Superconducting magnet coil quenching detection method and related device

Abstract

The invention discloses a superconducting magnet coil quenching detection method and a related device, and relates to the technical field of superconducting magnets, and the method comprises the steps: collecting the total voltage of two ends of a superconducting magnet coil and the current of any end of the superconducting magnet coil in real time in the operation process of a superconducting magnet; based on a relation curve that the inductance of the superconducting magnet coil changes along with the current, the resistive voltage in the superconducting magnet coil is determined through the total voltage at the two ends of the superconducting magnet coil and the current at any end of the superconducting magnet coil; when it is detected that the resistive voltage is increased to a voltage threshold value or the continuous increasing duration of the resistive voltage exceeds a duration threshold value, it is judged that the superconducting magnet coil is quenched; through the method, the problems that an existing superconducting magnet quenching detection method is complex and low in detection sensitivity are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of superconducting magnets, in particular to a superconducting magnet coil quench detection method and related device. BACKGROUND

[0002] Superconducting magnets are usually operated at ultra-low temperature, and the coils are in a superconducting state. The process of superconducting material in the superconducting coil changing into a normal conductor is called quench. The magnetic energy released during the quench process will cause the local temperature of the magnet to rise rapidly. If the temperature rises too high, it will damage the structure inside the superconductor and even burn the magnet. Therefore, in the process of designing and manufacturing superconducting magnets, quench should be prevented as much as possible, and the quench protection device or measures should be started in time when quench occurs.

[0003] A superconducting magnet is generally composed of one or more coils. During the quench process, the physical quantities of the coil (such as superconductor temperature, voltage, power, current, stress and strain, etc.) will change. Based on these physical quantities, existing quench detection methods include voltage difference detection method, distributed optical fiber temperature measurement method, stray capacitance detection method, active power method, ultrasonic detection method, etc. Among them, the voltage difference detection method as a widely used electrical detection method has obvious limitations: on the one hand, the small resistive voltage signal generated during quench is easily submerged by the large inductive and capacitive voltage noise of the magnet, making it difficult to achieve high sensitivity detection; on the other hand, the superconducting magnet must be designed as multiple coils with similar structures, and the voltage signals of each coil must be detected at the same time to compare the weak resistive voltage difference between different coils, which requires strict consistency of the coil structure and inductance, limiting the flexibility of the magnet design. SUMMARY

[0004] The purpose of the present application is to provide a superconducting magnet coil quench detection method and related device, aiming to solve the problem of complex existing superconducting magnet quench detection method.

[0005] The purpose of the present application is achieved by adopting the following technical solutions:

[0006] In a first aspect, the present application provides a superconducting magnet coil quench detection method, which comprises:

[0007] During the operation of the superconducting magnet, the total voltage V(t) across the superconducting magnet coil and the current I(t) at any one end of the superconducting magnet coil are collected in real time;

[0008] Based on the relationship curve L(I(t)) of the inductance of the superconducting magnet coil changing with the current, the resistive voltage V IR (t) in the superconducting magnet coil is determined based on the total voltage across the superconducting magnet coil and the current at any one end of the superconducting magnet coil.

[0009] When the resistive voltage increases to a voltage threshold, and / or the duration of the continuous increase of the resistive voltage exceeds a time threshold, it is determined that a quench occurs in the superconducting magnet coil.

[0010] Preferably, the relationship curve between the superconducting magnet coil inductance and current is obtained by the following method, including:

[0011] Establish a curve showing the relationship between the inductance of a superconducting magnet coil and the change in current through simulation; or,

[0012] When the superconducting magnet coil is in a superconducting state, during the current dynamic adjustment process, the inductance values ​​corresponding to different current values ​​are synchronously measured to establish and correct the relationship curve between the superconducting magnet coil inductance and current change;

[0013] During the operation of the superconducting magnet, the established curve of the superconducting magnet coil inductance variation with current is used to predict the inductance variation with current of a larger superconducting magnet in a larger current range.

[0014] Preferably, the method of determining the resistive voltage in the superconducting magnet coil according to a curve showing a relationship between the inductance of the superconducting magnet coil and the current, by the total voltage across the superconducting magnet coil and the current at any one end of the superconducting magnet coil, comprises:

[0015] Based on the curve of the superconducting magnet coil inductance versus current and the real-time current acquisition, the inductance value L(t) corresponding to the current I(t) and the derivative of the inductance with respect to the current dL / dI are determined.

[0016] According to the current inductance value L(t), the current current change rate and the current current I(t), the ideal inductive voltage V under the ideal superconducting state is calculated. ideal (t);

[0017]

[0018] in, is the current inductance change rate, which can be approximated to zero for magnets whose inductance changes slowly with time; is the current change rate;

[0019] By comparing the total voltage V(t) with the ideal inductive voltage V ideal (t), determine the resistive voltage V in the superconducting magnet coil IR (t);

[0020] V IR (t)=V(t)-V ideal (t)

[0021] Preferably, the determination logic for the resistive voltage increasing to a voltage threshold includes:

[0022] When the resistive voltage exceeds its starting voltage threshold for the first time, the continuous growth timer starts, and a quench is determined if any of the following conditions is met:

[0023] The resistive voltage exceeds the final voltage threshold before reaching the duration threshold;

[0024] When the resistive voltage reaches the duration threshold, it exceeds the final voltage threshold;

[0025] The resistive voltage continuously increases for a period exceeding a time threshold;

[0026] Among them, the starting voltage threshold is set based on the maximum noise peak value of normal operation; the final voltage threshold is set based on the exponential growth characteristics of the resistive voltage, and the final voltage threshold is determined by calculating the maximum temperature of the hot spot inside the magnet coil when the voltage is reached through a quench simulation experiment.

[0027] Preferably, the resistive voltage includes an original resistive voltage and a filtered resistive voltage, and the decision logic is independently executed on the original resistive voltage and the filtered resistive voltage at the same time.

[0028] Preferably, the voltage threshold is an adaptive threshold that is dynamically modified based on real-time operating conditions, including:

[0029] During the operation of the superconducting magnet, the ambient temperature of the coil, the pressure of the cooling medium, and the noise component of the total voltage across the coil are collected in real time. The noise component is extracted by high-frequency filtering of the total voltage signal and represented by the standard deviation of the voltage signal within a sliding window.

[0030] Dynamically adjusting the voltage threshold based on a preset correction coefficient matrix and a noise component; the correction coefficient matrix includes a temperature correction term, a pressure correction term, and a noise correction term;

[0031] When it is detected that the relationship curve has been calibrated online, the voltage threshold is corrected synchronously according to the calibration ratio.

[0032] Preferably, the correction coefficient matrix is ​​dynamically updated in the following manner:

[0033] Collect non-quench abnormal data and corresponding voltage noise data of superconducting magnet coils under different operating conditions, and construct an abnormal sample set with noise characteristics;

[0034] The temperature correction coefficient, pressure correction coefficient and noise correction coefficient are iteratively optimized using the gradient descent algorithm, so that the misjudgment rate of the voltage threshold in the abnormal sample set is lower than the set value, and the detection delay of the small resistive voltage in the quench simulation experiment does not exceed the preset threshold.

[0035] Preferably, the duration threshold is preset as a dynamic threshold associated with the resistive voltage growth slope and the voltage noise level, including:

[0036] Calculate the real-time growth slope of the resistive voltage, which is the ratio of the difference between the resistive voltage values ​​at adjacent moments to the time interval;

[0037] Real-time calculation of voltage noise fluctuation coefficient, which is the ratio of the current noise component to the baseline noise value under rated operating conditions;

[0038] Presetting multiple sets of slope thresholds, first reference times, and second reference times for different magnets or different operating power ranges of the same magnet;

[0039] When the real-time growth slope is not less than the slope threshold, the duration threshold is set to the product of the first reference time and the noise fluctuation coefficient; when the real-time growth slope is less than the threshold, the duration threshold is set to the product of the second reference time and the noise fluctuation coefficient, and the first reference time is less than the second reference time, and the noise fluctuation coefficient is ≥1;

[0040] The slope threshold, the first reference time, and the second reference time are determined through a quench evolution simulation experiment.

[0041] Optionally, the method further includes:

[0042] Determine the equivalent resistance of the current superconducting magnet coil by using the resistive voltage and the current at either end of the current superconducting magnet coil;

[0043] If the equivalent resistance exceeds its equivalent resistance threshold, or the equivalent resistance continuously increases for a period exceeding a time threshold and its integral value exceeds the energy integral upper limit corresponding to the safe hotspot temperature, it is determined that the superconducting magnet coil has quenched; the integral upper limit is determined based on the enthalpy characteristics of the superconducting coil material and the energy absorption upper limit at a safe temperature.

[0044] Preferably, the equivalent resistance threshold includes a starting resistance threshold and a final resistance threshold; the duration threshold is a third duration threshold, and a quench is determined when any of the following conditions is met:

[0045] After the equivalent resistance exceeds the starting resistance threshold, the continuous increase timing is started, and the final resistance threshold is exceeded before the third time threshold is reached; or,

[0046] The equivalent resistance exceeds the final resistance threshold when reaching the third time threshold;

[0047] Among them, the starting resistance threshold is set according to the resistance value corresponding to the maximum noise during normal operation, and the starting resistance threshold is not less than the critical resistance corresponding to the minimum propagation area; the final resistance threshold is set according to the resistance value corresponding to the safe hotspot temperature in the quench simulation experiment, and is calibrated according to the resistivity change characteristics of the superconducting material in the normal state with temperature and the coil structure parameters; the third time threshold is determined according to the equivalent resistance growth slope and the noise level.

[0048] In a second aspect, the present invention provides a superconducting magnet coil quench detection device, the device comprising:

[0049] One or more current sensors for measuring the current at any end of the superconducting magnet coil in real time;

[0050] One or more voltage sensors for measuring the total voltage across the superconducting coil in real time;

[0051] A processing unit is configured to execute the steps of the superconducting magnet coil quench detection method.

