compensation capabilities, and the key performance indicators do not show instantaneous degradation, thus meeting the requirements of dynamic operating conditions.

[0042] (5) Adaptability and maintainability: After simulating coating aging (A=0.65), the system can still maintain the above (1)-(3) performance indicators within an acceptable range (such as gain drift <±0.5% / ℃) by automatically increasing the electronic compensation weight, and actively issue maintenance reminders, thus realizing the prediction and buffering of performance degradation.

[0043] Example 2 is the second embodiment of the present invention. Based on Example 1, this embodiment constructs a more complex application scenario to verify the system's compensation capability and adaptive mechanism under harsh conditions such as non-uniform temperature fields, high count rate loads, and simulated coating aging. The system is deployed in a multi-channel photodetector module composed of a 4×4 scintillation crystal array (as a sensing element array), with 16 corresponding SiPM (as a photoelectric device) readout channels.

[0044] In this embodiment, the input parameters of the digital twin model of the modeling module 2 include temperature distribution data, coating state data, and detector operating parameters; the temperature distribution data includes the temperature of the gap between the sensitive elements, the temperature around the optoelectronic device, and its location label; the coating state data includes the real-time transmittance, refractive index, and aging factor calculated based on the cumulative working time of the temperature-sensitive phase change coating; the detector operating parameters include the real-time count rate, the detector working time, and the dark count rate of the optoelectronic device. The digital twin model is a hybrid neural network model with an integrated attention mechanism layer. The attention mechanism layer dynamically weights the coating state parameters to establish a mapping relationship between the temperature field, coating state, and detector performance.

[0045] The digital twin model is trained and optimized through the following process: A training set was constructed by collecting historical data covering different temperature points, coating aging states, and count rate conditions. A multi-objective loss function was constructed using the prediction errors of coating compensation contribution value, electronic compensation demand coefficient, and performance drift prediction parameters for model training. Feedback module 5 incrementally trains the digital twin model by periodically using quality control data obtained from standard tests.

[0046] 1. Extended system configuration and testing conditions: (1) Complex environment setup: Temperature field: A non-uniform temperature gradient environment is artificially created within the detection module, so that the maximum temperature difference between crystals at different locations in the 4×4 array reaches 7°C, simulating the "hot spot" problem commonly found in actual equipment.

[0047] Operating load: Increase the detector count rate to 2×10⁻⁶ using an external radiation source. 6 cps (counts per second) simulates a high-throughput operating state.

[0048] Coating aging simulation: By injecting data of 8000 hours of cumulative working time into the software, the system calculates the aging factor based on this time to simulate the performance degradation of the coating after long-term operation.

[0049] (2) Enhanced deployment of sensing module 1: Implantable temperature sensing network 11: A total of 24 high-precision temperature sensors are deployed in the crystal gaps and key thermal nodes to continuously collect data streams of temperature between sensitive components and the temperature around optoelectronic devices with precise location tags.

[0050] Coating condition monitoring: A fiber optic spectrometer synchronously monitors the real-time transmittance and refractive index of all crystal surface coatings at a frequency of 1 kHz.

[0051] (3) Parameterization of the model and algorithm: This embodiment specifically applies the following mathematical model: The aging factor in the coating condition parameters is calculated using the following formula: ; in, As an aging factor, This refers to the cumulative operating time of the temperature-sensitive phase change coating. For predefined maintenance thresholds, For time scaling parameters, time scaling parameters Related to the material degradation rate of the coating; For shape parameters, shape parameters Related to the nonlinear acceleration characteristics of the coating aging process, For position parameters, position parameters Related to the characteristic time point when the coating undergoes significant aging, For scale parameters, scale parameters The consistency of coating aging rate varies across different batches of sensitive components, and The range of values ​​is ,in This indicates that the coating is in brand new condition and has not aged. = This indicates that the coating has reached the maintenance threshold and needs to be replaced or calibrated. The value range linearly maps the degree of coating performance degradation, ensuring the long-term operational stability of the system; when ≤ When the feedback module triggers a maintenance reminder, the system increases the electronic compensation weight, ensuring that the aging system still maintains stable performance.

