A precision control method for liquid nitrogen ultra-low temperature test

Abstract

This application discloses a precision control method for liquid nitrogen cryogenic testing, comprising: deploying primary and secondary temperature sensors and a triaxial accelerometer in the near field of the specimen and synchronizing them with a Hopkinson bar system to predict the stress wave interference time window; calculating vibration intensity in real time, arbitrating and identifying spurious temperature sensor signals caused by impact vibration, and switching to a reliable virtual true temperature value; simultaneously, defining a cold energy health index by monitoring liquid nitrogen storage and system status, and dynamically switching control modes based on this index; and implementing a special strategy of locking and smooth recovery during impact transients. This invention effectively shields measurement distortion caused by dynamic interference and achieves a balance between control precision and system safety under resource constraints, significantly improving the environmental stability and process controllability of cryogenic impact testing.

Description

Technical Field

[0001] This invention relates to the field of cryogenic testing technology, and in particular to a method for precision control in liquid nitrogen cryogenic testing. Background Technology

[0002] Achieving high-precision temperature control in cryogenic Hopkinson bar (SHPB / SHTB) impact tests based on liquid nitrogen cooling faces two major challenges. First, the strong stress wave generated during the impact loading process is transmitted through the mechanical structure, causing severe vibrations in the test chamber and internal sensors. This vibration can lead to piezoelectric effects or signal distortion in critical temperature sensors mounted adjacent to the specimen, resulting in the output of momentary spurious temperature jump signals. If the control system misinterprets this spurious signal as a real change in ambient temperature, it will trigger completely erroneous or even reversed compensation adjustments, severely compromising the stability of the test conditions and potentially interfering with the measurement of the material's transient mechanical response.

[0003] Secondly, the continuous consumption of liquid nitrogen during the experiment leads to a constant decrease in the refrigerant reserve. Existing control methods typically monitor liquid level and pressure independently, failing to deeply couple the remaining cold energy—a critical resource—with the closed-loop control logic. When the liquid level is low, if the control system mechanically executes strong cooling commands based on a high liquid level model, it can easily lead to premature depletion of liquid nitrogen, unexpected interruption of the experiment, and even the risk of the evaporator burning out due to insufficient refrigerant. Summary of the Invention

[0004] This application provides a precision control method for liquid nitrogen cryogenic testing, which can ensure environmental accuracy and safety throughout the entire testing process under extreme dynamic loads and limited resource constraints.

[0005] This application provides a method for precision control in liquid nitrogen cryogenic testing, including: S1. Install main and auxiliary temperature sensors and a triaxial high-frequency acceleration sensor in the near field region of the specimen, establish communication with the data acquisition system of the Hopkinson rod, and acquire the stress wave initiation signal detected by the bullet trigger or strain gauge on the rod in real time. S2, based on the known wave velocity and propagation path length, predicts in advance the time window for the stress wave to reach the test chamber and temperature sensor; simultaneously reads the liquid level and pressure of the test chamber to calculate the liquid level change rate, and defines a cooling capacity health index in combination with the current air pressure and evaporator status; S3, In the impact interference test, the amplitude of the triaxial acceleration composite vector is calculated in real time. When the amplitude exceeds the preset vibration interference threshold or the current time is within the predicted stress wave arrival window, the data of the main temperature sensor is immediately marked as suspected interference, and an arbitration comparison group is established to determine whether the data of the main temperature sensor is a false signal. S4, during the period when it is determined to be a false signal, the input of the main temperature sensor data is cut off. The weighted average of the measured value and the short-term predicted value of the auxiliary temperature sensor is used as the virtual true value of the current test chamber temperature. When the vibration amplitude is less than 0.2g and the data is stable for three consecutive sampling cycles, the main temperature sensor is smoothly switched back. S5 dynamically selects the control mode based on the cooling capacity health index and couples it with a special strategy for shock transients. It generates closed-loop execution monitoring and event recording analysis through coordinated commands.

[0006] Preferably, the cold energy health index specifically includes: real-time reading of the liquid nitrogen storage tank's liquid level and pressure signals via a communication interface; continuous sampling and differential calculation of the liquid level signals to obtain the real-time liquid level change rate; preset a safe liquid level threshold and a critical liquid level threshold; and dynamically determining the index value using different calculation rules based on the real-time liquid level relative to the aforementioned thresholds: when the liquid level is higher than the safe threshold, the index is set to a fixed value representing abundant resources; when the liquid level is between the critical threshold and the safe threshold, the index is calculated using a linear interpolation method; when the liquid level is lower than the critical threshold, the index is calculated based on a composite function of liquid level, pressure, and liquid level change rate, and its upper limit is limited to represent a severely constrained resource state.

[0007] Preferably, the step of establishing an arbitration comparison group to determine whether the data from the main temperature sensor is a spurious signal includes: simultaneously acquiring the readings of the main temperature sensor marked as potentially interfered, the readings of the auxiliary temperature sensor, and a short-term temperature prediction value established based on historical temperature data to form an arbitration comparison group; comparing the main temperature sensor reading with the auxiliary temperature sensor reading and the short-term temperature prediction value respectively, and comparing them with a preset temperature deviation threshold; simultaneously, comparing the auxiliary temperature sensor reading with the short-term temperature prediction value, and also comparing them with the same preset temperature deviation threshold; if the difference between the main temperature sensor reading and the other two exceeds the threshold, while the difference between the auxiliary temperature sensor reading and the short-term temperature prediction value is less than the threshold, then the current data from the main temperature sensor is determined to be a spurious signal.