[0052] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any method described in the present invention or the functions of the device described in the present invention when executing the computer program.

[0053] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the storage medium stores computer instructions. When a computer reads the computer instructions, the computer executes the steps of any one of the methods described in the present invention.

[0054] Compared with the prior art, the present invention has at least the following beneficial effects: by precisely separating the resistive voltage (eliminating the inductive voltage interference caused by the nonlinearity of the inductance) and combining it with an adaptive threshold, it can effectively identify tiny resistive voltage signals, solving the problem of their being easily submerged by noise; at the same time, there are no special requirements for the structure of the superconducting magnet, in particular, it is not necessary for the multiple coil structures and inductance inside the superconducting magnet to be approximately consistent, and there is no need to compare the voltage signals of the multiple coils inside the superconducting magnet; only the voltage of the entire superconducting magnet needs to be measured. The present invention has no special requirements for the winding process of the superconducting magnet coil, and does not require various sensing structures such as optical fibers or temperature sensors to be pre-embedded during coil winding. It is not even necessary to pre-embed voltage measurement leads on the coils inside the magnet, nor is it necessary to establish an equivalent superconducting magnet outside the superconducting magnet for auxiliary measurement or calculation; quench detection can be achieved by simply measuring the voltage of the entire superconducting magnet, which simplifies the detection method and improves the detection sensitivity; through online calibration of the inductance curve and dynamic correction of the threshold, it is applicable to superconducting magnets under different operating states, improving the reliability of detection under complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a simplified schematic diagram of a magnetic circuit structure of an embodiment of the present invention;

[0056] Figure 2 1 is a schematic diagram of the resistive voltage of a flux jump signal according to an embodiment of the present invention;

[0057] Figure 3 Schematic diagram of a superconducting magnet coil quench detection method according to an embodiment of the present invention;

[0058] Figure 4 Schematic diagram of a magnet quench signal according to an embodiment of the present invention. DETAILED DESCRIPTION

[0059] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concepts of the example embodiments to those skilled in the art. Identical reference numerals in the drawings represent identical or similar structures, and thus repeated descriptions thereof will be omitted.

[0060] The words expressing positions and directions described in the present invention are all explained with reference to the accompanying drawings as examples, but can be modified as needed, and all such modifications are within the scope of protection of the present invention.

[0061] Reference Figure 1 : Attached Figure 1 To simplify the magnet circuit structure, there are two coils inside the magnet, with corresponding inductances L1 and L2 respectively;

[0062] The total voltage V across the magnet should satisfy the following formula:

[0063]

[0064] In the ideal superconducting state, the resistive voltage IR in the coil should satisfy:

[0065]

[0066] It is an inductive voltage (generated by the change of coil magnetic flux, reflecting the electromagnetic coupling characteristics of the superconducting magnet) generated by the coil resistance. In the superconducting state, the ideal resistance R = 0, so the resistive voltage IR is theoretically 0.

[0067] However, when a quench occurs, the resistance of the superconducting material suddenly changes, and the resistive voltage begins to rise from zero. In existing technologies, voltage detection methods have become the mainstream because they directly utilize the resistive voltage characteristics. However, they are limited by inductive voltage noise interference (large inductance causes the inductive voltage to be much larger than the weak resistive voltage) and multi-coil coupling errors (dynamic changes in coil self-inductance and mutual inductance affect the inductive voltage calculation). Traditional methods have difficulty accurately separating the resistive voltage, resulting in low quench detection sensitivity and a high misjudgment rate.

[0068] Based on the above principles, the present invention addresses the core pain points of superconducting magnet quench detection and proposes a resistive voltage separation method that combines inductance-current characteristics to achieve accurate quench determination.

[0069] The present invention focuses on solving the problems of complex existing methods and low detection sensitivity. It proposes to determine the quench by measuring or calculating the curve of the superconducting magnet's inductance changing with current, combining the voltage and current at both ends of the coil during operation, and calculating the resistive voltage (the resistive voltage in the ideal superconducting state is zero, and the resistive voltage will increase abnormally during quench). However, the implementation of this invention faces multiple technical difficulties: First, the experimental conditions are harsh, the superconducting magnet structure is complex and the cost is high, the number of high-field superconducting magnets in the world is scarce, the equipment and opportunities for conducting quench detection experiments are scarce, and the experiments are expensive and risky; second, the sensor accuracy requirements are extremely high. Superconducting magnets usually operate under high current and low voltage (close to zero voltage) conditions, relying on high-precision and wide-range current and voltage sensors; third, signal separation is difficult, the quench signal is weak, and high current operation relies on high-power power supply, which is easy to introduce various types of noise. How to separate the effective quench signal from the complex noise becomes a key challenge.

[0070] In terms of signal processing and quench determination, the existing noise interference problem is prominent: general noise can be reduced by high-precision sensors and shielding and vibration reduction measures, but it is difficult to completely eliminate it; AC noise (such as 50Hz power system frequency) suppression requires balancing the signal acquisition cycle and detection speed, and high-speed and high-precision collectors are often combined with effective value calculation and filtering noise reduction processing; in addition, the flux jump signals generated by superconducting magnet excitation, demagnetization and mechanical disturbances are easily confused with the real quench signal. Figure 2 , requiring a targeted detection method—the flux jump signal, after rising, then falls back due to energy release / transfer, allowing for differentiation from a true quench. These noise and interference factors further complicate accurate quench detection, highlighting the necessity and value of this invention in breaking through detection bottlenecks and ensuring the safe operation of superconducting magnets.

[0071] Refer to the attached Figure 3 , a superconducting magnet coil quench detection method, the method comprising:

[0072] During the operation of the superconducting magnet, the total voltage at both ends of the superconducting magnet coil and the current at any end of the superconducting magnet coil are collected in real time;

[0073] Based on the relationship curve of the superconducting magnet coil inductance versus current, the resistive voltage V in the superconducting magnet coil is determined by the total voltage V(t) across the superconducting magnet coil and the current I(t) at either end of the superconducting magnet coil. IR (t);

[0074] When it is detected that the resistive voltage increases to a voltage threshold, and / or the duration of the continuous increase of the resistive voltage exceeds a time threshold, it is determined that a quench occurs in the superconducting magnet coil.

[0075] The working principle of the above technical solution is:

[0076] The total voltage across the coil and the current at one end are collected in real time to capture the dynamic changes of electromagnetic parameters during operation. The relationship curve between inductance and current is used to separate the resistive voltage through the difference between the total voltage and the expected inductive voltage in the ideal superconducting state (calculated based on inductance, current and its rate of change). The resistive voltage is theoretically zero in the superconducting state, and resistive voltage is generated due to the appearance of resistance during quenching. The resistive voltage is used as the core criterion and combined with dynamic threshold logic (voltage threshold, duration threshold) to determine quenching.

[0077] The effects of the above technical solution are:

[0078] By accurately separating the resistive voltage (eliminating the inductive voltage interference caused by the nonlinearity of the inductance) and combining it with an adaptive threshold, it is possible to effectively identify tiny resistive voltage signals and solve the problem of them being easily submerged by noise. At the same time, there are no special requirements for the structure of the superconducting magnet, especially the multiple coil structures inside the superconducting magnet do not need to be approximately consistent with the inductance, and there is no need to compare the multiple coil voltage signals inside the superconducting magnet. It is only necessary to measure the voltage of the entire superconducting magnet. The present invention has no special requirements for the winding process of the superconducting magnet coil. It does not require various sensing structures such as optical fibers or temperature sensors to be pre-embedded during coil winding. It does not even require pre-embedded voltage measurement leads on the coils inside the magnet, nor does it require an equivalent superconducting magnet to be established outside the superconducting magnet for auxiliary measurement or calculation. It only needs to measure the voltage of the entire superconducting magnet to achieve quench detection. Through online calibration of the inductance curve and dynamic correction of the threshold, it is applicable to superconducting magnets under different operating conditions, improving the reliability of detection under complex working conditions.

[0079] In a possible implementation, the relationship curve between the superconducting magnet coil inductance and current is obtained by the following method, including:

[0080] Establish a curve showing the relationship between the inductance of a superconducting magnet coil and the change in current through simulation; or,

[0081] When the superconducting magnet coil is in a superconducting state, during the current dynamic adjustment process, the inductance values ​​corresponding to different current values ​​are synchronously measured to establish and correct the relationship curve between the superconducting magnet coil inductance and current change;

[0082] During steady-state operation of the superconducting magnet, the relationship curve is calibrated online by detecting the rate of change of the magnet coil current and using the total voltage across the superconducting magnet coil and the current at either end of the superconducting magnet coil at the current moment; and the established relationship curve of the superconducting magnet coil inductance versus current is used to predict the relationship between the inductance and current of a larger superconducting magnet within a larger current range.

[0083] The working principle of the above technical solution is:

[0084] Use simulation (such as finite element analysis) to preliminarily establish the inductance-current relationship curve, and calculate the change of inductance value with current based on the magnet structure parameters (such as the number of coil turns and geometric dimensions) and superconducting material characteristics (such as magnetization curve);

[0085] Or, in the superconducting state, by dynamically adjusting the current (such as the current increase and decrease process), synchronously measuring the inductance corresponding to different current values ​​(using the voltage-current phase difference or energy storage method), directly experimentally modeling and correcting the simulation curve to ensure consistency with the actual characteristics.

[0086] When the magnet is operating in steady state, the current change rate is detected in real time. Combined with the total coil voltage and current, the current inductance value is inferred through the formula (total voltage = inductance × current change rate + resistive voltage, and the resistive voltage is approximately 0 in the superconducting state). The established relationship curve is compared and dynamically calibrated to offset inductance deviations caused by temperature drift, mechanical stress, etc.