[0052] The attention mechanism layer is a coating property attention mechanism layer used to calculate the attention weights of the coating state parameters, where for the Each coating state parameter has an attention weight. Calculated using the following formula: ; in For the first Attention weights for each coating state parameter, This refers to the temperature value in the temperature field parameters. For the first The transmittance or refractive index of each coating state parameter The number of coating state parameters, The shape parameter of the Gamma function. For the first The scaling parameter of the error function corresponding to each coating state parameter. For the first The scaling parameter of the logarithmic function corresponding to each coating state parameter. For the first The parameters of the Riemann Zeta function for each coating state parameter, For the summation index, Let j be the current value of the j-th coating state parameter. The shape adjustment parameter of the Gamma function corresponding to the j-th coating state parameter is... The temperature coupling scaling parameter corresponds to the j-th coating state parameter. Let be the numerical scaling parameter for the j-th coating state parameter. The Riemann Zeta function adjustment parameter is the parameter corresponding to the j-th coating state parameter, and The range of values ​​is (0,1) and all The sum of 1 indicates the relative importance of each coating state parameter in the attention mechanism. The closer the value is to 1, the greater its contribution to temperature drift compensation. The accuracy of the output model is optimized through dynamic weighting.

[0053] The role of the attention mechanism in this embodiment is as follows: This mechanism enables the digital twin model to: increase the importance of the transmittance parameter when the temperature changes rapidly, increase the weight of the aging factor when the coating aging is obvious, and strengthen the attention to the refractive index when the coating refractive index fluctuates greatly, thereby increasing the model's adaptability to complex and dynamic working conditions.

[0054] In this embodiment, the aging factor model parameters are: a predefined maintenance threshold is set. =0.7. The parameters in the formula are calibrated based on accelerated aging experiments of the coating as follows: time scaling parameter. =1000 hours, shape parameters =1.2, position parameter =1.5, scale parameter =0.3.

[0055] Attention mechanism parameters: By training, the parameters in the coating characteristic attention mechanism layer are determined, so that the model can dynamically adjust the attention to different coating state parameters according to real-time operating conditions.

[0056] 2. Working principle of this embodiment two: Under the complex operating conditions set in this embodiment, the advanced functions of system collaborative compensation and adaptive optimization are fully demonstrated: (1) Multi-source heterogeneous data fusion and 3D feature modeling: The system inputs spatial temperature field data consisting of 24 temperature nodes into the CNN part of the digital twin model to generate a temperature distribution heatmap. Simultaneously, the coating transmittance, refractive index, and calculated aging factor A from 16 channels are input into the LSTM part, calculated based on 8000 hours of cumulative operation. At this point, the A value drops to approximately 0.72, close to the predefined aging threshold of 0.7. The coating characteristic attention mechanism layer then plays a crucial role: in regions with rapidly rising temperatures, this mechanism automatically assigns higher attention weights αi to the transmittance parameter, as the impact of transmittance changes on luminous flux is most significant at this point; while for coatings with long cumulative operating times, the weight of the aging factor is increased, alerting the model that its compensation capability may be declining. This dynamic weighting achieves a high-dimensional and accurate mapping of the complex nonlinear relationship between "temperature-coating state-performance".

[0057] (2) Cooperative decision-making of dynamic weight allocation algorithm under extreme conditions: The decision module 3 receives the coating compensation contribution value, electronic compensation demand coefficient, and various performance drift prediction parameters output by the modeling module 2. In this embodiment: Faced with a spatial temperature difference of up to 7°C, the algorithm combines temperature distribution data to independently calculate the electronic compensation demand coefficient for each channel. For high-temperature "hotspot" channels, the coefficient value is higher, triggering stronger local electronic compensation.

[0058] Since the simulated coating aging factor A=0.72, the algorithm increases the weight of electronic compensation on a global baseline to partially offset the physical compensation efficiency that may be reduced due to simulated aging, thus reflecting the predictive maintenance logic.

[0059] In 2×10 6 At a high count rate of cps, the model can effectively distinguish between systematic gain changes caused by temperature drift and signal fluctuations caused by random noise, thereby ensuring that the calculated gain drift and timing performance changes are more accurate.