[0008] Preferably, the impact interference test specifically includes: S31, after each formal impact test, when the test chamber temperature has returned to room temperature and is in an idle window before the next test, controlling the test chamber to execute a standard, low-intensity temperature pulse excitation in an unloaded, specimen-free state; S32, during the execution of the temperature pulse excitation, simultaneously recording the excitation signal, the full-space temperature response, and the data of the associated system status at a sampling frequency higher than that of conventional control; S33, analyzing the full-space temperature response data, calculating a set of key characteristic parameters characterizing the thermal dynamic properties, and using the key characteristic parameters obtained each time... Define the current thermal dynamic feature fingerprint; S34, compare the thermal dynamic feature fingerprint with the baseline feature fingerprint established after the initial calibration or the last maintenance item by item, and generate a digital performance drift vector by calculating the percentage of relative deviation of each feature parameter relative to the baseline value; S35, establish a drift-parameter mapping relationship model, and automatically calculate the adjustment suggestions for multiple key adjustable parameters in the main control model based on the performance drift vector using the drift-parameter mapping relationship model; S36, perform data simulation based on the adjustment suggestions, and verify whether the main temperature sensor is a spurious signal by judging the simulation results.

[0009] Preferably, the key characteristic parameters specifically include: the time constant of the central control system, the ratio of steady-state temperature change to excitation, the time difference and temperature difference between sensors at different positions reaching their peak values, and the natural recovery decay rate.

[0010] Preferably, defining the key feature parameters obtained each time as the current thermal dynamic feature fingerprint includes: S331, when the test chamber is in baseline health, conducting an impact interference test to induce typical faults, recording its complete response under standard diagnostic excitation, and constructing a local physical feature signature library for various physical drift sources; S332, in each self-diagnosis, matching the high-dimensional features collected in real time with the local signature library, and matching it with the local physical feature signature library, calculating and outputting a quantified probability distribution vector characterizing the current performance degradation caused by each known physical drift source; S333, based on the calculated root cause probability distribution vector, accurately locating and adjusting the corresponding parameters in the main control model. For specific parameter subsets associated with the model, the adjustment scheme must first be validated in the local digital twin model before being silently updated to the online main control model. In S334, the locally validated diagnostic and compensation cases are packaged into a standard knowledge package and uploaded to the cloud knowledge hub. When the test box in the group detects signs of initial drift of high-dimensional feature vectors, a pre-diagnosis query can be initiated to the cloud knowledge hub, and the cloud recommends a matching, high-trust group knowledge scheme. In S335, after receiving the group knowledge scheme recommended by the cloud, the test box performs personalized pre-validation in its local digital twin model. After confirming that it is effective for its own hardware status, it is then applied to its own main control model.

[0011] Preferably, the physical feature signature library includes: extracting a set of high-dimensional feature vectors characterizing thermal dynamics: time-domain features, such as the time constant and steady-state gain of temperature rise / fall at each measuring point; frequency-domain features, such as the frequency and amplitude of major temperature fluctuations extracted by FFT transformation; and spatial features, such as the time difference and temperature difference matrix of sensors at different locations reaching peak temperatures. The extracted feature vectors are strongly correlated with the physical drift source corresponding to the experiment to form a unique physical feature signature for the drift source.

[0012] Preferably, the precise location and adjustment of the specific parameter subset associated with the main control model specifically involves: determining the physical drift source with the highest probability based on the root cause probability distribution vector; locating the specific parameter subset associated with the drift source based on the preset mapping relationship between the drift source and the control model parameters; calculating the required adjustment amount and direction for the parameter subset; loading the currently running main control model parameters into the local digital twin model and applying the calculated adjustment amount to the corresponding parameter subset for simulation verification; performing simulation verification in the digital twin model to analyze the expected control performance of the adjusted model; and updating the adjusted specific parameter subset to the running main control model if the simulation verification results meet the preset performance acceptance conditions.

[0013] Preferably, the cloud-based knowledge hub performs the following operations: receiving and clustering the unknown drift pattern characteristics reported by each test chamber; when a specific pattern report reaches a group threshold, it is determined to be a new type of group drift challenge; defining and digitally representing a high-dimensional strategy search space for the challenge, dividing it into multiple sub-regions and assigning IDs using an intelligent segmentation algorithm; matching the current unknown pattern characteristics with known strategy clusters in the knowledge base; if the match is successful, directly associating and recommending the corresponding verified strategy subspace or historical solution as a temporary knowledge solution to the unknown challenge.

[0014] Preferably, the special strategy for coupled shock transients specifically includes: locking the output state of the main control actuators before the predicted time of the shock loading arrives, keeping it at its current value; suspending the model-based active adjustment function of the control loop during the duration of the shock disturbance, maintaining the output of each actuator unchanged; after the shock event ends, reusing the reliable temperature value confirmed by signal arbitration as the control feedback, and using a smooth recovery function to safely and gradually return the control output and temperature setpoint to normal operating conditions; and calculating and generating a unified control command by combining the current control mode, the shock state, and the reliable temperature value.

[0015] One or more technical solutions provided in this application have at least the following technical effects or advantages: By integrating vibration sensing and temperature arbitration under impact loads, this method effectively identifies and shields sensor spurious signals caused by stress waves, ensuring the reliability of core temperature feedback. Through a dynamic coupling control strategy based on the calculated cold energy health index, the method adaptively adjusts the control objective when liquid nitrogen resources are limited, achieving a balance between accuracy and safety under extreme disturbances and resource constraints. This significantly improves the temperature control stability and experimental process controllability of cryogenic impact tests.

