Method for automatic fault repair based on memory unit architecture and related device

Abstract

The application relates to the technical field of intelligent storage, and provides an automatic fault repairing method based on a storage unit architecture and related equipment. A dynamic defect entropy of an internal defect accumulation degree of a storage medium is calculated by collecting defect-sensitive electrical parameters of a storage unit; when the dynamic defect entropy exceeds a preset repairing threshold, a voltage amplitude and a pulse width of a repairing pulse are adaptively generated; a repairing excitation matched with a storage unit type is applied to the storage unit according to the voltage amplitude and the pulse width of the repairing pulse, so as to drive atoms, dipoles or magnetic moments in the storage medium to be rearranged; defect-sensitive electrical parameters of the same address unit in the storage unit after repairing are collected, the dynamic defect entropy after repairing is recalculated, and if the repairing is invalid, the repairing parameters are strengthened and the repairing is repeatedly executed. Active sensing of internal metastable defects of the storage medium and material-level lattice reconstruction repairing are realized, so that the effective service life of the storage unit is significantly prolonged under the condition of radiation accumulation.

Description

Technical Field

[0001] This application relates to the field of intelligent storage technology, and more specifically, to an automatic fault repair method and related equipment based on a storage unit architecture. Background Technology

[0002] Spaceborne storage devices are core components of spacecraft electronic systems, facing continuous threats from total dose effects and single-event effects in the space radiation environment. Existing radiation-hardened storage technologies mainly develop along four lines: hardware hardening, circuit redundancy, software fault tolerance, and physical shielding. These technologies employ radiation-hardened processes, error-correcting coding, redundant backups, data refresh, and metallic shielding to detect, correct, and avoid data errors that have already occurred. However, current technologies generally overlook the potential metastable defects generated within the storage medium during radiation accumulation, such as ferroelectric layer charge traps, interface dangling bonds, two-dimensional material vacancy defects, or amorphous magnetic moment disturbances. These defects do not immediately lead to bit flips in the early stages but rather gradually alter intrinsic electrical characteristics such as threshold voltage, subthreshold swing, and write efficiency, ultimately causing irreversible performance degradation.

[0003] Therefore, how to sense the accumulation level of metastable defects inside the storage medium in real time and proactively implement intrinsic material-level repair before the defects develop to an irreversible degree, thereby maintaining the intrinsic performance of the storage cell and extending the effective on-orbit lifespan of the storage device under conditions of continuous radiation dose accumulation, is a technical problem that urgently needs to be solved in this field. Currently, there is no feasible technology in the field capable of dynamically monitoring the evolution of defects inside the storage medium and performing closed-loop lattice reconstruction repair. Summary of the Invention

[0004] This application provides an automatic fault repair method and related equipment based on a storage cell architecture, which can at least partially solve the problem of dynamically monitoring the evolution of internal defects in storage media and performing closed-loop lattice reconstruction repair.

[0005] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.

[0006] According to one aspect of this application, an automatic fault repair method based on a memory cell architecture is provided, comprising: acquiring defect-sensitive electrical parameters of the memory cell, the defect-sensitive electrical parameters including threshold voltage drift, subthreshold swing degradation, and impulse response attenuation coefficient; calculating a dynamic defect entropy characterizing the degree of defect accumulation within the memory medium based on the defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature; when the dynamic defect entropy exceeds a preset repair threshold, adaptively generating a repair pulse with voltage amplitude and pulse width based on the dynamic defect entropy; applying a repair excitation matching the memory cell type to the memory cell based on the voltage amplitude and pulse width of the repair pulse to drive the atoms, dipoles, or magnetic moments in the memory medium to rearrange; acquiring defect-sensitive electrical parameters of the same address cell in the repaired memory cell, recalculating the repaired dynamic defect entropy, and judging the repair effect based on the comparison results of the repaired dynamic defect entropy with multiple preset thresholds; if the repair is ineffective, strengthening the repair parameters and repeating the repair until the maximum number of repairs is reached, after which it is marked as permanently failed.