[0087] By utilizing the verified inductance-current relationship curves for small-sized magnets or small current ranges, combined with electromagnetic theory (such as scale effect and current density distribution law), the inductance change trend of larger-sized magnets (such as longer coils and higher field strengths) in a larger current range can be predicted, reducing the cost and risk of large magnet experiments.

[0088] The effects of the above technical solution are:

[0089] The modeling approach, which combines simulation and experimentation, uses simulation to rapidly cover multiple operating conditions while correcting theoretical deviations through experimentation. Online calibration provides real-time compensation for dynamic disturbances during operation (such as vibration and magnetic field distortion), ensuring that the curve always reflects the true state of the magnet. Predicting the inductance characteristics of large magnets with high currents and large dimensions using small-scale data reduces the number of tests required for large magnets under extreme operating conditions, shortening the R&D cycle and minimizing the risk of equipment loss. Accurate inductance-current curves are the basis for calculating inductive voltages and separating resistive voltages (referring to the principle of resistive voltage detection). The accuracy of these curves directly improves the sensitivity of quench determination, avoiding misjudgments or missed detections due to inductance calculation errors.

[0090] In a possible implementation, the current dynamic regulation process adopts an adaptive segmented scanning strategy, including:

[0091] The current range is divided into a core interval and a non-core interval. The core interval uses a first step size (small step size, for example, a step size of 0.5%-1% of the rated current) to ensure data density under critical operating conditions, while the non-core interval uses a second step size (large step size, for example, a step size of 2%-5% of the rated current) to improve overall scanning efficiency. The core interval is dynamically determined based on the high-frequency operating range of the superconducting magnet's historical operation and the curvature change of the inductance-current curve.

[0092] Real-time monitor the inductance change rate, when the absolute value of the inductance change rate exceeds a first threshold (for example, 0.1 H / A), temporarily scale down the step size of the current interval, restore after not less than 3 steps, and the scale-down ratio is dynamically adjusted according to the trend of the inductance change rate; the scale-down ratio is the first threshold / absolute value of the inductance change rate;

[0093] After scanning, the core interval data is complemented by using a cubic spline interpolation with a noise weight factor, and the non-core interval data is complemented by using a piecewise linear-spline hybrid interpolation, to establish a relationship curve of the inductance of the superconducting magnet coil with the current change;

[0094] In a possible implementation, the method further includes:

[0095] After scanning, the preliminary constructed relationship curve is substituted into the superconducting magnet steady-state operation data for verification, and if the deviation between the theoretical inductance value and the actual measured value exceeds a calibration threshold, the interval with the deviation is scanned again and the interpolation curve is corrected;

[0096] In combination with a quench risk evaluation model of the superconducting magnet, a current interval with a high quench probability is given a sampling priority, and when the total scanning time is limited, the first step size sampling of the high-risk interval is preferentially ensured.

[0097] The above technical solution has the following effects:

[0098] Through the segmented step size design, high-density data is obtained in the core interval of the magnet main work, and the non-core interval is sampled redundantly through a large step size, which greatly reduces the scanning time while ensuring the accuracy of the key area; the step size is dynamically reduced for the inductance mutation area, avoiding the loss of curve details caused by the fixed step size, and it is especially suitable for scenes with obvious nonlinear characteristics such as magnetic saturation at high field; the cubic spline interpolation of the core interval can effectively fit the smooth nonlinear change of the inductance with the current, and the linear interpolation of the non-core interval simplifies the calculation, and the overall curve not only meets the accuracy requirement of the inductive voltage calculation in the quench detection, but also reduces the data storage and processing cost; without relying on the inherent structure of the magnet or the preset model, the superconducting magnet with different winding processes and different inductance characteristics is adapted through dynamic adjustment, providing reliable basic data support for subsequent resistive voltage separation. Through the steady-state operation data verification and the secondary scanning correction of the deviation interval, the accuracy of the inductance change relationship curve with the current is further improved, providing a reliable basis for subsequent quench detection and other applications. In combination with the quench risk evaluation model, the small step size sampling of the high quench probability interval is preferentially ensured when the total scanning time is limited, the resources are reasonably allocated, and the safety of the superconducting magnet operation is improved.

[0099] In a possible implementation, the current dynamic adjustment process uses a symmetric triangular wave current sequence:

[0100] Apply a triangular wave current with increasing amplitude. Each triangular wave cycle includes a forward excitation phase (current increases from 0 to peak value I k ) and reverse excitation phase (current from I k to 0), where Ik=k×ΔI (k is a positive integer, ΔI is the current step size, I k The maximum value does not exceed 110% of the rated current);

[0101] Record the inductance value L corresponding to the current I during the forward excitation phase + (I), in the reverse excitation stage, record the inductance value L corresponding to the same current I - (I);

[0102] Calculate the mean inductance L(I) for each current I = [L + (I)+L - (I)] / 2, based on the corresponding relationship between L(I) and I, a curve showing the relationship between the inductance of the superconducting magnet coil and the change of current (LI curve) is established.

[0103] The effects of the above technical solution are:

[0104] By leveraging the symmetry of forward and reverse excitation, the inductance measurement deviation caused by the hysteresis effect is offset through averaging, improving the accuracy of the LI curve. This is particularly suitable for superconducting magnets with significant hysteresis characteristics. The triangular wave amplitude is increased in fixed steps to ensure that sampling points are present throughout the entire current range from 0 to 110% of the rated current, avoiding data loss in critical current segments. Data acquisition is completed through periodic scanning of the symmetrical triangular wave without the need for complex dynamic step adjustment logic, reducing the complexity of the control algorithm. Bidirectional sampling and averaging reduce the random error of a single measurement, making the constructed LI curve more consistent in repeated experiments and providing a reliable inductance benchmark for subsequent resistive voltage separation.

[0105] In a possible implementation, the current dynamic regulation process further includes:

[0106] Establish a database to store the relationship curve of superconducting magnet coil inductance versus current (LI curve) and corresponding magnet state parameters obtained from previous scans; the magnet state parameters include magnet cooling time and cumulative number of operations;

[0107] Through the trained machine learning model, the current magnet state parameters are input to predict the possible change range of the LI curve [I start , I end ];

[0108] Only in [I start , I end] interval to perform dynamic adjustment scanning, and the other intervals call the LI data of the most recent scan. If the change amplitude predicted by the model exceeds the amplitude threshold (for example, 5%), the scanning range is extended to [I star t-ΔI, I end +ΔI].

[0109] The working principle of the above technical solution is: establish a database to continuously store the LI curves and corresponding magnet state parameters (including cooling time, cumulative number of operations, etc.) obtained from previous dynamic adjustment scans, forming a historical data set covering different magnet states, providing a basis for model training and data reuse.

[0110] A machine learning model is trained based on a historical dataset. By inputting current magnet state parameters (cooling time, cumulative number of runs), it predicts the interval [Istart, Iend] where the LI curve is likely to change significantly. This interval represents the current range where magnet state changes most significantly affect inductance characteristics. Real-time dynamic adjustment scanning is performed only within the predicted interval [Istart, Iend] to obtain the latest LI data. For other current intervals, the most recently scanned LI data in the database is directly used to reduce redundant scans. If the model's predicted LI curve change exceeds a preset threshold, the scan range is automatically expanded to [Istart-ΔI, Iend+ΔI] to ensure coverage of potentially significant change areas.

[0111] The effects of the above technical solution are:

[0112] The machine learning model accurately predicts the change range of the LI curve, performs real-time scanning only on the key range, and reuses historical data for other ranges, significantly reducing the invalid scanning range and reducing the time cost and energy consumption of dynamic adjustment. Combined with state parameters such as the cooling time of the magnet and the cumulative number of operations, machine learning is used to capture the influence of magnet aging and state changes on the LI curve, so that the scanning range dynamically matches the actual changes in the magnet characteristics, avoiding the blindness of "full-range scanning". When the predicted change amplitude exceeds the threshold, the scanning range is expanded to ensure that the key change area is not missed, taking into account the efficiency of focusing on the core range and the accuracy of covering significant changes. It is especially suitable for scenarios with large fluctuations in the magnet state. The previous LI curves stored in the database provide reliable data support for non-critical intervals, reducing the waste of resources caused by repeated scanning, and at the same time ensuring the continuity of the curve in the non-changing interval through the most recent data call.

[0113] In a possible implementation, the resistive voltage V in the superconducting magnet coil is determined based on the relationship curve of the superconducting magnet coil inductance versus current (LI curve) by the total voltage V(t) across the superconducting magnet coil and the current I(t) at any end of the superconducting magnet coil. IR (t); including:

[0114] Based on the curve of the superconducting magnet coil inductance versus current and the real-time current acquisition, the inductance value L(t) corresponding to the current I(t) and the derivative of the inductance with respect to the current dL / dI are determined; L(t) = L(I(t)).

[0115] Based on the current inductance value L(t), the current current change rate and the current current I(t), the ideal inductive voltage V under the ideal superconducting state is calculated. ideal (t);

[0116]

[0117] in, is the current rate of change of inductance. For magnets whose inductance changes slowly over time, it can be approximated to zero. It can be obtained by real-time interpolation of the slope of the curve L(I) showing the relationship between the superconducting magnet coil inductance and current change. dI(t) / dt is the current rate of change of current. Represents the inductive voltage component caused by the change of current over time; Represents the inductive voltage component caused by the change of inductance over time;

[0118] By comparing the total voltage V(t) with the ideal inductive voltage V ideal (t), determine the resistive voltage V in the superconducting magnet coil IR (t); total voltage V(t) and ideal inductive voltage V ideal The difference between (t) is the resistive voltage V IR (t), v IR (t)=V(t)-V ideal (t).