[0016] By automatically performing standardized temperature pulse excitation and response analysis during test intervals, the long-term drift of the test chamber's thermal dynamic characteristics can be quantitatively assessed, and a performance drift vector can be generated. Based on this vector, the system can automatically and accurately adjust the core parameters of the main control model, achieving online adaptive calibration and optimization of the control strategy. This method effectively compensates for the negative impacts caused by equipment aging, frosting, or sensor performance degradation, fundamentally improving the long-term robustness, accuracy consistency, and reliability of the control system. It also transforms the maintenance mode from reactive remediation to state-based predictive maintenance, significantly reducing the risk of test failure due to slow performance degradation.

[0017] By constructing a physical feature signature library of drift sources and performing online matching, this method achieves decoupled diagnosis of the physical root causes of performance degradation, moving from overall performance monitoring to specific fault location. It can target and accurately compensate a subset of key parameters of the main control model based on root cause probability, avoiding the risks and instability caused by global model adjustments. Combined with the collective intelligence of a cloud-based knowledge hub, effective experience from a single test chamber can be quickly and reliably extended to the entire test chamber cluster, achieving a leap from individual adaptation to collective collaborative evolution. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a method for precision control in a liquid nitrogen cryogenic test according to an embodiment of the present invention. Detailed Implementation

[0019] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the present invention.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the terminology used herein includes any and all combinations of one or more of the associated listed items.

[0021] Example 1: Figure 1 This is a flowchart illustrating a method for precision control in a liquid nitrogen cryogenic test according to an embodiment of the present invention.

[0022] like Figure 1 As shown, a method for precision control in liquid nitrogen cryogenic testing includes the following steps: S1. Install primary and secondary temperature sensors and a triaxial high-frequency accelerometer in the near-field region of the specimen, establish communication with the data acquisition system of the Hopkinson bar, and acquire stress wave initiation signals triggered by bullets or detected by strain gauges on the bar in real time.

[0023] Specifically, the specimen mounting base is selected as the near-field region. A co-located mounting point for the main temperature sensor and the triaxial high-frequency accelerometer is chosen on the specimen mounting base, ensuring tight mechanical contact between the two. Inside the test chamber (referring to the liquid nitrogen cryogenic Hopkinson bar impact test chamber), an auxiliary temperature sensor is placed at a location far from the impact force transmission path and with a relatively stable temperature, serving as a temperature reference during vibration interference. The signal lines of the main temperature sensor, auxiliary temperature sensor, and triaxial accelerometer are connected to the central data acquisition card to complete the hardware connection.

[0024] A communication link is established between the central control system (a control system for controlling the accuracy of liquid nitrogen cryogenic testing, including a control decision module and a data analysis module) and the Hopkinson bar independent data acquisition system via Ethernet or a hardware trigger line, and the corresponding data communication protocol is configured. In the central control system, the bullet firing trigger signal from the Hopkinson bar system or the stress wave initiation signal detected by the strain gauges on the incident bar is received and analyzed in real time. Based on the received initiation trigger signal, the precise stress wave initiation time point is extracted.

[0025] S2, based on the known wave velocity and propagation path length, predicts in advance the time window for the stress wave to reach the test chamber and temperature sensor; simultaneously reads the liquid level and pressure of the test chamber to calculate the liquid level change rate, and defines a cooling capacity health index in combination with the current air pressure and evaporator status.

[0026] Specifically, the known longitudinal wave velocity of the incident rod material of the Hopkinson rod device is obtained. Next, the axial straight-line distance from the bullet trigger point to the outer surface of the test chamber mounting base is measured; this distance is the stress wave propagation path length. Then, by calculating the propagation path length / longitudinal wave velocity, the basic time delay required for the stress wave to propagate from the trigger point to the specimen mounting base is obtained. Based on this basic time delay, 1 millisecond is subtracted forward (as a lead time) and 2 milliseconds are added backward (as a delay time), thus defining a possible time interval for the stress wave to arrive; this interval is the predicted time window for strong vibration disturbance.

[0027] In another parallel task, the liquid level signal value of the test chamber's liquid level gauge is read in real time via RS485 or analog input channels, and the pressure signal value of the test chamber's gas phase space is also read in real time. Differential calculations are performed on the continuously sampled liquid level signals to obtain the current liquid level change rate. Two liquid level thresholds are defined: a safe liquid level threshold and a critical liquid level threshold, as benchmarks for judging resource status. When the real-time liquid level is higher than the safe threshold, the cooling capacity health index is set to a maximum value of 1. When the real-time liquid level is between the critical and safe thresholds, a linear interpolation formula is used to calculate the cooling capacity health index. When the real-time liquid level is lower than the critical threshold, the index is calculated based on a composite function of liquid level and pressure, and the upper limit of the calculation result is limited to 0.5 to clearly indicate a state of severe resource shortage.

[0028] S3. In the impact interference test, the amplitude of the triaxial acceleration composite vector is calculated in real time. When the amplitude exceeds the preset vibration interference threshold or the current time is within the predicted stress wave arrival window, the data of the main temperature sensor is immediately marked as suspected interference, and an arbitration comparison group is established to determine whether the data of the main temperature sensor is a spurious signal.

[0029] Specifically, firstly, instantaneous signals in the X, Y, and Z directions from a triaxial accelerometer mounted on the main temperature sensor base are acquired in real time from a central data acquisition card. Next, for each sampling moment, the square root of the sum of the squares of the three axial acceleration components is calculated to obtain the magnitude of the composite acceleration vector at that moment. A preset vibration disturbance threshold parameter is then obtained; this threshold is typically set empirically to 0.5 times the gravitational acceleration.