[0007] In this application, based on the aforementioned scheme, the step of calculating the dynamic defect entropy, which characterizes the degree of defect accumulation inside the storage medium, according to the defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature, includes: normalizing the defect-sensitive electrical parameters to obtain a normalized parameter vector; constructing a hyperbolic tangent function based on the product of the normalized parameter vector, the normalized cumulative radiation dose, and the normalized chip junction temperature, and introducing a natural logarithm operation to generate a defect potential function; inputting the defect potential function into a sigmoid function and a sinusoidal square modulation function respectively, and multiplying the outputs of the two to obtain the dynamic defect entropy.

[0008] In this application, based on the aforementioned scheme, the acquisition of defect-sensitive electrical parameters of the memory cell includes: applying a linearly increasing gate voltage to the memory cell multiple times and reading the channel current abrupt change point; comparing the median of the multiple measurement results with the factory nominal value to obtain the threshold voltage drift; measuring the gate voltage change required for the leakage current to change from a first preset ratio to a second preset ratio in the subthreshold region; comparing the median of the multiple measurement results with the initial subthreshold swing value to obtain the subthreshold swing degradation; and measuring the channel conductance after applying a standard write pulse to the memory cell, and obtaining the pulse response attenuation coefficient based on the statistical data of the multiple measurement results.

[0009] In this application, based on the aforementioned scheme, the step of adaptively generating the voltage amplitude and pulse width of the repair pulse according to the dynamic defect entropy when the dynamic defect entropy exceeds a preset repair threshold includes: when the dynamic defect entropy exceeds the preset repair threshold, reading the difference between the dynamic defect entropy and the preset repair threshold; dynamically adjusting the voltage multiplier using a hyperbolic tangent function based on the difference, and then multiplying the voltage multiplier by a reference voltage to obtain the voltage amplitude of the repair pulse, wherein the voltage multiplier saturates as the difference increases; dynamically adjusting the pulse width multiplier using a natural exponential function based on the square of the difference, and multiplying the pulse width multiplier by a reference pulse width to obtain the pulse width of the repair pulse, wherein the pulse width multiplier increases exponentially as the difference increases.

[0010] In this application, based on the aforementioned scheme, the step of applying a repair excitation to the memory cell that matches the memory cell type according to the voltage amplitude and pulse width of the repair pulse includes: when the memory cell is a ferroelectric field-effect transistor, applying a reverse polarized triangular wave and dynamically adjusting the falling edge slope of the triangular wave according to the ferroelectric domain wall migration rate; when the memory cell is a magnetoelectric random access memory, applying a double exponentially decaying oscillating magnetic field pulse corresponding to the equivalent voltage amplitude, and the number of oscillations being dynamically determined by the dynamic defect entropy; when the memory cell is a two-dimensional material molybdenum disulfide, applying a low-voltage pulse sequence that linearly increases from a preset proportional voltage to the voltage amplitude of the repair pulse to excite lattice phonon modes.

[0011] In this application, based on the aforementioned scheme, the defect-sensitive electrical parameters of the same address unit in the repaired storage unit are collected, the dynamic defect entropy after repair is recalculated, and the repair effect is judged based on the comparison results of the dynamic defect entropy after repair with multiple preset thresholds. This includes: collecting defect-sensitive electrical parameters of the same address unit in the repaired storage unit and recalculating the dynamic defect entropy after repair; if the dynamic defect entropy after repair is less than or equal to a first threshold, the repair is determined to be successful and the health status table is updated; if the dynamic defect entropy after repair is between the first threshold and the preset repair threshold, the repair is determined to be partially effective, and the current repair parameters are recorded for subsequent adaptive learning; if the dynamic defect entropy after repair is greater than or equal to the preset repair threshold, the repair is determined to be invalid, and the width of the repair pulse is increased by a predetermined step size and the repair operation is repeated until the maximum number of repairs is reached, after which the storage unit is marked as permanently failed and the backup block mapping is started.

[0012] In this application, based on the aforementioned scheme, the value range of the dynamic defect entropy is configured to be 0 to 1.25, the preset repair threshold is configured to be 0.75, and an automatic fault repair process is triggered when the dynamic defect entropy exceeds 0.75.

[0013] In this application, based on the aforementioned scheme, the statistical data of the multiple measurement results includes the median value of eight measurement results.