[0119] The effects of the above technical solution are:

[0120] By dynamically correlating the inductance parameters with the LI curve, the two physical components of the inductive voltage (the inductive voltage caused by current changes and inductance changes) are completely decomposed, eliminating the interference of inductive nonlinearity on resistive voltage detection. This solves the problem of inductive voltage aliasing in traditional methods that makes it difficult to accurately extract the resistive voltage. Based on the slope interpolation and dynamic parameter calculation of the LI curve, it is possible to track the fluctuations of the inductance characteristics caused by current changes in real time, adapt to the nonlinear changes of the inductance during the operation of the superconducting magnet (such as different current ranges and magnet aging), and ensure the accuracy of resistive voltage detection under all operating conditions. There is no need to embed complex sensing structures (such as distributed voltage / temperature sensors) inside the magnet; resistive voltage separation can be achieved solely through external total voltage and current measurement, reducing system complexity and the difficulty of magnet design and manufacturing. The accurately extracted resistive voltage is a direct electrical signature of the quench (the quench is accompanied by resistance generation).

[0121] In a possible implementation, the current current change rate is a ratio of a difference between current values ​​at adjacent moments collected in real time to a time interval.

[0122] In another possible implementation, an adaptive sliding window difference algorithm is used to determine the current current change rate, and the window width is dynamically adjusted according to the current change rate:

[0123]

[0124] τ min is the minimum window width, τ max is the maximum width of the window, α is the adjustment coefficient;

[0125] Among them, in the superconducting magnet coil simulation fast quench experiment, high-frequency voltage noise (for example: amplitude is 0.1% of the rated voltage, frequency covers 10kHz and above), with the current mutation signal response requirement and voltage noise tolerance as dual benchmarks; when the window width is reduced to a certain value, the current change rate calculation result captures the high-frequency mutation error less than the preset error threshold (for example, 0.5%), and the calculation deviation introduced by the voltage noise is lower than the noise tolerance threshold (for example, the noise variance increase is less than 1%), and this value is τ min ;

[0126] Under the steady-state operation condition of the superconducting magnet, high-frequency current noise (for example, high-frequency current noise with an amplitude of 0.1% of the rated current) and high-frequency voltage noise (for example, high-frequency voltage noise with an amplitude of 0.1% of the rated voltage) are injected; when the window width is increased to a certain value, the noise variance of the current change rate calculation result is reduced to below a preset threshold (for example, 5% of the original variance), and the tracking error of the slow change of the current change rate below the critical value of the steady-state current change rate does not exceed the accuracy threshold (for example, 0.1%), and at the same time, the probability of false change rate misjudgment caused by voltage noise is lower than the preset probability threshold (for example, 1%). This value is τ max ;

[0127] Through dynamic experiments on superconducting magnet coils under rated operating conditions, sample data with different current change rates and different voltage noise intensities (represented by voltage noise variance or amplitude) are collected; with the dual goals of minimizing the sum of squares of the deviations between the calculated current change rate and the actual sensor value and keeping the misjudgment rate introduced by voltage noise below a preset probability threshold (for example, 5%), the least squares method is used to fit the adjustment coefficient α.

[0128] The working principle and effect of the above technical solution are as follows: it provides two current rate-of-change calculation paths. The first is a simple difference method, which directly calculates the current rate of change by comparing the difference between adjacent currents to the time interval. The second is an adaptive sliding window difference method, which dynamically adjusts the sliding window width. By dynamically adjusting the window width, a small window ensures accurate capture of high-frequency changes in rapid quench scenarios (such as sudden current changes), with an error that meets a preset threshold (e.g., within 0.5%). In steady-state, a large window effectively suppresses noise, allowing the tracking error of slow current changes to fall below an accuracy threshold (e.g., 0.1%), thus resolving the problem of a fixed window that "pays attention to one thing while neglecting another." The adjustment coefficient k is optimized through dual-objective fitting, minimizing the deviation between the calculated current rate of change and the measured value under various noise and rate-of-change conditions, improving calculation accuracy across all operating conditions. It also offers strong anti-interference capabilities: the experimentally calibrated maximum and minimum window widths limit the calculation bias introduced by noise (e.g., the voltage noise variance increase is less than 1%) and the probability of false rate-of-change misjudgment (e.g., less than 1%), respectively, significantly reducing the interference of voltage and current noise on the current rate-of-change calculation. Dynamic window adjustment mechanism: when the noise is large and the current change rate is low, the large window smoothes the noise; when the noise is small and the change rate is high, the small window focuses on the sudden change, without manual intervention.

[0129] Refer to the attached Figure 4 In one possible implementation, the logic for determining whether the resistive voltage increases to a voltage threshold includes:

[0130] When the resistive voltage exceeds its starting voltage threshold for the first time, the continuous growth timer starts, and a quench is determined if any of the following conditions is met:

[0131] The resistive voltage exceeds the final voltage threshold before reaching the duration threshold;

[0132] When the resistive voltage reaches the duration threshold, it exceeds the final voltage threshold;

[0133] The resistive voltage continuously increases for a period exceeding a time threshold;

[0134] The resistive voltage includes an original resistive voltage and a filtered resistive voltage, and the decision logic is independently executed on the original resistive voltage and the filtered resistive voltage;

[0135] The starting voltage threshold is set based on the maximum noise peak value of the corresponding signal during normal operation (the original signal is not less than 1.5 times the noise peak value, and the filtered signal can be adjusted due to noise reduction); the final voltage threshold is set based on the exponential growth characteristic of the resistive voltage;

[0136] In a possible implementation, the final voltage threshold is determined by calculating the maximum temperature of the hot spot inside the magnet coil when the voltage is reached through a quench simulation experiment.

[0137] In one possible implementation, the final voltage threshold is the product of the resistive voltage at the determination starting point and a growth coefficient calibrated by the quench simulation experiment, and the growth coefficient is calibrated by the average growth slope of the resistive voltage in the quench simulation experiment; the duration threshold is positively correlated with the corresponding signal noise level;

[0138] For example, the resistive voltage is an original resistive voltage, and the determination logic of whether the resistive voltage increases to a voltage threshold includes:

[0139] Setting a first starting voltage threshold, a final voltage threshold of the original signal, and a first duration threshold;

[0140] When the original resistive voltage exceeds the first starting voltage threshold for the first time, the quench determination timing is triggered, the current moment is recorded as the determination starting point and the continuous growth timing is started;

[0141] If any of the following conditions is met, it is considered a quench:

[0142] During the continuous growth process, the original resistive voltage exceeds the final voltage threshold of the original signal before reaching the first duration threshold;

[0143] When the original resistive voltage continuously increases to the first duration threshold, it has exceeded the final voltage threshold of the original signal;

[0144] The duration of the continuous increase of the original resistive voltage exceeds a first duration threshold;

[0145] The first starting voltage threshold is not less than N times the maximum noise peak during normal operation; N≥1.5, and the final voltage threshold of the original signal is set based on the exponential growth characteristics of the resistive voltage, satisfying the determination of the resistive voltage at the starting point × the growth coefficient, and the growth coefficient is calibrated by the average growth slope of the resistive voltage in the quench simulation experiment.

[0146] For example, the resistive voltage is a filtered resistive voltage, and the determination logic for whether the resistive voltage increases to a voltage threshold includes:

[0147] The original resistive voltage signal collected in real time is filtered and denoised to obtain a smoothed resistive voltage curve;

[0148] Setting a second starting voltage threshold (which may be the same as or different from the first starting voltage threshold), a final voltage threshold of the filtered signal, and a second duration threshold;

[0149] When the smooth curve exceeds the second starting voltage threshold for the first time, the continuous growth timing is started;

[0150] If any of the following conditions is met, it is considered a quench:

[0151] During the continuous growth process, the smooth curve exceeds the final voltage threshold of the filtered signal before reaching the second time threshold;

[0152] When the smooth curve continuously increases to the second time threshold, it has exceeded the final voltage threshold of the filtered signal;

[0153] The duration of the continuous growth of the smooth curve exceeds a second duration threshold; and the window length of the smoothing algorithm is positively correlated with the noise fluctuation coefficient.

[0154] The continuous growth of the original resistive voltage and the smoothed resistive voltage is defined as:

[0155] For the original signal: a single fluctuation drop due to noise interference is allowed, and the drop shall not exceed a preset proportion of the voltage value before the fluctuation (the proportion is set based on the noise level of the original signal, for example, 15%), and the voltage after the fluctuation must remain no lower than the first starting voltage threshold; at the same time, within the set time interval, the cumulative duration of the non-negative voltage difference at adjacent moments shall not be less than a first preset percentage (the percentage is determined based on the noise characteristics of the original signal, for example, 90%). Meeting the above conditions is regarded as continuous growth.

[0156] For the smoothed resistive voltage signal (processed by weighted moving average filtering): because the noise is suppressed, the allowed single fluctuation decline does not exceed the second preset proportion of the voltage value before the fluctuation (this proportion is lower than the preset proportion of the original signal, adapted to the characteristics of the smoothed signal, such as 5%); and within the set time interval (this interval is shorter than the time interval of the original signal because the smoothed signal is more stable), the cumulative duration of the non-negative voltage difference at adjacent moments is not less than the second preset percentage (this percentage is higher than the first preset percentage of the original signal, such as 95%). Meeting the above conditions is regarded as continuous growth.

[0157] The continuous growth timing of both signals starts from the first time it exceeds its corresponding starting voltage threshold (the original signal corresponds to the first starting voltage threshold, and the smoothed signal corresponds to the second starting voltage threshold). If the voltage after fluctuation is lower than its corresponding starting voltage threshold, the timing is reset. Under ideal conditions, a new timing will be started when the resistive voltage is greater than the starting voltage threshold at any time, regardless of whether other timing tasks are currently in progress.

[0158] When the original resistive voltage curve that has not been smoothed exceeds its set starting voltage threshold (triggering the quench determination timing), if the smoothed resistive voltage curve decreases during the timing process and the decrease reaches or exceeds the preset amplitude, the current continuous growth timing is paused; after the smoothed curve resumes the growth trend, the timing is restarted, and the cumulative number of such interruptions (for example, 2 times) is limited to no more than a preset value.