[0030] Obtain the stress wave arrival time window parameters calculated in step S2, including the window start and end times. Obtain the current precise experimental progress time. Determine whether the current time is within the predicted stress wave arrival time window, or whether the current synthetic acceleration amplitude exceeds the preset vibration disturbance threshold. If either condition is met, add a suspected disturbance status marker to the current reading of the main temperature sensor in the data stream.

[0031] An arbitration comparison group is established, comprising: labeled primary temperature sensor readings, auxiliary temperature sensor readings, and short-term temperature predictions from an ARIMA time-series prediction model based on historical temperature data. The absolute difference between the primary and auxiliary temperature sensor readings is calculated and compared to a preset arbitration temperature deviation threshold (e.g., 5K). The absolute difference between the primary temperature sensor reading and the short-term prediction is calculated and compared to the same arbitration threshold. The absolute difference between the auxiliary temperature sensor reading and the short-term prediction is also calculated and compared to the same arbitration threshold. If the differences between the primary and auxiliary temperature sensor readings, and the short-term prediction, all exceed the arbitration threshold, while the difference between the latter two is less than the arbitration threshold, the current primary temperature sensor data is determined to be a spurious signal.

[0032] S4, during the period determined to be a spurious signal, cut off the input of the main temperature sensor data, and use the weighted average of the measured value and the short-term predicted value of the auxiliary temperature sensor as the virtual true value of the current test chamber temperature. When the vibration amplitude is less than 0.2g and the data is stable for three consecutive sampling cycles, smoothly switch back to the main temperature sensor.

[0033] Specifically, upon determining that the main temperature sensor data is a spurious signal, the control loop immediately interrupts the real-time data stream input from the main temperature sensor. It retrieves the latest measurement value from the auxiliary temperature sensor in the data buffer. It then obtains the predicted temperature value for the corresponding time moment from the ARIMA short-term prediction model. Based on preset arbitration weights (e.g., auxiliary temperature sensor weight 0.7, predicted value weight 0.3), the two values ​​are weighted and averaged to obtain the current virtual true temperature value. This virtual true value is input into the subsequent temperature control loop as feedback input to the control algorithm. During signal switching, the magnitude of the composite vector of triaxial acceleration is continuously monitored. It is then determined whether the magnitude of this composite vector has consistently fallen below a stable threshold (e.g., 0.2g).

[0034] After the vibration meets the stability condition, temperature data from the auxiliary temperature sensor is continuously acquired for the next three sampling periods, and its standard deviation is calculated. It is then determined whether this standard deviation is less than a preset data stationarity threshold. If both the vibration amplitude and the auxiliary temperature data meet the above stability condition, the main temperature sensor data recovery process is initiated. Using a first-order inertial filtering algorithm, within five preset sampling periods, the temperature input signal of the control loop is gradually and smoothly transitioned from the current virtual true value to new, valid data from the main temperature sensor that has been confirmed to have recovered and stabilized.

[0035] S5 dynamically selects the control mode based on the cooling capacity health index and couples it with a special strategy for shock transients. It generates closed-loop execution monitoring and event recording analysis through coordinated commands.

[0036] Specifically, the control decision module compares the current cooling capacity health index with preset high and low thresholds to determine the resource status. When the cooling capacity health index is higher than 0.7, a precision-priority control mode is activated, which uses high-gain PID parameters to pursue optimal control precision. When the cooling capacity health index is between 0.3 and 0.7, a resource protection control mode is activated. In this mode, control output commands (such as air pressure setpoints) are multiplied by the cooling capacity health index to reduce the cooling rate and moderately relax the allowable temperature fluctuation range. When the cooling capacity health index is lower than 0.3, the system switches to a safety maintenance mode, significantly reducing the evaporator heating power limit, prohibiting rapid cooling commands, and shifting the control objective to delaying temperature drift. One second before the predicted impact loading time, the control decision module issues a command to lock the output status of all major control actuators (such as solenoid valves and heaters). During the duration of the impact disturbance, the control loop suspends model-based active adjustment, and each actuator maintains its output before the lockout.

[0037] After the impact event ends, based on the signal arbitration result, a reliable temperature value (from the main temperature sensor or the virtual true value) is reselected as the feedback. A gentle ramp function is used to restore the control output from the current value to the normal operating state, and the temperature setpoint also approaches the target value again at a safe rate. Combining the current control mode, the impact state, and the reliable temperature value after arbitration, the control decision module calculates a unified target evaporator pressure and target valve opening.

[0038] The generated target instructions are sent to the underlying programmable logic controller for execution. The actual temperature change trend of the test chamber after instruction execution is monitored in real time and compared with the expected direction. The liquid level drop rate is monitored synchronously to confirm that it does not exceed the limits of the current control mode. All key events, state transitions, sensor data, and control instructions are recorded in a time-series database. Based on historical data analysis, the calculation model parameters for vibration interference threshold, arbitration temperature deviation threshold, and cooling capacity health index are periodically optimized.

[0039] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: By integrating vibration sensing and temperature arbitration under impact loads, this method effectively identifies and shields sensor spurious signals caused by stress waves, ensuring the reliability of core temperature feedback. Through a dynamic coupling control strategy based on the calculated cold energy health index, the method adaptively adjusts the control objective when liquid nitrogen resources are limited, achieving a balance between accuracy and safety under extreme disturbances and resource constraints. This significantly improves the temperature control stability and experimental process controllability of cryogenic impact tests.