[0014] According to one aspect of this application, an automatic fault repair apparatus based on a storage cell architecture is provided, comprising: The acquisition unit is used to acquire defect-sensitive electrical parameters of the storage unit, including threshold voltage drift, subthreshold swing degradation, and impulse response attenuation coefficient. The calculation unit is used to calculate the dynamic defect entropy, which characterizes the degree of defect accumulation inside the storage medium, based on the defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature. The parameter tuning unit is used to adaptively generate the voltage amplitude and pulse width of the repair pulse based on the dynamic defect entropy when the dynamic defect entropy exceeds the preset repair threshold. A driving unit is used to apply a repair excitation matching the type of the storage cell to the storage cell according to the voltage amplitude and pulse width of the repair pulse, so as to drive the atoms, dipoles or magnetic moments in the storage medium to rearrange. The feedback unit is used to collect the defect-sensitive electrical parameters of the same address unit in the repaired storage unit, recalculate the dynamic defect entropy after repair, and judge the repair effect based on the comparison results of the dynamic defect entropy after repair and multiple preset thresholds. If the repair is ineffective, the repair parameters are strengthened and the repair is repeated until the maximum number of repairs is reached and then it is marked as permanently failed. Based on the defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature, the dynamic defect entropy, which characterizes the degree of defect accumulation within the storage medium, is calculated. This includes: normalizing the defect-sensitive electrical parameters to obtain a normalized parameter vector; constructing a hyperbolic tangent function based on the product of the normalized parameter vector, the normalized cumulative radiation dose, and the normalized chip junction temperature, and introducing a natural logarithm operation to generate a defect potential function; inputting the defect potential function into a sigmoid function and a sinusoidal square modulation function respectively, and multiplying their outputs to obtain the dynamic defect entropy.

[0015] According to one aspect of this application, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the automatic fault repair method based on a storage cell architecture as described in the above embodiments.

[0016] According to one aspect of this application, an electronic device is provided, comprising: one or more processors; and a storage device for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement the automatic fault repair method based on the storage cell architecture as described in the above embodiments.

[0017] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the automatic fault repair method based on a storage cell architecture provided in the various optional implementations described above.

[0018] The main differences and technical effects of the technical solution of this application compared with the prior art are as follows: Existing radiation-hardened storage technologies primarily focus on passively correcting and replacing explicit data errors, lacking the ability to proactively detect and intervene in potential metastable defects during radiation accumulation. This scheme achieves direct quantitative sensing of the lattice disorder and trap density within the storage medium by collecting three defect-sensitive electrical parameters: threshold voltage drift, subthreshold swing degradation, and impulse response attenuation coefficient. Furthermore, it integrates radiation dose and temperature information to construct a dynamic defect entropy, enabling accurate identification of metastable defects before they reach irreversible levels. This elevates the radiation-hardened strategy from post-error correction and replacement to proactive sensing and lattice reconstruction repair during defect accumulation. Through defect entropy-driven adaptive repair excitation generation, the voltage amplitude and pulse width of the repair pulse can be dynamically adjusted according to the severity of the defect. Saturation control of the voltage avoids overstress damage, while exponential pulse width growth provides a sufficient repair window for deep defects. These two aspects work together to maximize repair effectiveness within the safety boundary. The closed-loop verification and hierarchical decision-making mechanism after repair further ensures the effectiveness and traceability of the repair. Parameters of some effective repairs are recorded for subsequent adaptive learning, while invalid repairs are reinforced and retried until the spare block is finally activated. This scheme improves radiation resistance from the traditional approach of tolerating errors to repairing intrinsic defects in materials, significantly extending the effective service life of storage cells under cumulative radiation doses.

[0019] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0020] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0021] Figure 1The flowchart illustrating an automatic fault repair method based on a storage cell architecture in one embodiment of this application is shown.

[0022] Figure 2 A flowchart illustrating the calculation of dynamic defect entropy is shown in one embodiment of this application.

[0023] Figure 3 The illustration shows a schematic diagram of an automatic fault repair device based on a storage cell architecture in one embodiment of this application.

[0024] Figure 4 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown. Detailed Implementation

[0025] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art.

[0026] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.

[0027] It should be noted that the data acquisition or information collection in this embodiment is performed after authorization by the user or the object of collection, and its process and purpose strictly follow the relevant regulations.

[0028] The block diagrams shown in the attached figures are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, or in one or more hardware modules composed of smart chips, smart integrated circuits, or application-specific integrated circuits (ASICs), or in different network and / or processor devices and / or microcontroller devices.