[0159] The determination of the number of interruptions is based on the comprehensive optimization results of noise characteristics, quench signal characteristics and engineering practice. The core goal is to strike a balance between anti-interference ability and detection sensitivity. Setting an interruption tolerance of less than 2 times can cover most normal noise scenarios.

[0160] The working principle of the above technical solution is:

[0161] The original resistive voltage and the filtered, smoothed resistive voltage are collected, and judgment parameters (starting voltage threshold, final voltage threshold, and duration threshold) are set to suit their signal characteristics. The original signal retains more details and responds faster, while the smoothed signal filters out noise and is more stable. The two work together to achieve complementary detection.

[0162] The process of "starting voltage threshold trigger - continuous growth timing - final voltage threshold / duration threshold judgment" is adopted for both signals: when the signal exceeds the starting voltage threshold for the first time, the timing is started. If the final voltage threshold is exceeded during the timing process, or the continuous growth duration exceeds the duration threshold, it is judged as a quench.

[0163] Continuous growth is defined differently based on signal characteristics: the original signal allows for larger noise fluctuations (such as a single drop ≤ 15%), and the cumulative proportion of non-negative voltage differences at adjacent moments is required to be lower (such as ≥ 90%); the smoothed signal is suppressed due to noise, and the allowed fluctuation amplitude is smaller (such as a single drop ≤ 5%), and the cumulative proportion of non-negative differences is required to be higher (such as ≥ 95%). If the voltage after the fluctuation is lower than the corresponding starting voltage threshold, the timing is reset directly. When the original signal triggers the timing, if the smoothed signal shows a significant drop (such as ≥ 5%), the timing is paused, and the timing is restarted after it resumes growth. The cumulative number of interruptions does not exceed the preset value (such as 2 times). This mechanism can filter out short-term noise interference and avoid misjudgment or missed judgment due to occasional fluctuations.

[0164] The effects of the above technical solution are:

[0165] Through dual-dimensional verification of the original signal and the smoothed signal, combined with differentiated continuous growth judgment criteria, the rapid response of the original signal to quench is retained, and the smoothed signal is used to suppress noise, effectively reducing the risk of misjudgment of single signal detection. The fluctuation tolerance, non-negative difference ratio requirements, and limited interruption fault tolerance mechanism designed for different signal characteristics can adapt to complex environments such as electromagnetic interference and measurement noise during the operation of superconducting magnets, ensuring stable identification of true quench signals. The hierarchical threshold setting (the starting voltage threshold is based on the noise peak, and the final voltage threshold is based on the quench growth characteristics) and dynamic timing logic ensure rapid response to quench while avoiding missed judgments through the duration threshold, achieving an optimal balance between sensitivity and reliability.

[0166] In a possible implementation, the resistive voltage includes an original resistive voltage and a filtered resistive voltage, and the determination logic is cooperatively executed as follows:

[0167] When the original resistive voltage triggers the starting voltage threshold and the filtered resistive voltage synchronously presents a growth trend within a preset delay, the corresponding time threshold is shortened by a preset proportion (for example, 50%-80% of the original value);

[0168] When only the filtered resistive voltage triggers the starting voltage threshold, a secondary verification process is started: if the original resistive voltage does not reach a preset proportion (80%) of the starting voltage threshold in the next sampling period, the timer of the filtered signal is reset;

[0169] The preset delay is set based on the response characteristics of the filtering process and the noise fluctuation period, and the sampling period is adapted to the signal acquisition frequency;

[0170] If the original resistive voltage reaches the preset proportion of the starting voltage threshold but does not exceed the starting voltage threshold in the secondary verification, the timer can be maintained and the time threshold can be adaptively adjusted by a time correction coefficient.

[0171] In the cooperative execution logic:

[0172] The value range of the preset proportion is determined according to engineering fault tolerance design and simulation experiment data of the double-feature synchronization of the loss of field signal;

[0173] The time correction coefficient is positively correlated with the fluctuation amplitude of the original signal;

[0174] The determination result of the secondary verification does not affect the execution of the original resistive voltage independent determination logic.

[0175] The working principle and effects of the above technical solutions are as follows:

[0176] When the original signal (original resistive voltage) triggers the starting voltage threshold and the filtered signal (filtered resistive voltage) synchronously grows within a preset delay, it indicates that the loss of field feature is clear (the original signal responds quickly + the filtered signal eliminates noise interference), and at this time, the time threshold is shortened by a preset proportion to speed up the loss of field determination speed. Shortening the time threshold when the double signals are synchronized can speed up the response speed in the scenario where the loss of field feature is clear, and gain time for superconducting magnet protection.

[0177] When only the filtered signal triggers the start-up voltage threshold, secondary verification using the original signal is required due to potential lag or oversmoothing in the filtering process. If the original signal does not reach the preset ratio of its start-up voltage threshold, the filtered signal may be a false trigger due to noise, and the timer is reset to avoid misjudgment. If the ratio is reached but the threshold is not exceeded, the duration threshold is adjusted to balance reliability and sensitivity. This secondary verification mechanism, when triggered by a single filtered signal, filters out false increases caused by noise through reverse verification of the original signal, making it particularly suitable for scenarios where the filtering algorithm may introduce lag.

[0178] Secondary verification only affects the timing of the filtered signal and does not interfere with the independent determination of the original signal, ensuring that the original signal's fast response advantage is not compromised. Dynamic adjustment of parameters such as preset ratios and duration correction coefficients allows adaptation to varying noise environments (such as changes in electromagnetic interference intensity) and quench characteristics (such as differences in growth rates), enhancing system robustness.

[0179] In one possible implementation, the voltage threshold is an adaptive threshold dynamically modified based on real-time operating conditions (including a starting voltage threshold and a final voltage threshold set in stages), including:

[0180] The initial value of the voltage threshold is determined through a quench simulation experiment of the superconducting magnet coil at rated current. The starting voltage threshold of the raw resistive voltage is set based on the voltage noise peak value of the quench simulation experiment at rated current and is not less than a preset multiple (≥1.5 times) of the peak value. The starting voltage threshold of the filtered resistive voltage is set based on the peak value of the filtered noise and can be appropriately lowered due to the noise suppression effect. The final voltage thresholds of the raw signal and the filtered signal are both set based on the exponential growth characteristics of the corresponding signals and are the product of the resistive voltage at the determination starting point and the growth coefficient calibrated by the quench simulation experiment.

[0181] During the operation of the superconducting magnet, the ambient temperature T(t) of the coil, the pressure of the cooling medium P(t), and the noise component Vn(t) of the total voltage across the coil are collected in real time.

[0182] The noise component Vn(t) is extracted by high-frequency filtering of the total voltage signal and represented by the standard deviation of the voltage signal within a sliding window;

[0183] Dynamically adjusting the voltage threshold based on a preset correction coefficient matrix and a noise component; Dynamically adjusting the voltage threshold based on a preset correction coefficient matrix and a noise component; the correction coefficient matrix includes a temperature correction term, a pressure correction term, and a noise correction term;

[0184] Vth(t)=Vth0×[1+kT×(T(t)-T0)+kP×(P(t)-P0)]+kN×Vn(t)

[0185] Wherein, Vth0 is the initial voltage threshold under the rated working condition, T0 is the rated ambient temperature, P0 is the rated cooling pressure, kT is the temperature correction coefficient, kP is the pressure correction coefficient, kN is the noise correction coefficient, and kT, kP are calibrated through multivariate quench sensitivity experiment, kN is determined through noise influence experiment, and the value range is 1.5-2.5;

[0186] When the relationship curve is detected to be calibrated online, the voltage threshold is corrected in synchronization with the calibration proportion, and the calculation window of the noise component Vn(t) is adaptively adjusted; if it is found after calibration that the current noise characteristics (such as noise frequency, fluctuation amplitude) deviate from the historical reference (for example, the proportion of high-frequency noise increases or the noise fluctuation intensifies), the time length of the sliding window is shortened (or the number of data points is reduced) to improve the response sensitivity to high-frequency noise; if the noise tends to be stable or the proportion of low-frequency components increases, the time length of the sliding window is lengthened (or the number of data points is increased) to reduce random noise interference and ensure the accuracy and real-time performance of the noise component calculation adapt to the current working condition.

[0187] In a possible implementation, the correction coefficient matrix is dynamically updated in the following manner:

[0188] Collect non-quenching abnormal data of the superconducting magnet coil under different working conditions and corresponding voltage noise data to construct an abnormal sample set containing noise characteristics;

[0189] The temperature correction coefficient, the pressure correction coefficient and the noise correction coefficient are iteratively optimized by using the gradient descent algorithm, so that the misjudgment rate of the voltage threshold in the abnormal sample set is lower than a set value, and the detection delay of the small resistive voltage in the quench simulation experiment does not exceed a preset threshold; wherein, the small resistive voltage refers to a resistive voltage signal with an amplitude of 1-3 times the voltage noise peak value under the working condition;

[0190] The optimized correction coefficients are updated synchronously with the online calibration cycle.

[0191] The working principle and effects of the above technical solutions are as follows:

[0192] Based on the quench simulation experiment of the superconducting magnet under the rated current, the voltage threshold is determined: the original resistive voltage starting voltage threshold is not less than a preset multiple (such as ≥1.5 times) of the noise peak value, to ensure the suppression of normal noise; the starting voltage threshold after filtering can be appropriately reduced due to the suppression of noise, to balance the sensitivity; the final voltage threshold is set based on the signal exponential growth characteristic, which is the product of the starting point resistive voltage and the experimental calibration growth coefficient, to ensure the accurate capture of the quench trend.

[0193] The environmental temperature T(t) and the cooling medium pressure P(t) are collected in real time during operation, and the noise component Vn(t) (represented by the standard deviation of the voltage in the sliding window) in the total voltage is extracted through high-frequency filtering, comprehensively reflecting the changes of the magnet operating environment and signal quality.