[0040] Example 2: By arbitrating the temperature signal at the moment of impact and switching the control mode based on the cooling capacity health index, transient disturbances and resource constraints are addressed. However, this method relies on preset fixed control model parameters and does not consider the slow drift of the test chamber's thermal dynamic characteristics (such as response speed and uniformity) due to long-term use, component aging, or internal frosting. Faced with this gradual performance degradation, fixed model parameters will lead to a gradual decrease in control accuracy, and the performance inflection point cannot be predicted. To fundamentally maintain the stability of long-term control accuracy and achieve a transition from passive response to active maintenance, a method that can evaluate the system performance status online and adaptively correct the control model is needed.

[0041] In some embodiments, in the impact interference test, step S3 further includes: S31, after each formal impact test has ended and the test chamber temperature has returned to room temperature, and during the idle window before the next test, control the test chamber to execute a standard, low-intensity temperature pulse excitation in an unloaded state without test specimens.

[0042] Specifically, the central control system monitors and confirms that the current formal impact test procedure has been completely completed. Subsequently, the control unit stops all active cooling, allowing the test chamber to naturally warm up to room temperature until the readings of all temperature sensors inside the chamber stabilize within the preset room temperature range. The operator or automatic detection program must confirm that there are no test specimens inside the test chamber and that all doors and ports are in a safe, unloaded closed state. The control unit determines whether there is currently an available idle window, i.e., no other test tasks to be performed.

[0043] From the control logic, the preset standardized temperature pulse excitation program is invoked and prepared to start, closing all main cooling and rapid cooling valves to ensure the system is in a stable initial thermal state. The heater is controlled to output a fixed, low heating power for a preset short period of time. The heating power and duration are recorded synchronously, and the total excitation amount of this pulse input is calculated.

[0044] Immediately after the pulse excitation ends, the heater output is cut off, active heat input is stopped, and the test chamber's airtightness and ventilation conditions remain unchanged, allowing the test chamber to enter the natural cooling and recovery phase. Throughout the excitation and recovery phases, the temperature changes inside the chamber are continuously monitored and recorded to ensure they do not exceed the safety threshold. Once the temperature inside the chamber has naturally recovered to near the initial temperature before excitation and reached stability, the excitation process is considered complete.

[0045] S32, during the execution of temperature pulse excitation, simultaneously records data on the excitation signal, the temperature response of the entire space, and the status of the associated system at a sampling frequency higher than that of conventional control.

[0046] Specifically, at the instant the temperature pulse excitation begins, the sampling frequency of the data acquisition channels of all temperature sensors, accelerometers, flow meters, and pressure transmitters within the test chamber is switched to high-resolution mode. The start timestamp and duration of the excitation signal output are accurately recorded and stored to ensure the accuracy of the excitation time base.

[0047] The system synchronously records the applied heating power or valve opening signal at a high sampling rate as a known input excitation signal; records the temperature reading change curves of the main temperature sensor and all auxiliary temperature sensors; records the triaxial acceleration data at the mounting base of the main temperature sensor to monitor vibration during excitation; records the pressure change data of the test chamber to analyze the impact of thermal disturbance on the pressure inside the tank; and records the liquid level change data of the test chamber to calculate the rate of influence of micro-evaporation on the liquid level.

[0048] During the natural recovery process after the excitation stops, the test chamber status data continues to be recorded at a high sampling rate until the test chamber returns to steady state. All high-frequency acquisition data from all channels are aligned with a unified high-precision clock signal and timestamped.

[0049] The complete synchronized dataset is stored as a separate diagnostic data file, which contains excitation parameters, sampling rate, sensor identifiers, and raw data.

[0050] S33 analyzes the temperature response data of the entire space, calculates a set of key characteristic parameters characterizing thermal dynamics, and defines the key characteristic parameters obtained each time as the thermal dynamic characteristic fingerprint of the current moment.

[0051] Key characteristic parameters include: the time constant of the central control system (the time required for the temperature change to reach 63.2% of the steady-state change), the system steady-state gain (the ratio of the steady-state temperature change to the excitation), the temperature field uniformity index (the time difference and temperature difference between sensors at different locations reaching their peak values), and the natural recovery decay rate (the rate at which the temperature decreases (decays). The set of all characteristic parameters calculated in this diagnostic test is defined as the current thermal dynamic characteristic fingerprint.

[0052] S34 compares the thermal dynamic feature fingerprint with the baseline feature fingerprint established after the initial calibration or the last maintenance item by item, and generates a digital performance drift vector by calculating the percentage of the relative deviation of each feature parameter relative to the baseline value.

[0053] The baseline feature fingerprint is obtained by strictly following steps S31 to S33 during the initial standardized diagnostic process after the test chamber has passed factory commissioning or after major maintenance such as sensor recalibration. The feature fingerprint calculated during the initial diagnosis is used as the sole reference standard for all subsequent comparisons and drift analyses; this is the baseline feature fingerprint.

[0054] S35. Establish a drift-parameter mapping relationship model. Based on the performance drift vector, use the drift-parameter mapping relationship model to automatically calculate the suggested adjustment values ​​for multiple key adjustable parameters in the main control model.

[0055] The control unit incorporates a pre-trained drift-parameter mapping model, which describes the quantitative correlation between the performance drift vector and key adjustable parameters of the main control model described in Example 1 (such as the PID parameters of the temperature correction model and the weighting coefficients of the cooling capacity health index calculation model). Based on the current performance drift vector calculated in step S34, the data analysis module uses this mapping model to automatically calculate suggested adjustment values ​​for one or more key adjustable parameters in the aforementioned main control model.

[0056] S36. Perform data simulation based on the recommended adjustment values, and verify whether the main temperature sensor is a spurious signal by analyzing the simulation results.