[0060] (3) Precise execution of dual-path compensation in non-uniform fields: The compensation module 4 performs differentiated compensation based on the collaborative execution parameters (optoelectronic device bias adjustment, signal discrimination threshold adjustment, and data weighting coefficients) generated for each channel according to the above decisions. For example, for channels in high-temperature regions, the system applies a larger bias adjustment to increase gain and fine-tunes the discriminator threshold adjustment for that channel to stabilize the time response. This "zonal management and precise control" mode is key to addressing uneven spatial temperature.

[0061] (4) Incremental training and state update of the model based on closed-loop feedback: During the week-long continuous testing, Feedback Module 5 periodically performed standard tests. When the deviation between the measured performance values ​​(such as energy resolution) and the model predictions of certain channels under high temperature and high load conditions exceeded a predefined deviation threshold (still 5%), the system automatically initiated incremental training. The training used all temperature distribution data, coating state data, and high count rate operating parameters collected in this round of testing to fine-tune the digital twin model, especially optimizing its predictive ability for coupled operating conditions of "large temperature difference + high load". At the same time, the system continuously monitored the aging factor, and when the simulated value approached or fell below 0.7, in addition to adjusting the algorithm weights, it also recorded a warning message in the log.

[0062] 3. Technical effects of this embodiment: Under the stringent testing conditions described in Example 2, this system achieved the following verification technical results: (1) Effectiveness of non-uniform temperature field compensation: Under the maximum inter-crystal temperature difference of 7°C, the system successfully controlled the gain consistency difference between channels within ±3%, and maintained the overall system coincidence time resolution fluctuation within ±15ps, effectively suppressing the performance differentiation caused by local hot spots.

[0063] (2) Prediction accuracy under high noise conditions: at 2×10 6Under high count rate noise backgrounds, the system's prediction error for gain drift is less than 0.8%, and its prediction error for energy resolution change rate is less than 0.5%, demonstrating the model's excellent noise resistance and feature extraction capabilities.

[0064] (3) System robustness under coating aging simulation: Under the coating aging simulation (A≈0.72), the electronic compensation weight was automatically increased by the decision algorithm, and the overall gain drift rate of the system was still stabilized at 0.4% / ℃. No compensation failure due to coating performance degradation occurred, which verified the system's adaptive capability.

[0065] (4) Interpretability and effectiveness of the attention mechanism: Data analysis shows that during the stage of drastic temperature change, the attention weight αi of the coating transmittance parameter in the attention mechanism increased by an average of about 40%, successfully guiding the model to focus on the most critical changing factors and improving the accuracy of prediction and decision-making.

[0066] (5) Overall performance maintenance under complex coupled conditions: Under multiple stress tests of “non-uniform temperature field + high count rate + simulated aging”, the system maintained the long-term (168 hours) stability of the energy resolution of all channels, with a fluctuation range better than ±2.0%, which comprehensively demonstrated its reliability and robustness under near-extreme conditions.

[0067] This embodiment fully demonstrates that the system described in this invention is not only applicable to ideal or single-factor changing scenarios, but also has the comprehensive temperature drift compensation capability to handle complex, coupled, and extreme working conditions in the real world.

[0068] Example 3 is the third embodiment of the present invention. This embodiment aims to verify the specific structure, execution mechanism, and long-term optimization effect of the system in a near-realistic, continuously operating application scenario. The test was conducted on a multi-channel photoelectric detection system that ran continuously for 30 days, during which the ambient temperature experienced daily periodic fluctuations (20℃-35℃) and several sudden temperature rises.

[0069] 1. Working principle of this embodiment three: In the long-term, dynamic operation of this embodiment, the system demonstrates its accurate, coordinated, and adaptive compensation mechanism: (1) Linkage between prediction and decision-making based on specific performance parameters: The performance drift prediction parameters output by modeling module 2 explicitly include the gain drift. The decision module receives these specific parameters, including the temporal performance change (ΔTres) and the energy resolution change rate (δER), and then executes a dynamic weight allocation algorithm.