[0194] Based on the preset correction coefficient matrix (including temperature, pressure, and noise correction terms), the voltage threshold is adjusted in real time through the formula:

[0195] The temperature correction term kT×(T(t)-T0): compensates for the influence of environmental temperature deviation from the rated value on superconducting characteristics (such as temperature rise may cause critical current to drop, threshold needs to be lowered to respond in advance);

[0196] The pressure correction term kP×(P(t)-P0): adapts to the change of cooling medium pressure (such as insufficient pressure, cooling effect decreases, risk of quench increases, threshold sensitivity needs to be adjusted);

[0197] The noise correction term kN×Vn(t): dynamically raises the threshold with the noise intensity to avoid noise false triggering (the greater the noise, the more the threshold is raised).

[0198] At the same time, when the inductance-current relationship curve is calibrated online, the threshold is corrected in proportion and the noise calculation window is adjusted simultaneously to ensure matching with the real-time state of the magnet.

[0199] By collecting non-quench abnormal data to construct a sample set, and using gradient descent algorithm to iteratively optimize kT, kP, and kN:

[0200] The goal is to make the misjudgment rate in abnormal samples lower than the set value, while ensuring that the detection delay of small resistive voltage (noise peak 1-3 times) is within the preset range;

[0201] The optimized correction coefficients are updated synchronously with the online calibration period, so that the threshold correction logic continuously adapts to the changes in the characteristics of the magnet during long-term operation.

[0202] In one possible implementation, the duration threshold is preset as a dynamic threshold associated with the resistive voltage growth slope and voltage noise level, including:

[0203] The duration threshold (including the first duration threshold of the original resistive voltage and the second duration threshold of the filtered resistive voltage) is a dynamic threshold associated with the resistive voltage growth slope and noise fluctuation coefficient of the corresponding signal (the signal includes the original signal and the filtered signal), including:

[0204] Respective slope thresholds are preset for the original resistive voltage and the filtered resistive voltage, which are the ratio of the difference between adjacent resistive voltage values and the time interval (the original signal slope threshold is higher than the filtered signal, which adapts to the noise characteristics);

[0205] Real-time growth slope of the corresponding signal resistance voltage is calculated (for example, the original signal is calculated at 10 ms intervals, the filtered signal is calculated at 5 ms intervals, and the adaptive smoothing characteristics are adopted);

[0206] The voltage noise fluctuation coefficient of the corresponding signal is calculated in real time, which is the ratio of the current noise component to the reference noise value under the rated working condition (the original signal is based on the high-frequency noise component, and the filtered signal is based on the standard deviation in the smoothing window);

[0207] For different magnets or different operating power segments of the same magnet, a plurality of combinations of slope threshold values, first reference time and second reference time are preset;

[0208] When the real-time growth slope is not less than the slope threshold value of the signal, the time threshold value is set as the product of the first reference time and the noise fluctuation coefficient; when it is less than the slope threshold value, it is set as the product of the second reference time and the noise fluctuation coefficient, and the first reference time < the second reference time, and the noise fluctuation coefficient ≥ 1;

[0209] The slope threshold values, the first / second reference times of the two types of signals are determined through the quench evolution simulation experiment (the reference time of the original signal is shorter than that of the filtered signal), which correspond to the response critical time of the fast quench scene and the cumulative judgment time of the slow quench scene, respectively.

[0210] The working principle and effect of the above technical solution are as follows:

[0211] According to the signal characteristics of the original resistance voltage and the filtered resistance voltage (the original signal has high noise and large fluctuation, and the filtered signal is smooth and stable), the slope threshold value (the ratio of the voltage difference between adjacent time points to the time interval) is preset respectively: the slope threshold value of the original signal is higher than that of the filtered signal, so as to adapt to the natural fluctuation caused by noise and avoid misjudgment of noise as "rapid growth".

[0212] According to the signal characteristics, the differential calculation interval is adopted (for example, 10 ms for the original signal and 5 ms for the filtered signal), so as to ensure that the original signal slope calculation can filter high-frequency noise, and the filtered signal slope calculation can capture the subtle growth trend after filtering.

[0213] The noise level of the corresponding signal is calculated in real time (for example, the original signal is based on the high-frequency noise component, and the filtered signal is based on the standard deviation in the smoothing window), which is represented by the ratio of the current noise value to the reference noise value under the rated working condition, and quantifies the signal quality.

[0214] The differential slope threshold value, calculation interval and reference time of the original signal and the filtered signal are set according to their inherent characteristics: the original signal filters noise through a higher slope threshold value and a slightly longer calculation interval, and the filtered signal captures subtle trends through a lower slope threshold value and a shorter calculation interval, so that the dynamic adjustment logic is consistent with the essential characteristics of the signal, and the judgment accuracy is improved.

[0215] When the real-time growth slope is greater than or equal to the preset slope threshold (indicating that the quench is developing rapidly): duration threshold = first reference time × noise fluctuation coefficient (the greater the noise, the longer the threshold should be to avoid misjudgment of rapid fluctuations caused by noise);

[0216] When the real-time growth slope is less than the preset slope threshold (indicating that the quench develops slowly): duration threshold = second reference time × noise fluctuation coefficient (a longer reference time ensures that slowly growing quenches are cumulatively identified, while the threshold is lengthened as noise increases to resist interference).

[0217] In addition, the reference time of the original signal is shorter than that of the filtered signal as a whole, which is adapted to its fast response characteristics to quench.

[0218] Different types of superconducting magnets (such as MRI magnets, nuclear fusion tokamaks, and superconducting energy storage magnets) have significant differences in coil structure, material properties, and cooling methods, resulting in different quench development patterns:

[0219] The resistive voltage of high-field magnets (such as superconducting magnets above 10T) increases rapidly and has a large slope after quenching, requiring a more sensitive slope threshold and a shorter reference time; the quench of low-field magnets develops more slowly and has a small slope, requiring a lower slope threshold and a longer reference time to avoid missed judgment.

[0220] The quench characteristics of the same magnet in different operating power ranges (such as 50%, 80%, and 100% of rated power) are also different:

[0221] When operating in the high power range, the current density is large, the Joule heat accumulates quickly after quenching, and the resistive voltage increases at a steep slope;

[0222] When operating in the low power range, the current density is small, the quench develops relatively slowly, and the slope is low.

[0223] Through quench evolution simulation experiments and measured data, a unique parameter combination is calibrated for each scenario (magnet type / power range):

[0224] For high power or high field magnets: preset high slope threshold + short first reference time + medium second reference time to adapt to their fast quench characteristics and prioritize response speed;

[0225] For low power segments or low field magnets: preset low slope threshold + medium first reference time + long second reference time to adapt to their slow quench characteristics and give priority to ensuring the reliability of cumulative judgment.

[0226] Based on the magnet type identification or real-time power monitoring results, the parameter combination (slope threshold, first / second reference time) of the corresponding scenario is automatically called to make the adjustment logic of the duration threshold accurately match the current quench risk level and development law.

[0227] In a possible implementation, determining whether a superconducting magnet coil has a timeout also includes multi-parameter fusion determination, including:

[0228] When the ratio of the original resistive voltage to the peak value of its own voltage noise reaches the first ratio (50%-80%) of the ratio of its starting voltage threshold to the noise peak value, or the ratio of the filtered resistive voltage to the peak value of its own voltage noise reaches the second ratio (40%-70%) of the ratio of its starting voltage threshold to the noise peak value, and the absolute value of the current change rate is not less than the preset threshold, the pre-failure warning is triggered;

[0229] In the early warning state, the first time threshold of the original resistive voltage and the second time threshold of the filtered resistive voltage are shortened to a third proportion of the original threshold (e.g., 50%-80%), and the coil local temperature monitoring and the real-time analysis of the voltage noise of the two types of signals are started at the same time;

[0230] If the rate of increase of the local temperature exceeds the critical value within the preset time after the warning, or the ratio of the growth amplitude of the original / filtered resistive voltage to the respective noise fluctuation amplitude exceeds the preset ratio, it is directly judged as a quench (not limited by the segment duration threshold); the preset ratio ∈ [2,5].

[0231] The working principle of the above technical solution is:

[0232] Based on the correlation characteristics of the resistive voltage and current changes, the warning trigger conditions are set: when the ratio of the raw resistive voltage to the peak value of the self-noise reaches a first ratio (e.g., 50%-80%) of the ratio corresponding to the starting voltage threshold, or the ratio of the filtered resistive voltage to the peak value of the self-noise reaches a second ratio (e.g., 40%-70%) of the ratio corresponding to the starting voltage threshold, it indicates that the voltage signal has deviated from the normal fluctuation trend;

[0233] At the same time, the absolute value of the current change rate is required to be no less than the preset threshold, eliminating simple voltage noise interference (such as electromagnetic disturbance) and ensuring that the warning is associated with the actual state change of the magnet (such as the risk of quench caused by current anomaly).

[0234] When the above conditions are met, the system enters the pre-failure warning state.

[0235] After the early warning is triggered, the system initiates two levels of reinforcement measures:

[0236] Dynamically shorten the duration threshold: shorten the duration thresholds of both the original signal and the filtered signal to a third ratio (e.g., 50%-80%) of the original thresholds to speed up the determination of subsequent growth signals.

[0237] Expanded monitoring dimensions: Local temperature monitoring of the starting coil (to capture the local temperature rise associated with quench) and real-time analysis of voltage noise of two types of signals (to distinguish noise from real signals) provide key features for the final judgment.

[0238] In the warning state, if any of the following strong characteristics appear, it will be directly judged as a quench (not subject to the original segment duration threshold):

[0239] The rate of increase of the local temperature within the preset time after the warning exceeds the critical value (the essence of quench is the destruction of the superconducting state caused by local heating, and the temperature characteristics have direct physical significance); the ratio of the growth amplitude of the original or filtered resistive voltage to the respective noise fluctuation amplitude exceeds the preset range (such as 2-5), indicating that the signal growth has been significantly stronger than the noise interference, and the quench characteristics are clear.