[0029] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0030] The implementation details of the technical solution of this application are described below: Figure 1 A flowchart illustrating an automatic fault recovery method based on a storage cell architecture according to an embodiment of this application is shown. (Refer to...) Figure 1 As shown, this automatic fault repair method based on storage cell architecture includes at least steps S110 to S150, which are described in detail below: S110, acquire the defect-sensitive electrical parameters of the storage unit, including threshold voltage drift, subthreshold swing degradation, and impulse response attenuation coefficient.

[0031] In this embodiment, the chip's built-in microcontroller unit first obtains the address of the target memory block to be detected from the task queue of the defect detection and management module. Simultaneously, the microcontroller unit initiates a data read request to both the on-chip temperature sensor and the onboard radiation monitoring unit via the on-chip integrated bus. The on-chip temperature sensor returns a digital reading of the current chip junction temperature, and the radiation monitoring unit returns the current spatial equivalent total dose rate. These two data points are temporarily stored in the microcontroller unit's register, serving as the basis for temperature and dose rate compensation of the measurement reference during subsequent electrical parameter measurements.

[0032] In one embodiment of this application, collecting defect-sensitive electrical parameters of the storage cell includes: The threshold voltage drift is obtained by repeatedly applying a linearly increasing gate voltage to the memory cell and reading the channel current abrupt change point. The median of the multiple measurement results is then compared with the factory nominal value. The gate voltage change required for the leakage current to change from a first preset ratio to a second preset ratio is measured in the subthreshold region. The median of multiple measurement results is then compared with the initial subthreshold swing value to obtain the subthreshold swing degradation amount. After applying a standard write pulse to the memory cell, the channel conductance is measured, and the pulse response attenuation coefficient is obtained based on the statistical data of multiple measurement results.

[0033] The microcontroller unit (MCU) selects the corresponding physical memory cell based on the address of the target memory block using the row decoder and column selector of the memory array. After confirming that the memory cell is isolated and is not currently undergoing read, write, or erase operations, the MCU first performs a threshold voltage drift measurement. Specifically, the MCU controls a built-in programmable gate voltage generator to apply a linearly increasing voltage signal to the gate of the memory cell, starting from a value below the normal threshold voltage, in very small steps. After each voltage application, the MCU immediately reads the channel current value through a sensitive amplifier and compares it with a preset current surge criterion. When the channel current first reaches and exceeds the criterion, the MCU records the current gate voltage value. Optionally, this measurement process is repeated eight times, with a brief reset of the memory cell after each measurement to eliminate charge accumulation effects. The results of the eight measurements are fed into an on-chip median filter circuit, which sorts the eight values ​​using a hardware comparator and selects the median value, effectively eliminating abnormal jumps caused by transient noise or particle impacts. The output of the median filter circuit is compared with the standard threshold voltage stored in the chip's one-time programmable memory at the factory. The difference between the two is the threshold voltage drift. This drift directly reflects the pinning effect of radiation-induced defect dipoles on the polarization reversal barrier in the ferroelectric layer.

[0034] After measuring the threshold voltage drift, the microcontroller immediately proceeds to measure the subthreshold swing degradation. The microcontroller biases the memory cell's operating area into the subthreshold region, where the gate voltage is slightly below the turn-on voltage. Then, the microcontroller restarts the programmable gate voltage generator, gradually increasing the gate voltage in finer steps within the subthreshold voltage range. Under each gate voltage level, the microcontroller precisely measures the leakage current using a high-gain current-sensing amplifier and calculates the logarithm of the leakage current in real time. When the leakage current reaches 90% of the maximum leakage current (10%), the microcontroller records the total change in gate voltage during this process. Optionally, similar to the threshold voltage measurement, the subthreshold swing measurement is repeated eight times, resetting the memory cell's charge state between each measurement. The results of the eight measurements are then processed by a median filter circuit to obtain the current subthreshold swing value. The microcontroller reads the initial subthreshold swing value calibrated at the factory from the read-only memory and outputs the ratio of the current value to the initial value as the subthreshold swing degradation. This ratio characterizes the radiation-induced increase in interface state density, which exacerbates trap-assisted tunneling of channel carriers, causing the same gate voltage change to cause larger leakage current fluctuations, thus manifesting as a degradation of the subthreshold swing.