[0057] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: By automatically performing standardized temperature pulse excitation and response analysis during test intervals, the long-term drift of the test chamber's thermal dynamic characteristics can be quantitatively assessed, and a performance drift vector can be generated. Based on this vector, the system can automatically and accurately adjust the core parameters of the main control model, achieving online adaptive calibration and optimization of the control strategy. This method effectively compensates for the negative impacts caused by equipment aging, frosting, or sensor performance degradation, fundamentally improving the long-term robustness, accuracy consistency, and reliability of the control system. It also transforms the maintenance mode from reactive remediation to state-based predictive maintenance, significantly reducing the risk of test failure due to slow performance degradation.

[0058] Example 3: In Example 2, quantitative compensation for overall performance drift was achieved by quantitatively evaluating the long-term drift of the test chamber's thermal dynamic characteristics and adjusting model parameters online. However, this method treats system performance degradation as a single overall phenomenon. While the generated performance drift vector quantifies the magnitude of degradation, it fails to reveal the specific physical root causes (such as sensor loosening, flow channel blockage, or heater aging). Faced with degradation caused by different physical reasons but potentially exhibiting similar overall performance indicators, a unified model parameter adjustment strategy inevitably limits the accuracy and long-term adaptability of compensation. Since the impact mechanisms of different physical drift sources on different parameter subsets in the main control model are fundamentally different, a method is needed to decouple the drift root causes and perform precise intervention based on physical characteristics and collective knowledge in order to achieve root cause diagnosis and targeted compensation of performance degradation, and to utilize collective experience for preventative optimization.

[0059] In some embodiments, step S33, which defines the key feature parameters obtained each time as the current thermal dynamic feature fingerprint, further includes: S331, when the test chamber is in baseline health, performs impact interference tests to induce typical faults, records its complete response under standard diagnostic excitation, and builds a local physical feature signature library for various physical drift sources.

[0060] The physical signature library includes a set of high-dimensional feature vectors characterizing thermal dynamics: time-domain features, including the time constants of temperature rise / fall and steady-state gain at each measuring point; frequency-domain features, extracted using FFT transformation to determine the main temperature fluctuation frequencies and amplitudes; and spatial features, including the time difference and temperature difference matrix between sensors at different locations when they reach peak temperatures. The extracted feature vectors are strongly correlated with the physical drift source corresponding to the experiment (e.g., sensor loosening) to form a unique physical signature for that drift source, which is then stored in the local database.

[0061] S332, in each self-diagnosis, matches the high-dimensional features collected in real time with the local signature library and with the local physical feature signature library, calculates and outputs a quantified probability distribution vector characterizing the current performance degradation caused by each known physical drift source.

[0062] Specifically, when the test chamber is unloaded, the temperature has returned to room temperature, and it is within an idle window, self-diagnosis is automatically triggered. Standardized low-intensity temperature pulse excitation is executed. At the same time, the sampling frequency of all temperature, acceleration, pressure, and flow sensors in the test chamber is switched to high-resolution mode (same as S32). Using the collected complete response data, the high-dimensional feature vector of the current test chamber is calculated strictly according to S331. The current high-dimensional feature vector is compared with each benchmark signature in the local physical feature signature library in a multi-dimensional similarity measure. The weighted Euclidean distance or cosine similarity algorithm is used to calculate the difference between the current feature and each benchmark feature. The calculated distance or similarity value is normalized to the interval [0, 1].

[0063] Based on the similarity calculation results, a probability value is assigned to each known physical drift source in the library (such as poor sensor contact or flow channel blockage). A higher probability indicates a closer match between the current test chamber state and the typical characteristics of that drift source. Finally, a structured root cause probability distribution vector is output, which is used as the direct input to step S333. S333, based on the calculated root cause probability distribution vector, accurately locates and adjusts a specific subset of parameters associated with it in the main control model. The adjustment scheme needs to be validated in the local digital twin model first, and then silently updated to the online main control model.

[0064] Specifically, the root cause probability distribution vector output from step S332 is read and parsed. In a pre-defined root cause-parameter mapping table, the subset of main control model parameters (parameters strongly correlated with a specific physical fault) corresponding to the drift source with the highest probability is queried. Based on the mapping relationship, the specific adjustment amount and direction for this parameter subset are calculated.

[0065] A snapshot of the parameters of the currently running master control model is loaded into the digital twin model, and the calculated parameter adjustments are applied to the corresponding subset of parameters in the digital twin model. Standardized temperature pulse excitation and typical shock test conditions are simulated within the digital twin environment. The simulation results are analyzed to quantitatively evaluate the expected improvement in control accuracy and stability changes after parameter adjustments.

[0066] Determine if the simulation results meet the preset performance improvement acceptance threshold. If the verification passes, generate a parameter update instruction and write the adjusted parameter subset into the online main control model. Perform a hot update of the model parameters to ensure seamless switching of the control process without interrupting any pending test tasks. Record a complete log of this parameter adjustment, including the root cause vector, adjusted parameters, simulation results, and update timestamp. If the verification fails, abandon the adjustment, retain the original model parameters, and generate a fault analysis report for subsequent processing.

[0067] S334 encapsulates locally validated diagnostic and compensation cases into standard knowledge packages and uploads them to the cloud knowledge hub. When it detects signs of initial drift of high-dimensional feature vectors in the test chamber within the group, it can initiate a pre-diagnosis query to the cloud knowledge hub, and the cloud recommends matching, high-trust group knowledge solutions to it.

[0068] Among these, the cloud recommends matching, high-trust group knowledge solutions, specifically including: The cloud-based knowledge hub continuously performs cluster analysis on all reported high-dimensional feature vectors and their verified effective solutions to form and maintain a structured policy knowledge base, where each cluster represents a known drift pattern and its corresponding policy search space.