[0070] The algorithm first considers the real-time monitored rate of temperature change. Coating aging factors and the individual sensitivity coefficient of each channel The electronic compensation demand coefficient is dynamically calculated according to a predefined formula. The dynamic calculation formula for the electronic compensation demand coefficient is as follows: ; in, Indicates the electronic compensation demand coefficient. Indicates the rate of temperature change. Time scaling parameter representing the rate of temperature change. This represents the individual sensitivity coefficient of the crystal.

[0071] It quantifies the proportion of compensation that the back-end electronic system needs to bear under specific operating conditions. It is a weighting coefficient between 0 and 1.

[0072] Error function part: This constitutes a smooth "S-shaped" switching function. Its input is the rate of temperature change. The value is obtained by taking the logarithm and normalizing. This reflects that the system's sensitivity to different rates of temperature change is non-linear. When the temperature changes drastically ( (Large), the function value approaches 1, requiring fast response from electronic compensation.

[0073] When the temperature changes slowly ( (Small), the function value approaches 0, relying more on the front-end coating for compensation.

[0074] Aging factors section: This demonstrates the impact of coating performance degradation on compensation strategies. As the coating ages (…), (reduce) Increasing this ratio means the system will dynamically increase the proportion of electronic compensation to compensate for the decrease in coating compensation effect; this is a predictive maintenance logic.

[0075] Individual sensitivity coefficient These are unique parameters calibrated for each crystal unit to compensate for individual performance differences caused by the manufacturing process, thus solving the uniformity problem caused by the "one-size-fits-all" calibration parameters in the prior art.

[0076] Explanation of the range: The range of values ​​is . =0 indicates that electronic compensation is not required under the current operating conditions (ideally, the coating will fully compensate for it). =1 indicates that the electronic compensation system needs to work at full capacity to offset performance drift, with intermediate values ​​providing proportional compensation. In a clinical system environment, this gives the compensation system the ability to intelligently adapt to complex temperature disturbances.

[0077] For example, during the period when temperatures rise rapidly at midday, Increase Accordingly, the system is preparing to initiate a stronger electronic compensation response.

[0078] Next, the algorithm utilizes The specific system gain drift is precisely calculated using a piecewise formula that includes an error function enhancement term, thereby accurately determining the bias adjustment of the optoelectronic device. .

[0079] Bias adjustment amount The formula for segmented calculation is: ; in, This is the bias voltage adjustment amount. This is the bias reference value. This represents the system gain drift. This is the system gain reference value. For electronic compensation demand coefficient, This is the amplitude parameter for bias adjustment. This is the scaling parameter for bias adjustment.

[0080] This formula is used to calculate the stable system gain. The amount of bias adjustment required.

[0081] Reference adjustment amount: It is the basic adjustment amount calculated based on the gain drift ratio, which is a feedforward control loop.

[0082] Electronic compensation weight: multiplied by This means that the final bias adjustment is linked to the electronic compensation requirement factor. If the coating has already compensated for most of the drift ( If the voltage is small, the electronic bias adjustment range will also decrease accordingly, thus avoiding overcompensation.

[0083] Dynamic response enhancements: This is the core of the innovative concept in this formula. The error function here acts as a dynamic gain controller: When the rate of temperature change When the value is high, this value will be greater than 1, thus amplifying the bias adjustment. This makes the system more agile in responding to rapid temperature changes, overcoming the bottleneck of lag in the response of traditional PID algorithms.

[0084] When the temperature is stable or changes slowly, this value is approximately equal to 1, returning to the basic precise adjustment mode.

[0085] Explanation of the range: The range is theoretically However, in actual hardware, due to the limitations of the DAC range, it is usually constrained to... Within a certain range. Its sign determines the direction of bias adjustment (increase or decrease), and the absolute value represents the magnitude of adjustment.

[0086] This formula is designed so that when When the value is large, the error function value increases, thus dynamically amplifying. The calculation results enable a rapid and strong response to sudden temperature changes.

[0087] Signal discrimination threshold adjustment amount The formula is: ; in, The threshold adjustment amount for signal discrimination. The proportional coefficient for threshold adjustment. To match the changes in time resolution The reference parameter for adjustment, The amplitude parameter to be adjusted. For adjusting the time scaling parameter, for function.

[0088] Core idea: This formula is used to stabilize the time resolution and compensate for time jitter caused by temperature by adjusting the threshold of the constant ratio discriminator. ).