[0035] Next, the microcontroller unit (MCU) acquires the pulse response decay coefficient. The MCU first applies a standard write pulse with a precisely controllable width, generated by a precision pulse generator, to the memory cell via the write drive circuit. After the pulse ends, the MCU immediately suspends all access operations to the same memory array, allowing a brief relaxation time for the ferroelectric domains or charge traps. After the relaxation time, the MCU measures the conductance of the memory cell channel via the conductance detection module in the read / write path. The magnitude of this conductance directly depends on the amount of polarization charge flipped in the ferroelectric layer by the write pulse or the charge density trapped in the memory layer. Optionally, similar to the two measurements mentioned above, the conductance measurement is repeated eight times, ensuring no additional read / write interference between the eight measurements. The eight results are again fed into a median filter, outputting the current stable post-write channel conductance value. The MCU then reads the initial channel conductance value recorded at the factory and sends the ratio of the current conductance value to the initial conductance value to the logarithmic calculation unit. The logarithmic calculation unit uses a hardware-implemented natural logarithm lookup table to convert this ratio into a value in the logarithmic domain, namely the impulse response decay coefficient. The reason for taking the logarithm is that the decrease in radiation-induced charge trapping efficiency is physically approximately exponential. Taking the logarithm maps this exponential decay to an approximately linear degradation measure, which facilitates unified processing by the subsequent defect entropy calculation module.

[0036] After completing all three measurements, the microcontroller combines the threshold voltage drift, subthreshold swing degradation, and impulse response attenuation coefficient into a parameter vector according to a fixed data format. This parameter vector, along with the temperature and radiation dose rate readings acquired at the start of S110, is written into the parameter buffer within the defect sensing module. The microcontroller updates the status flag of this storage block in the parameter buffer, indicating that the block has completed data acquisition for S110, and then transfers control to the defect entropy calculation module of S120. Throughout the execution of S110, the microcontroller continuously monitors the status of each measurement operation. If any measurement timeout or data anomaly occurs, the current process is immediately interrupted and an error status is reported. Simultaneously, the storage block is marked as requiring retesting or entering isolation observation mode.

[0037] The above process, by collecting three defect-sensitive electrical parameters—threshold voltage drift, subthreshold swing degradation, and impulse response decay coefficient—achieves multi-dimensional and highly sensitive sensing of radiation-induced defects within the storage medium. Threshold voltage drift directly reflects the degree of fixed charge accumulation due to radiation in the ferroelectric layer or charge-trapping layer, which is the core degradation characteristic of the total dose effect. Subthreshold swing degradation reveals the increase in interface state density; interface states are the direct sites where charge carriers are trapped and released, and their density changes directly affect the switching characteristics of transistors. The impulse response decay coefficient quantifies the decrease in write efficiency, comprehensively reflecting the degradation of polarization charge reversal capability or charge trapping efficiency. The combined acquisition of these three parameters avoids the limitations of a single parameter, ensuring that no aspect of defect accumulation is overlooked. The operation of taking the median of multiple measurements effectively eliminates transient noise interference from space particle impacts, ensuring the stability and reliability of the acquired data and laying a reliable raw data foundation for the subsequent accurate calculation of defect entropy.

[0038] S120, calculate the dynamic defect entropy, which characterizes the degree of defect accumulation inside the storage medium, based on the defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature.

[0039] In one embodiment of this application, after executing S110 and obtaining the parameter vector, cumulative radiation dose, and chip junction temperature, these three sets of data are used as inputs to S120 to begin the numerical calculation process of dynamic defect entropy. S120 plays a core computational role in the entire automatic fault repair method. The complete processing procedure of S120 is described in detail below.

[0040] like Figure 2 As shown, in one embodiment of this application, S120 calculates the dynamic defect entropy, which characterizes the degree of defect accumulation inside the storage medium, based on the defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature, including S210~S230: S210, The defect-sensitive electrical parameters are normalized to obtain a normalized parameter vector; S220, based on the product of the normalized parameter vector, the normalized cumulative radiation dose, and the normalized chip junction temperature, a hyperbolic tangent function is constructed, and a natural logarithm operation is introduced to generate a defect potential function. S230, the defect potential function is input into the Sigmoid function and the sinusoidal square modulation function respectively, and the output results of the two are multiplied to obtain the dynamic defect entropy.