[0240] The effects of the above technical solution are:

[0241] Through the coordinated warning logic of voltage ratio + current change rate, potential quench risks can be identified before the resistive voltage reaches the starting voltage threshold (such as the 50%-80% threshold stage). This is several response cycles earlier than the traditional single threshold judgment, buying valuable time for magnet protection (especially for rapid quench scenarios).

[0242] Multi-parameter fusion (voltage, current change rate, temperature) avoids the limitations of single voltage signal judgment. The introduction of temperature characteristics provides direct physical evidence of quench, significantly reducing the risk of misjudgment. The judgment logic based on the signal / noise ratio (rather than the absolute value) can adapt to the noise level under different operating conditions, ensuring that the real signal growth can still be accurately identified in a strong noise environment.

[0243] In the early warning state, the time threshold is shortened to accelerate the judgment after the suspected risk occurs; and the final judgment relies on strong features such as rapid temperature rise or significant excess of signal-to-noise ratio to avoid false operations due to excessive sensitivity after the early warning, and find the optimal balance between fast response and robust judgment.

[0244] In a possible implementation, the method further includes:

[0245] Determine the equivalent resistance of the current superconducting magnet coil by using the resistive voltage and the current at either end of the current superconducting magnet coil;

[0246] If the equivalent resistance exceeds its equivalent resistance threshold, or the equivalent resistance continuously increases for a period exceeding a time threshold and its integral value exceeds the energy integral upper limit corresponding to the safe hotspot temperature, it is determined that the superconducting magnet coil has quenched; the integral upper limit is determined based on the enthalpy characteristics of the superconducting coil material and the energy absorption upper limit at a safe temperature.

[0247] Among them, the enthalpy value of the material (the relationship between the amount of heat absorbed per unit mass and the temperature) is the core thermodynamic parameter, which is mainly obtained through low-temperature experimental measurements. For example, in the temperature range of 4.2K to 300K, the enthalpy change curve of key materials in the superconducting coil (such as superconducting wire, copper stabilization layer, insulating material, etc.) is directly measured by calorimetry to establish a "temperature-enthalpy value" database. This is because the specific heat capacity characteristics of materials at low temperatures (which affect the enthalpy value) are very different from those at room temperature, and experimental data can better reflect the real physical properties. Simulation calculations (such as using thermodynamic simulation software) can further infer the enthalpy distribution under different working conditions (such as enthalpy changes in non-uniform temperature fields) based on the basic data measured experimentally, but they need to be calibrated based on experimental data.

[0248] The maximum energy carrying limit at a safe temperature, which must be determined based on the material's temperature tolerance and the energy-enthalpy balance relationship:

[0249] First, the safe hotspot temperature limit of the superconducting coil (e.g., 150K or 200K) was experimentally determined. Exceeding this temperature can lead to material degradation or structural damage. Based on the relationship "energy = mass × enthalpy increment," the maximum energy the coil can absorb (i.e., its energy carrying capacity) was calculated, combining the mass proportions of the various materials in the coil with the corresponding enthalpy values ​​at the safe temperature.

[0250] The working principle of the above technical solution is:

[0251] The equivalent resistance (R = resistive voltage / current) is calculated using the real-time collected resistive voltage (the pure resistive component after eliminating the inductive voltage of the superconducting coil) and coil current. The equivalent resistance is a key indicator reflecting the size of the normal region (non-superconducting state) of the superconducting coil. During superconducting failure, the normal region expands, causing a significant increase in resistance. During normal operation, the resistance is close to zero (containing only slight noise).

[0252] Adopting dual judgment logic of threshold breakthrough and duration-energy joint accumulation:

[0253] Condition 1 (Resistance Threshold Exceeded): When the equivalent resistance directly exceeds the preset equivalent resistance threshold, it indicates that the normal range has expanded to a dangerous level, and a quench is directly determined. This threshold is based on the resistance characteristics of the superconducting material at a safe hotspot temperature (e.g., 150-200K), ensuring that protection is triggered before temperature runaway.

[0254] Condition 2 (Duration-Energy Joint Determination): If the equivalent resistance does not immediately exceed the threshold, but shows a continuous growth trend:

[0255] When the continuous growth time exceeds the dynamic time threshold (adaptive noise level and resistance growth rate), and the integral value of the non-superconducting equivalent resistance over time exceeds the energy integral upper limit, it is determined to be quenched.

[0256] The setting of the integral upper limit is related to the enthalpy characteristics of the superconducting coil material (the material's ability to absorb energy) and the maximum energy carrying limit at a safe temperature, to avoid excessive hot spot temperatures due to energy accumulation.

[0257] The effects of the above technical solution are:

[0258] Taking equivalent resistance as the core detection indicator, it directly reflects the development status of the normal area of ​​the superconducting coil. Its threshold setting combines the minimum propagation zone (MPZ) characteristics with the resistance value corresponding to the safe hotspot temperature, which can effectively filter out false triggers caused by noise or tiny recoverable normal areas; at the same time, through the time dimension judgment of "continuous growth duration", it further suppresses transient interference, ensures that only the real quench trend is responded to, and greatly reduces the risk of misjudgment.

[0259] When the equivalent resistance exceeds the threshold, quench is directly determined, which can make an immediate response to the rapidly developing quench process and buy time for subsequent protection actions; the combined judgment logic of duration and integral value, for slowly developing quench scenarios, ensures that even if the quench process is slow, protection can be triggered in time before the hot spot temperature exceeds the standard through the energy integral upper limit (based on the material enthalpy value and the energy absorption limit at the safe temperature), thereby avoiding irreversible damage to the coil due to overheating and comprehensively covering quench conditions at different rates.

[0260] In a possible implementation, the equivalent resistance threshold includes a starting resistance threshold and a final resistance threshold; the duration threshold is a third duration threshold, and a quench is determined when any of the following conditions is met:

[0261] After the equivalent resistance exceeds the starting resistance threshold, the continuous increase timing is started, and the final resistance threshold is exceeded before the third time threshold is reached; or,

[0262] The equivalent resistance exceeds the final resistance threshold when reaching the third time threshold;

[0263] Among them, the starting resistance threshold is set according to the resistance value corresponding to the maximum noise during normal operation, and the starting resistance threshold is not less than the critical resistance corresponding to the minimum propagation area; ensure that the resistance generated in the normal area smaller than the minimum propagation area will not trigger the judgment; set the final resistance threshold according to the resistance value corresponding to the safe hot spot temperature in the quench simulation experiment, and calibrate it according to the resistivity change characteristics of the superconducting material in the normal state with temperature and the coil structure parameters; determine the third time threshold according to the equivalent resistance growth slope and the noise level.

[0264] Continuous growth of equivalent resistance means: in the original signal, a single drop is allowed to be less than or equal to a first drop (e.g., ≤15%), and the cumulative proportion of non-negative differences between adjacent moments is greater than or equal to the first proportion (e.g., ≥90%); in the smoothed signal, a single drop is allowed to be less than or equal to a second drop (e.g., ≤5%), and the cumulative proportion of non-negative differences is greater than or equal to the second proportion (e.g., ≥95%). If the resistance falls below the start threshold after the fluctuation, the timer is reset. When timing the original signal, if the smoothed signal drops by a third drop (e.g., ≥5%), the timer is paused (with no more than two cumulative interruptions) and restarted after growth resumes, thus defining a continuous growth trend.

[0265] In a possible implementation, determining the third duration threshold according to the equivalent resistance growth slope and the noise level includes:

[0266] The equivalent resistance growth slope (the ratio of the resistance difference between adjacent moments to the time interval) and the noise fluctuation coefficient (the ratio of the current noise resistance to the rated operating condition reference noise resistance) are calculated in real time. When the growth slope is not less than the preset slope threshold, the third duration threshold is the product of the third reference time and the noise fluctuation coefficient; when the growth slope is less than the slope threshold, the third duration threshold is the product of the fourth reference time and the noise fluctuation coefficient (when the third reference time is less than the fourth reference time, the noise fluctuation coefficient is ≥1), and the slope threshold, third reference time, and fourth reference time are determined through quench evolution simulation experiments.

[0267] The working principle and effects of the above technical solution are as follows:

[0268] This threshold is based on the maximum noise resistance value during normal operation and is no less than the critical resistance corresponding to the minimum propagation zone (MPZ). It filters out minor interference. When the equivalent resistance is below this threshold, it is considered normal noise or a self-recovering minor normal zone, and the timing is not triggered. Only when the equivalent resistance exceeds this threshold does the continuous increase timing start, effectively eliminating interference from the minor normal zone (which can recover) and avoiding misjudgments.

[0269] The final resistance threshold is set based on the resistance value corresponding to the safe hotspot temperature (e.g., 150-200K) in quench simulation experiments, and calibrated based on the resistivity-temperature characteristics of the superconducting material in its normal state and the coil structure parameters (e.g., cross-sectional area and length). Its physical meaning is that when the equivalent resistance reaches this value, it indicates that the normal range has expanded to a dangerous level that could cause the hotspot temperature to exceed the specified limit. The final resistance threshold is linked to the safe hotspot temperature and the material resistivity characteristics, directly reflecting the safety margin of the magnet. The dynamic duration parameter is calibrated through simulation experiments to align with the actual quench evolution, making the judgment logic more consistent with the physical nature and improving detection accuracy.