[0089] Basic adjustments: It is the basic part, which is proportionally adjusted according to the amount of drift in time resolution and the requirements of electronic compensation.

[0090] Function enhancements: This is the innovative aspect of this formula. The characteristic of the function here is that it grows faster than linearly but slower than exponentially. When the rate of temperature change... When it increases, The function value will increase significantly, thereby dynamically increasing the adjustment level of the threshold.

[0091] This design is based on a precise modeling of the nonlinear physical process of the crystal fluorescence decay time constant changing with temperature (-5.5 ns / ℃), which enables electronic compensation to better match the changes in the underlying physical mechanism, something that simple linear compensation cannot achieve.

[0092] Interpretation of the range: Its theoretical range is However, its practical applications are limited by circuitry. It precisely provides the amount of voltage fine-tuning required to maintain the accuracy of time measurement.

[0093] The formula for the data weighting coefficient is: ; Weighting coefficients for the data, For electronic compensation demand coefficient, The rate of change of energy resolution. The current temperature. As the reference temperature, This is a scaling parameter for changes in energy resolution. This refers to the temperature range parameter.

[0094] (2) Specific execution of electronic compensation unit 42: The electronic compensation unit 42, based on the aforementioned coordinated execution parameters, performs a millisecond-level response through three dedicated circuits, referring to... Figure 2 : Bias voltage regulation circuit 421: Receives the specific bias voltage regulation amount It outputs analog voltage signals through its core high-precision 16-bit DAC chip (adjustment accuracy ±0.01mV, which is the first predetermined threshold) to directly adjust the operating point of each optoelectronic device and accurately correct the gain.

[0095] Threshold adjustment circuit 422: Receives specific signals and determines the threshold adjustment amount. Through a high-speed comparator and a programmable reference source, the output adjusted voltage is sent to a constant ratio discriminator (discriminator) to stabilize the time measurement.

[0096] Data correction circuit 423: Receives specific data weighting coefficients. Within the FPGA, the amplitude of each digitized pulse is subjected to real-time scalar multiplication to optimize the energy spectrum shape.

[0097] The total response time from receiving parameters to completing all circuit adjustments is less than 10 μs, which is better than the requirement of the second predetermined threshold.

[0098] (3) The source compensation effect of the temperature-sensitive coating: The temperature-sensitive phase change coating unit 41 continuously operates as a physical compensation pathway. The coating has a transmittance >98% (first transmittance threshold) at a reference temperature of 25°C, and its thickness is less than 100 nm (predefined thickness threshold). Its positive temperature coefficient of transmittance is design-complementary to the negative temperature coefficient of light output from the sensitive element. When the temperature changes, this physical layer provides rapid, passive initial compensation, synergistically working with electronic compensation.

[0099] (4) Feedback-driven continuous self-optimization: Feedback module 5 continuously runs its self-optimization process. The system automatically performs standard tests daily to obtain measured performance values ​​such as gain and energy resolution. When the deviation between the measured value of any indicator and the predicted value of modeling module 2 exceeds a predefined deviation threshold (set to 5% in this embodiment), the system automatically triggers incremental training. This process utilizes newly acquired temperature, coating, and operating condition data to optimize and update the parameters of the digital twin model. Simultaneously, the system calculates the aging factor A based on monitoring data. When the value of A is lower than the predefined aging threshold (0.7), it automatically increases the baseline weight of the electronic compensation demand coefficient in the decision-making algorithm and records maintenance prompts, thus achieving predictive management of performance degradation.

[0100] 2. Technical effects of this embodiment: During a continuous dynamic operation test lasting 30 days, this system demonstrated the following verifiable technical effects: (1) High stability of gain under dynamic temperature: During daily temperature cycles and sudden temperature rises, the maximum instantaneous drift of the system gain is limited to within 0.4% / ℃, and the standard deviation of gain fluctuation over the entire cycle is less than 0.8%.

[0101] (2) Precise maintenance of timing performance: Throughout the entire test period, the system strictly controlled the fluctuations in time resolution (the manifestation of the change in timing performance) within ±18ps. Even during the rapid temperature change phase, there was no deterioration, which verified the accuracy of the calculation and adjustment of the signal discrimination threshold.