[0041] When S120 starts up, the defect entropy calculation module inside the chip first reads the parameter vector output by S110 from the parameter buffer. This vector contains three electrical parameters: threshold voltage drift, subthreshold swing degradation, and impulse response decay coefficient. Since these three parameters are completely different in physical dimensions and numerical ranges, the first step of the calculation module is to normalize each parameter, mapping them all to a dimensionless range between 0 and 1. The normalization is based on the theoretical maximum values ​​of each parameter obtained through radiation acceleration experiments before the chip leaves the factory. The maximum value of the threshold voltage drift corresponds to the saturation drift value when the ferroelectric layer is completely degraded; the maximum value of the subthreshold swing degradation corresponds to the ratio when the interface state density reaches a critical value; and the maximum value of the impulse response decay coefficient corresponds to the logarithmic decay value when the write capability is almost completely lost. The calculation module divides the current measured value of each parameter by its corresponding calibrated maximum value to obtain three normalized defect sensitivity factors, denoted as the first factor, the second factor, and the third factor. These three factors together constitute the normalized parameter vector, serving as the direct input for subsequent defect potential function calculations.

[0042] After normalization, the process moves to solving the defect potential function. The defect potential function is an intermediate variable used to fuse three independent electrical degradation parameters and environmental stress parameters into a comprehensive potential energy index. The calculation module first calculates the square of the first factor. This square operation amplifies the contribution of the threshold voltage drift to the defect potential function, reflecting the threshold accumulation characteristic of the radiation-induced polarization charge pinning effect—that is, the drift is slow in the early stages but accelerates as it approaches the failure boundary. Then, the calculation module calculates the second power of the natural constant e. This exponential operation causes the contribution of subthreshold swing degradation to the defect potential function to increase rapidly after reaching a moderate level, simulating the physical process of carrier trap-assisted tunneling sharply enhancing after the interface state density reaches a certain level. Next, the calculation module multiplies the square of the first factor by the second power of e. This product realizes the second-order coupling between threshold voltage drift and subthreshold swing degradation; that is, this product only increases significantly when both exhibit significant degradation simultaneously, consistent with the actual mechanism of space radiation simultaneously damaging ferroelectric layer polarization and interface states.

[0043] After obtaining the above product, the calculation module divides it by the sum of a third factor. The third factor represents the logarithmic decay of the write pulse response; a larger value indicates lower write efficiency. The division operation makes the denominator of the defect potential function smaller when the write response decays severely, increasing the overall ratio and thus incorporating the degradation of the memory cell's write capability into the defect potential evaluation. At this point, the calculation module has obtained a ratio result that integrates the three electrical parameters. The cumulative total dose value currently experienced by the chip is read from the radiation monitoring register, and the current absolute temperature value (in Kelvin) is read from the temperature register. The cumulative total dose value is divided by the reference total dose of 100 kL silicon preset at the chip's factory to obtain the normalized radiation dose factor. The absolute temperature value is divided by the reference temperature of 300 Kelvin to obtain the normalized temperature factor. These two factors are multiplied to form a joint acceleration factor of radiation and temperature, which is fed into a hyperbolic tangent function, compressing the output range to [-1, 1] and monotonically saturating. This design ensures that when the cumulative radiation dose is much less than the reference dose or the temperature is much lower than the reference temperature, the combined acceleration factor approaches 0, and the defect potential function is almost unaffected by the environment. When the combined effect exceeds a certain level, the hyperbolic tangent function output rapidly approaches one, indicating that radiation and temperature have fully driven the defect evolution, the environmental stress factor reaches saturation, and the defect potential function is prevented from growing indefinitely.

[0044] The ratio of the three combined electrical parameters is multiplied by the output of the hyperbolic tangent function, the product is then added to the number 1, and the natural logarithm of the whole is taken. This natural logarithmic operation compresses the dynamic range, preventing excessively rapid subsequent saturation, and simultaneously uses the result in the logarithmic domain as the output value of the defect potential function. for:

[0045] in, These represent the normalized parameters obtained by normalizing the threshold voltage drift, subthreshold swing degradation, and impulse response attenuation coefficient, respectively; D represents the total ionizing dose accumulated by the real-time reading of the spaceborne radiation monitoring unit, in units of krad(Si). =100krad(Si) represents the reference total dose, corresponding to the upper limit of the cumulative dose during the typical design life of a commercial satellite, and is used as a normalization benchmark; T represents the real-time read chip junction temperature. This represents the reference temperature, which can be 300K, corresponding to a room temperature of approximately 27°C, and serves as the normalization baseline. According to the engineering design, the output value of this defect potential function... The value ranges from 0 to 1.2. The higher the value, the greater the lattice disorder and trap density inside the storage medium.