[0069] When a single test chamber identifies an unknown drift pattern in its self-diagnosis (the matching degree with all known signatures in the local feature library is lower than the set threshold, such as <0.3), it will encrypt and package the high-dimensional feature vector of the pattern, the device identifier, and the operating condition snapshot, and automatically upload it as an unknown problem report.

[0070] The cloud-based central hub continuously receives these reports and uses clustering algorithms to analyze their feature vectors. When the number of reports with similar features reaches a group threshold within a preset time window, the cloud identifies it as a novel, widespread group drift challenge.

[0071] First, based on domain knowledge, the cloud platform defines an initial high-dimensional policy search space for this type of problem and digitally represents it. Then, using intelligent segmentation algorithms (such as initial segmentation based on Latin hypercube sampling or heuristic partitioning based on the distribution of historical successful policies), the entire policy space is divided into multiple discrete sub-exploration regions, each of which is assigned a unique region ID.

[0072] The cloud-based system attempts to find a quick solution for the feature vector of this unknown problem: it matches this feature vector with each cluster in the policy knowledge base (i.e., each known policy search space). If an existing cluster with a high matching degree is found, the system directly recommends the validated policy subspace or historical solutions associated with that cluster as temporary knowledge solutions to the current unknown problem.

[0073] If an unknown problem after splitting can correspond to a region ID with a high degree of matching, then it is included in the temporary knowledge scheme until all the location problems after splitting have found corresponding region IDs. The temporary knowledge scheme is then integrated as a special knowledge scheme for the current unknown drift mode.

[0074] After receiving the group knowledge solution recommended by the cloud, the S335 test chamber performs personalized pre-verification in its local digital twin model to confirm its effectiveness on its own hardware status before applying it to its main control model.

[0075] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: By constructing a physical feature signature library of drift sources and performing online matching, this method achieves decoupled diagnosis of the physical root causes of performance degradation, moving from overall performance monitoring to specific fault location. It can target and accurately compensate a subset of key parameters of the main control model based on root cause probability, avoiding the risks and instability caused by global model adjustments. Combined with the collective intelligence of a cloud-based knowledge hub, effective experience from a single test chamber can be quickly and reliably extended to the entire test chamber cluster, achieving a leap from individual adaptation to collective collaborative evolution.

[0076] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A precision control method of liquid nitrogen ultra-low temperature test, characterized in that, include: S1. Install main and auxiliary temperature sensors and a triaxial high-frequency accelerometer in the near-field region of the specimen, establish communication with the data acquisition system of the Hopkinson bar, and acquire the stress wave initiation signal detected by the bullet-triggered Hopkinson bar or the strain gauge on the bar in real time. S2, based on the known wave velocity and propagation path length, predicts in advance the time window for the stress wave to reach the test chamber and temperature sensor; simultaneously reads the liquid level and pressure of the test chamber to calculate the liquid level change rate, and defines a cooling capacity health index in combination with the current air pressure and evaporator status; S3, In the impact interference test, the amplitude of the triaxial acceleration composite vector is calculated in real time. When the amplitude exceeds the preset vibration interference threshold or the current time is within the predicted stress wave arrival window, the data of the main temperature sensor is immediately marked as suspected interference, and an arbitration comparison group is established to determine whether the data of the main temperature sensor is a false signal. S4, during the period when it is determined to be a false signal, the input of the main temperature sensor data is cut off. The weighted average of the measured value and the short-term predicted value of the auxiliary temperature sensor is used as the virtual true value of the current test chamber temperature. When the vibration amplitude is less than 0.2g and the data is stable for three consecutive sampling cycles, the main temperature sensor is smoothly switched back. S5 dynamically selects the control mode based on the cooling capacity health index and couples it with a special strategy for shock transients. It generates closed-loop execution monitoring and event recording analysis through coordinated commands.

2. The precision control method for liquid nitrogen cryogenic testing according to claim 1, characterized in that, The cooling capacity health index specifically includes: real-time reading of the liquid nitrogen storage tank's liquid level and pressure signals via a communication interface; continuous sampling and differential calculation of the liquid level signal to obtain the real-time liquid level change rate; preset safe liquid level thresholds and critical liquid level thresholds; and dynamically determining the index value using different calculation rules based on the real-time liquid level relative to the aforementioned thresholds: when the liquid level is higher than the safe threshold, the index is set to a fixed value representing abundant resources; when the liquid level is between the critical threshold and the safe threshold, the index is calculated using a linear interpolation method; when the liquid level is lower than the critical threshold, the index is calculated based on a composite function of liquid level, pressure, and liquid level change rate, and its upper limit is limited to represent a severely constrained resource state.

3. The precision control method for liquid nitrogen cryogenic testing according to claim 1, characterized in that, The process of establishing an arbitration comparison group to determine whether the data from the main temperature sensor is a spurious signal includes: simultaneously acquiring the readings of the main temperature sensor (marked as potentially affected by interference), the readings of the auxiliary temperature sensor, and a short-term temperature prediction value based on historical temperature data to form an arbitration comparison group; comparing the main temperature sensor reading with the auxiliary temperature sensor reading and the short-term temperature prediction value, respectively, and comparing them with a preset temperature deviation threshold; simultaneously comparing the auxiliary temperature sensor reading with the short-term temperature prediction value, and also comparing them with the same preset temperature deviation threshold; if the difference between the main temperature sensor reading and the other two exceeds the threshold, while the difference between the auxiliary temperature sensor reading and the short-term temperature prediction value is less than the threshold, then the current data from the main temperature sensor is determined to be a spurious signal.