[0102] (3) Long-term consistency of energy resolution: Thanks to the real-time correction of data weighting coefficients and the self-optimization of the model, the system energy resolution changes at an excellent rate of ±1.5% over 30 days, demonstrating the long-term stability of the closed-loop system.

[0103] (4) Rapid response to sudden operating conditions: Two sudden temperature rises (heating rate > 0.3℃ / s) were simulated during the test. The system completed the compensation response within 100 milliseconds by significantly increasing the signal discrimination threshold adjustment and bias voltage adjustment. There was no visible step degradation in key performance indicators, and the total response time met the second predetermined threshold (10μs) requirement, which confirmed the rapid execution capability of the electronic compensation unit 42.

[0104] (5) Actual effectiveness of system self-optimization: During the test, the system triggered three incremental model trainings. After training, the average deviation between the model predictions and subsequent measured values ​​decreased from 4.2% before the trigger to 2.1%, which is significantly lower than the deviation threshold of 5%.

[0105] This embodiment fully demonstrates the high reliability and superior performance of the system of the present invention in real and complex application environments through long-term and dynamic field testing.

[0106] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A temperature drift collaborative compensation control system for a photodetector, characterized in that, include: The sensing module includes an implantable temperature sensing network and a coating state monitoring unit. The implantable temperature sensing network is used to collect temperature distribution data of the sensitive element array in the photodetector, and the coating state monitoring unit is used to collect coating state data of the temperature-sensitive phase change coating applied to the surface of the sensitive element. The modeling module, connected to the sensing module, is used to receive the temperature distribution data and coating state data, and, in combination with the detector operating parameters, process them through a digital twin model to output the coating compensation contribution value, electronic compensation demand coefficient, and performance drift prediction parameters. The decision module, connected to the modeling module, is used to receive the coating compensation contribution value, electronic compensation demand coefficient and performance drift prediction parameters, and process them through a dynamic weight allocation algorithm. The dynamic weight allocation algorithm dynamically calculates the collaborative execution parameters of coating compensation and electronic compensation based on the temperature change rate, coating aging state and individual characteristics of sensitive components. The compensation module, connected to the decision module, includes a temperature-sensitive phase change coating unit and an electronic compensation unit. The temperature-sensitive phase change coating unit is coated on the surface of the sensitive element, and its optical transmission characteristics are adaptively adjusted with temperature changes to compensate for the light output drift of the sensitive element from the source. The electronic compensation unit is used to perform back-end electronic compensation according to the cooperative execution parameters. The feedback module, connected to the compensation module and the modeling module, is used to collect the measured performance values ​​of the photodetector and compare the measured values ​​with the output values ​​of the modeling module to drive the self-optimization update of the digital twin model and coating state parameters.

2. The temperature drift collaborative compensation control system according to claim 1, characterized in that: The input parameters of the digital twin model of the modeling module include the temperature distribution data, the coating state data, and the detector operating parameters; The temperature distribution data includes the temperature between sensitive elements, the temperature around optoelectronic devices, and their location labels. The coating status data includes the real-time transmittance, refractive index, and aging factor calculated based on the cumulative working time of the temperature-sensitive phase change coating. The detector operating parameters include real-time count rate, detector operating time, and photoelectric device dark count rate. The digital twin model is a hybrid neural network model with an integrated attention mechanism layer. The attention mechanism layer dynamically weights the coating state parameters to establish a mapping relationship between the temperature field, coating state, and detector performance.

3. The temperature drift collaborative compensation control system according to claim 2, characterized in that: The digital twin model is trained and optimized through the following process: A training set was constructed by collecting historical data covering different temperature points, coating aging states, and count rate conditions. A multi-objective loss function is constructed based on the prediction errors of the coating compensation contribution value, electronic compensation demand coefficient, and performance drift prediction parameters for model training; The feedback module incrementally trains the digital twin model by periodically using quality control data obtained from standard tests.