[0046] The dynamic defect entropy is obtained by multiplying two sub-terms. The first sub-term is the standard logistic function, i.e., the Sigmoid function, with the difference between the defect potential function and the defect potential threshold as the independent variable. The calculation module first reads the steepness factor (default 12.5) and the defect potential threshold (default 0.38) from the configuration register. The calculation module subtracts the defect potential threshold from the defect potential function value to obtain the difference. If the difference is negative, it indicates that the defect level is still within a safe range; if it is positive, it indicates that the threshold has been exceeded. The calculation module multiplies the steepness factor by the difference, takes the negative sign, and uses it as the exponent of the natural constant e. It calculates e raised to the power of this exponent, and then adds 1 to the exponent as the denominator to calculate the first sub-term. The output value of this sub-term varies continuously between 0 and 1. When the defect potential is much lower than the threshold, it approaches 0; when the defect potential is much higher than the threshold, it approaches 1. A steep but continuous transition band is formed near the threshold to make a gradual judgment on the ambiguous state between health and failure, rather than a hard decision of black and white.

[0047] The second sub-term is one plus a sinusoidal squared modulation term. First, the saturation potential (default 1.2) and nonlinear modulation depth (default 0.25) are read from the configuration register. The defect potential function value is divided by the saturation potential value, and the ratio is multiplied by half of pi. The product is then fed into the independent variable of the sine function. After the sine function calculation is complete, the result is squared to obtain an oscillation value between 0 and 1. This oscillation value is multiplied by the modulation depth (default 0.25) and added to the number 1 to obtain the second sub-term. As the defect potential function value gradually increases from 0 to the saturation potential value, the sinusoidal squared term exhibits periodic fluctuations, causing the second sub-term to oscillate between 0.75 and 1.25. This oscillation modulation simulates the metastable oscillation behavior of lattice defects between local relaxation and stress accumulation in actual physical processes. That is, defect accumulation is not monotonically smooth; sometimes, due to local lattice micro-reconstruction, a brief performance recovery illusion occurs, followed by a relapse.

[0048] The final calculation result of the dynamic defect entropy is the product of the first and second sub-terms. The first sub-term provides a probabilistic judgment on whether the defect exceeds the repair threshold, while the second sub-term superimposes the physical oscillations and fluctuations onto the probabilistic judgment. The product of the two yields the dynamic defect entropy value. for:

[0049] in, These represent the value of the defect potential function and the defect potential threshold, respectively. The defect potential threshold is obtained by fitting a large amount of radiation acceleration experimental data and can be 0.38. This represents the saturation potential, corresponding to the defect potential. The upper limit reference value is 1.2; k represents the steepness factor of the Sigmoid function, which can be 12.5. This represents the nonlinear modulation depth, and can take a value of 0.25. Through the above calculation process, the value range of dynamic defect entropy is compressed to between 0 and 1.25. When the value is below 0.75, it is determined that the storage unit is still in a healthy or acceptable degradation range, and no repair process needs to be initiated. When the value exceeds 0.75, it is determined that the storage unit has entered a state that requires active repair, triggering the repair flag. The dynamic defect entropy value output by S120 is written into the defect entropy register, and the repair flag is set or cleared to 0. After S120 ends, control is transferred to the adaptive repair excitation parameter generation module of S130. If the repair flag is true, repair pulse parameters are further generated based on the specific value of the dynamic defect entropy.