[0270] The third time threshold is adjusted in real time according to the growth rate of the equivalent resistance and the noise level:

[0271] Calculate the equivalent resistance growth slope (the ratio of the resistance difference between adjacent moments to the time interval) to reflect the quench development speed;

[0272] Calculate the noise fluctuation coefficient (the ratio of the current noise resistance to the rated working condition reference noise resistance) to reflect the intensity of environmental interference;

[0273] When the growth slope is greater than or equal to the preset slope threshold (rapid quench), the third duration threshold is the product of the third reference time (shorter time) and the noise fluctuation coefficient to ensure a rapid response;

[0274] When the growth slope is less than the slope threshold (slow quench), the third duration threshold is the product of the fourth reference time (longer time) and the noise fluctuation coefficient to avoid false triggering due to noise;

[0275] Among them, the slope threshold and the third / fourth reference time are determined through quench evolution simulation experiments to adapt to the quench characteristics of different magnets; a shorter third reference time is used for fast quench (large growth slope) to ensure timely triggering of protection; a longer fourth reference time is used for slow quench (small growth slope) to avoid misjudgment due to instantaneous fluctuations, thereby achieving full coverage of quench scenarios with different rates.

[0276] When the equivalent resistance exceeds the starting resistance threshold and the timer starts, a quench is determined if any of the following conditions is met:

[0277] During the timing process, the equivalent resistance exceeds the final resistance threshold before reaching the third duration threshold;

[0278] When the timing reaches the third time threshold, the equivalent resistance still exceeds the final resistance threshold.

[0279] An embodiment of the present invention further provides a superconducting magnet coil quench detection device, the device comprising:

[0280] One or more current sensors for measuring the current at any end of the superconducting magnet coil in real time;

[0281] One or more voltage sensors for measuring the total voltage across the superconducting coil in real time;

[0282] A processing unit is used to execute the steps of the superconducting magnet coil quench detection method described in an embodiment of the present invention.

[0283] The working principle and effect of the above technical solution are the same as those in the embodiment of the method of the present invention, and will not be described in detail here.

[0284] An embodiment of the present invention further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the steps of any one of the methods described in the embodiments of the present invention or the functions of the device described in the embodiments of the present invention.

[0285] An embodiment of the present invention also provides a computer-readable storage medium, which is used to store a computer program. When the computer program is executed, the steps of the method in the embodiment of the present invention are implemented. Its specific implementation method is consistent with the implementation method and the technical effect achieved in the above-mentioned method embodiment, and some contents will not be repeated here.

[0286] In the present invention, a readable storage medium can be any tangible medium that contains or stores a program that can be used by or in combination with an instruction execution system, device, or device. A program product can employ any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0287] A computer-readable storage medium may include a data signal propagated in baseband or as part of a carrier wave, carrying readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium that can transmit, propagate, or transfer a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, or any suitable combination thereof. The program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar programming languages. The program code may be executed entirely on the user computing device, partially on an associated device, as a standalone software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. Where a remote computing device is involved, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0288] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limiting the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the invention without departing from the principles and purpose of the present invention. All such changes shall fall within the scope of protection of the claims of the present invention.

Claims

1. A method for detecting a quench in a superconducting magnet coil, characterized in that: The method comprises: During the operation of the superconducting magnet, the total voltage at both ends of the superconducting magnet coil and the current at any end of the superconducting magnet coil are collected in real time; Based on a curve showing the relationship between the inductance of the superconducting magnet coil and the current, the resistive voltage in the superconducting magnet coil is determined by the total voltage across the superconducting magnet coil and the current at any one end of the superconducting magnet coil. When the resistive voltage increases to a voltage threshold, or the duration of the continuous increase of the resistive voltage exceeds a time threshold, it is determined that a quench occurs in the superconducting magnet coil.

2. The method according to claim 1, characterized in that The relationship curve of the superconducting magnet coil inductance versus current is obtained by the following method, including: Establish a curve showing the relationship between the inductance of a superconducting magnet coil and the change in current through simulation; or, When the superconducting magnet coil is in a superconducting state, during the current dynamic adjustment process, the inductance values ​​corresponding to different current values ​​are synchronously measured to establish and correct the relationship curve between the superconducting magnet coil inductance and current change; During the steady-state operation of the superconducting magnet, the relationship curve is calibrated online using the total voltage across the superconducting magnet coil and the current at any one end of the superconducting magnet coil at the current moment.

3. The method according to claim 1, characterized in that The method of determining the resistive voltage in the superconducting magnet coil according to the relationship curve between the inductance of the superconducting magnet coil and the current, through the total voltage at both ends of the superconducting magnet coil and the current at any end of the superconducting magnet coil, comprises: Based on the curve of the superconducting magnet coil inductance versus current and the real-time current acquisition, the inductance value L(t) corresponding to the current I(t) and the derivative of the inductance with respect to the current dL / dI are determined. According to the current inductance value L(t), the current current change rate and the current current I(t), the ideal inductive voltage V in the ideal superconducting state is calculated. ideal (t); in, is the current inductance change rate; is the current change rate; Determine the resistive voltage in the superconducting magnet coil by dividing the total voltage by the ideal inductive voltage.

4. The method according to claim 1, wherein The determination logic for the resistive voltage increasing to the voltage threshold includes: When the resistive voltage exceeds its starting voltage threshold, the continuous growth timer starts, and a quench is determined if any of the following conditions is met: The resistive voltage exceeds the final voltage threshold before reaching the duration threshold; When the resistive voltage reaches the duration threshold, it exceeds the final voltage threshold; The resistive voltage continuously increases for a period exceeding a time threshold; The starting voltage threshold is set based on the maximum noise peak value during normal operation; and the final voltage threshold is set based on the exponential growth characteristic of the resistive voltage.

5. The method according to claim 4, characterized in that The resistive voltage includes an original resistive voltage and a filtered resistive voltage, and the decision logic is independently executed on the original resistive voltage and the filtered resistive voltage at the same time.

6. The method according to claim 1, characterized in that The voltage threshold is an adaptive threshold that is dynamically modified based on real-time operating conditions, including: During the operation of the superconducting magnet, the ambient temperature of the coil, the pressure of the cooling medium, and the noise component of the total voltage across the coil are collected in real time. The noise component is extracted by high-frequency filtering of the total voltage signal and represented by the standard deviation of the voltage signal within a sliding window. Dynamically adjusting the voltage threshold based on a preset correction coefficient matrix and a noise component; the correction coefficient matrix includes a temperature correction term, a pressure correction term, and a noise correction term; When it is detected that the relationship curve has been calibrated online, the voltage threshold is corrected synchronously according to the calibration ratio.

7. The method according to claim 6, characterized in that The correction coefficient matrix is ​​dynamically updated in the following way: Collect non-quench abnormal data and corresponding voltage noise data of superconducting magnet coils under different operating conditions, and construct an abnormal sample set with noise characteristics; The temperature correction coefficient, pressure correction coefficient and noise correction coefficient are iteratively optimized using the gradient descent algorithm, so that the misjudgment rate of the voltage threshold in the abnormal sample set is lower than the set value, and the detection delay of the small resistive voltage in the quench simulation experiment does not exceed the preset threshold.

8. The method according to claim 1, characterized in that The duration threshold is preset as a dynamic threshold associated with the resistive voltage growth slope and the voltage noise level, including: Calculate the real-time growth slope of the resistive voltage, which is the ratio of the difference between the resistive voltage values ​​at adjacent moments to the time interval; Real-time calculation of voltage noise fluctuation coefficient, which is the ratio of the current noise component to the baseline noise value under rated operating conditions; Presetting multiple sets of slope thresholds, first reference times, and second reference times for different magnets or different operating power ranges of the same magnet; When the real-time growth slope is not less than the slope threshold, the duration threshold is set to the product of the first reference time and the noise fluctuation coefficient; when the real-time growth slope is less than the threshold, the duration threshold is set to the product of the second reference time and the noise fluctuation coefficient, and the first reference time is less than the second reference time, and the noise fluctuation coefficient is ≥1; The slope threshold, the first reference time, and the second reference time are determined through a quench evolution simulation experiment.

9. The method according to claim 1, characterized in that The method further comprises: Determine the equivalent resistance of the current superconducting magnet coil by using the resistive voltage and the current at either end of the current superconducting magnet coil; If the equivalent resistance exceeds its equivalent resistance threshold, or the equivalent resistance continuously increases for a period exceeding a time threshold and its integral value exceeds the energy integral upper limit corresponding to the safe hotspot temperature, it is determined that the superconducting magnet coil has quenched; the integral upper limit is determined based on the enthalpy characteristics of the superconducting coil material and the energy absorption upper limit at a safe temperature.

10. The method according to claim 9, characterized in that The equivalent resistance threshold includes a starting resistance threshold and a final resistance threshold; the duration threshold is a third duration threshold, and a quench is determined when any of the following conditions is met: After the equivalent resistance exceeds the starting resistance threshold, the continuous increase timing is started, and the equivalent resistance exceeds the final resistance threshold before reaching the third time threshold; or, the equivalent resistance exceeds the final resistance threshold when reaching the third time threshold; Among them, the starting resistance threshold is set according to the resistance value corresponding to the maximum noise during normal operation, and the starting resistance threshold is not less than the critical resistance corresponding to the minimum propagation area; the final resistance threshold is set according to the resistance value corresponding to the safe hotspot temperature in the quench simulation experiment, and is calibrated according to the resistivity change characteristics of the superconducting material in the normal state with temperature and the coil structure parameters; the third time threshold is determined according to the equivalent resistance growth slope and the noise level.

11. A superconducting magnet coil quench detection device, characterized in that: The device comprises: One or more current sensors for measuring the current at any end of the superconducting magnet coil in real time; One or more voltage sensors for measuring the total voltage across the superconducting coil in real time; A processing unit, configured to execute the steps of the superconducting magnet coil quench detection method according to any one of claims 1 to 10.

12. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method according to any one of claims 1 to 10 or the functions of the device according to claim 11 when executing the computer program.

13. A computer-readable storage medium, characterized in that The storage medium stores computer instructions. When a computer reads the computer instructions, the computer executes the steps of the method according to any one of claims 1 to 10.

Images