4. The temperature drift collaborative compensation control system according to claim 2, characterized in that: The aging factor in the coating condition parameters is calculated using the following formula: ； in, As an aging factor, This refers to the cumulative operating time of the temperature-sensitive phase change coating. For predefined maintenance thresholds, The time scaling parameter is... Related to the material degradation rate of the coating, The shape parameter is a shape parameter. Related to the nonlinear acceleration characteristics of the coating aging process, The position parameter is a location parameter. Related to the characteristic time point when the coating undergoes significant aging, The scale parameter is the scale parameter. The consistency of coating aging rate varies across different batches of sensitive components, and The range of values ​​is ,when ≤ At that time, the feedback module triggers a maintenance reminder.

5. The temperature drift collaborative compensation control system according to claim 2, characterized in that: The attention mechanism layer is a coating characteristic attention mechanism layer, used to calculate the attention weights of the coating state parameters, wherein for the Each coating state parameter has an attention weight. Calculated using the following formula: ； in For the first Attention weights for each coating state parameter, This refers to the temperature value in the temperature field parameters. For the first The transmittance or refractive index of each coating state parameter The number of coating state parameters, The shape parameter of the Gamma function. For the first The scaling parameter of the error function corresponding to each coating state parameter. For the first The scaling parameter of the logarithmic function corresponding to each coating state parameter. For the first The parameters of the Riemann Zeta function for each coating state parameter, For the summation index, Let j be the current value of the j-th coating state parameter. The shape adjustment parameter of the Gamma function corresponding to the j-th coating state parameter is... The temperature coupling scaling parameter corresponds to the j-th coating state parameter. Let be the numerical scaling parameter for the j-th coating state parameter. The Riemann Zeta function adjustment parameter is the parameter corresponding to the j-th coating state parameter, and The range of values ​​is (0,1) and all The sum is 1.

6. The temperature drift collaborative compensation control system according to claim 1, characterized in that: The performance drift prediction parameters include gain drift, timing performance change, and energy resolution change rate; the cooperative execution parameters include optoelectronic device bias adjustment, signal discrimination threshold adjustment, and data weighting coefficient.

7. The temperature drift cooperative compensation control system according to claim 6, characterized in that: The dynamic weight allocation algorithm of the decision module is specifically used for: The electronic compensation requirement coefficient is dynamically calculated based on the temperature change rate, the coating aging state, and the pre-stored individual sensitivity coefficients of the sensitive elements. The bias adjustment of the optoelectronic device is calculated based on the electronic compensation demand coefficient, the gain drift, and the temperature change rate. The signal discrimination threshold adjustment amount is calculated based on the electronic compensation demand coefficient and the time-series performance change amount, combined with the temperature change rate factor. The data weighting coefficient is generated using the energy resolution change rate, the difference between the current temperature and the reference temperature, and the electronic compensation requirement coefficient.

8. The temperature drift collaborative compensation control system according to claim 7, characterized in that: The electronic compensation unit includes: A bias voltage adjustment circuit is used to receive the bias voltage adjustment amount of the optoelectronic device and output a corresponding bias voltage adjustment signal to the optoelectronic device. A threshold adjustment circuit is used to receive the signal, determine the threshold adjustment amount, and output the threshold adjustment signal to the discriminator; A data correction circuit is used to receive the data weighting coefficients and perform real-time weighted correction on the detector output data.

9. The temperature drift collaborative compensation control system according to claim 1, characterized in that: The temperature-sensitive phase change coating unit is a composite coating applied to the surface of the sensitive element, and its components include a temperature-sensitive polymer substrate, inorganic nanoparticle functional fillers, and interface modifiers. The coating has a thickness below a predefined thickness threshold, its transmittance at a reference temperature is higher than a first transmittance threshold, and the rate of increase of its transmittance with increasing temperature is configured to be inversely complementary to the rate of decrease of the light output of the sensitive element with increasing temperature.

10. The temperature drift collaborative compensation control system according to claim 1, characterized in that: The model self-optimization process of the feedback module is as follows: The measured values ​​of the detector performance are compared with the output values ​​of the modeling module. When the deviation between the two exceeds a predefined deviation threshold, the incremental training process is automatically started. The parameters of the digital twin model are optimized and updated using the currently collected temperature distribution data, coating state data, and detector operating condition data.

Images