4. The precision control method for liquid nitrogen cryogenic testing according to claim 1, characterized in that, The impact interference test specifically includes: S31, after each formal impact test, when the test chamber temperature has returned to room temperature and is in an idle window before the next test, controlling the test chamber to execute a standard, low-intensity temperature pulse excitation in an unloaded, specimen-free state; S32, during the execution of the temperature pulse excitation, simultaneously recording the excitation signal, the full-space temperature response, and the data of the associated system status at a sampling frequency higher than that of conventional control; S33, analyzing the full-space temperature response data, calculating a set of key characteristic parameters characterizing thermal dynamics, and defining the key characteristic parameters obtained each time. S34. The thermal dynamic feature fingerprint at the current moment is obtained; S35. The thermal dynamic feature fingerprint is compared item by item with the baseline feature fingerprint established after the initial calibration or the last maintenance. By calculating the percentage of the relative deviation of each feature parameter relative to the baseline value, a digital performance drift vector is generated; S36. A drift-parameter mapping relationship model is established. Based on the performance drift vector, the drift-parameter mapping relationship model is used to automatically calculate the adjustment suggestions for multiple key adjustable parameters in the main control model; S37. Data simulation is performed based on the adjustment suggestions. The simulation results are used to verify whether the main temperature sensor is a spurious signal.

5. The precision control method for liquid nitrogen cryogenic testing according to claim 4, characterized in that, The key characteristic parameters specifically include: the time constant of the central control system, the ratio of steady-state temperature change to excitation, the time difference and temperature difference between sensors at different positions reaching their peak values, and the natural recovery decay rate.

6. The precision control method for liquid nitrogen cryogenic testing according to claim 4, characterized in that, The step of defining the key feature parameters obtained each time as the current thermal dynamic feature fingerprint includes: S331, when the test chamber is in baseline health, conducting impact interference tests to induce typical faults, recording its complete response under standard diagnostic excitation, and constructing a local physical feature signature library for various physical drift sources; S332, in each self-diagnosis, matching the high-dimensional features collected in real time with the local signature library, and matching it with the local physical feature signature library, calculating and outputting a quantified probability distribution vector characterizing the current performance degradation caused by each known physical drift source; S333, based on the calculated root cause probability distribution vector, accurately locating and adjusting the associated parameters in the main control model. For a specific subset of parameters, the adjustment scheme must first be validated in the local digital twin model before being silently updated to the online main control model. In S334, successfully validated diagnostic and compensation cases are packaged into a standard knowledge package and uploaded to the cloud knowledge hub. When signs of initial high-dimensional feature vector drift are detected in the test chambers within the group, a pre-diagnosis query can be initiated to the cloud knowledge hub, which will recommend a matching, highly trusted group knowledge scheme. In S335, after receiving the group knowledge scheme recommended by the cloud, the test chamber performs personalized pre-validation in its local digital twin model. After confirming its effectiveness for its own hardware status, it applies it to its own main control model.

7. The precision control method for liquid nitrogen cryogenic testing according to claim 6, characterized in that, The physical signature library includes: extracting a set of high-dimensional feature vectors characterizing thermal dynamics: time-domain features, such as the time constant and steady-state gain of temperature rise / fall at each measuring point; frequency-domain features, such as the frequency and amplitude of major temperature fluctuations extracted by FFT transformation; and spatial features, such as the time difference and temperature difference matrix of sensors at different locations reaching peak temperatures. The extracted feature vectors are strongly correlated with the physical drift source corresponding to the experiment to form a unique physical signature of the drift source.

8. The precision control method for liquid nitrogen cryogenic testing according to claim 6, characterized in that, The precise location and adjustment of the specific parameter subset associated with the main control model specifically involves: determining the physical drift source with the highest probability based on the root cause probability distribution vector; locating the specific parameter subset associated with the drift source based on the preset mapping relationship between the drift source and the control model parameters; calculating the required adjustment amount and direction for the parameter subset; loading the currently running main control model parameters into the local digital twin model and applying the calculated adjustment amount to the corresponding parameter subset for simulation verification; performing simulation verification in the digital twin model and analyzing the expected control performance of the adjusted model; and updating the adjusted specific parameter subset to the online main control model if the simulation verification results meet the preset performance acceptance conditions.

9. The precision control method for liquid nitrogen cryogenic testing according to claim 6, characterized in that, The cloud-based knowledge hub performs the following operations: receives and clusters the features of unknown drift patterns reported by each test chamber; when a specific pattern report reaches a group threshold, it is determined to be a new type of group drift challenge; defines and digitally represents a high-dimensional strategy search space for the challenge, divides it into multiple sub-regions using an intelligent segmentation algorithm and assigns IDs; matches the current unknown pattern features with known strategy clusters in the knowledge base; if the match is successful, it directly associates and recommends the corresponding verified strategy subspace or historical solution as a temporary knowledge solution to the unknown challenge.

10. The precision control method for liquid nitrogen cryogenic testing according to claim 1, characterized in that, The specific strategies for coupled shock transients include: locking the output state of the main control actuators before the predicted time of the shock loading, keeping it at its current value; suspending the model-based active adjustment function of the control loop during the duration of the shock disturbance, maintaining the output of each actuator unchanged; after the shock event ends, re-selecting the reliable temperature value confirmed by signal arbitration as the control feedback, and using a smooth recovery function to safely and gradually return the control output and temperature setpoint to normal operating conditions; and calculating and generating a unified control command by combining the current control mode, the shock state, and the reliable temperature value.

Images