[0050] The above process integrates discrete, dimensionlessly variable electrical parameters and environmental stress factors into a continuous, monotonic, and physically meaningful dynamic defect entropy. Normalization eliminates the dimensional differences between parameters, enabling nonlinear combinations of the three at the same scale. The defect potential function amplifies the accelerated degradation effect of threshold voltage drift through squaring, captures the initially slow-then-rapid characteristic of subthreshold swing degradation through exponential operations, introduces a penalty factor for write response decay through division, and incorporates the combined acceleration effect of radiation dose and temperature into the evaluation through a hyperbolic tangent function. Finally, the dynamic range is compressed using the natural logarithm, ensuring the potential function maintains good sensitivity throughout the degradation process. Furthermore, the sigmoid function maps the defect potential to a probabilistic activation degree of defect severity, avoiding edge jitter caused by hard decision-making; sinusoidal square modulation introduces the simulation of local relaxation and metastable oscillation behavior during lattice defect accumulation, allowing the entropy value to reflect the real physical process of performance briefly recovering before continuing to deteriorate. The final output of dynamic defect entropy can not only distinguish between the three state ranges of health, early warning, and repair, but also provide fine-grained and continuous guidance for adjusting the intensity of subsequent repair incentives.

[0051] S130, when the dynamic defect entropy exceeds the preset repair threshold, the voltage amplitude and pulse width of the repair pulse are adaptively generated according to the dynamic defect entropy.

[0052] In this embodiment, when the dynamic defect entropy calculated by S120 exceeds the repair threshold of 0.75 and triggers the repair flag, S130 is executed to adaptively generate repair stimulus parameters. The input of S130 includes the dynamic defect entropy value output by S120 and the type identifier of the current memory cell to be repaired read from the chip configuration register. This type identifier is a one-time information burned into the chip at the factory according to the physical structure of the memory array, used to distinguish ferroelectric field-effect transistors, magnetoelectric random access memory amorphous memory cells, and two-dimensional material molybdenum disulfide memory cells.

[0053] In one embodiment of this application, the adaptive repair parameter generation module inside the chip first reads the value in the dynamic defect entropy register and confirms that the value has undergone normalization and nonlinear transformation processing in S120, and its value ranges from 0.75 to 1.25, where 0.75 represents the critical state that just triggers repair, and 1.25 represents an extremely severe defect accumulation state. Since S130 will not be triggered when the dynamic defect entropy is below 0.75, the entropy value entering this step must be greater than or equal to 0.75 to ensure that the repair operation is only performed when necessary, avoiding unnecessary stress on healthy memory cells.

[0054] When calculating the voltage amplitude of the repair pulse, two preset parameters are read from the read-only configuration register: the normal write voltage value and the reference multiplier required for voltage amplitude calculation. Based on the physical characteristics of ferroelectric field-effect transistors, the normal write voltage is the standard programming voltage calibrated at the chip's factory, used to complete polarization reversal in a healthy state. To drive the reorientation of ferroelectric domains that have fallen into a defect pinning state after radiation, the repair voltage needs to be increased based on the normal write voltage, but the increase cannot be infinitely large to avoid breaking down the gate oxide layer. 1.2 times the normal write voltage is used as the reference voltage value. First, the difference between the dynamic defect entropy and the 0.75 threshold is calculated. This difference reflects the margin that the current defect severity exceeds the repair trigger point. A positive difference, and a larger value, indicates a more severe defect. The module multiplies the difference by a magnification factor (which can be 2.5) to obtain a scaled difference. This scaling factor is used to adjust the sensitivity of the subsequent hyperbolic tangent function. The value of 2.5 ensures that when the entropy changes from 0.75 to 1.25, the scaled difference changes from 0 to 1.25, which precisely covers the nonlinear range of the hyperbolic tangent function from 0 to near saturation.

[0055] The generation module then calculates the hyperbolic tangent function, using the scaled difference as the input variable. The hyperbolic tangent function is an odd-symmetric sigmoid saturation function; it outputs 0 when the independent variable is 0, monotonically increases from 0 and gradually approaches 1 when the independent variable is positive and continuously increases, and outputs negative when the independent variable is negative. Since the scaled difference is always non-negative in this step, the hyperbolic tangent output is always between 0 and 1. For example, when the defect entropy is exactly at the critical value of 0.75, the scaled difference is 0, the hyperbolic tangent output is 0, and the voltage adjustment factor is 1 plus (0.4 multiplied by 0), resulting in 1. The repair voltage is equal to 1.2 times the normal write voltage of the reference voltage. When the defect entropy reaches a severe level of 1.25, the scaled difference also reaches 1.25, and the hyperbolic tangent output approaches 0.99 or higher. At this point, the voltage adjustment coefficient approaches 1 plus 0.4, resulting in approximately 1.4. The repair voltage is approximately 1.4 multiplied by 1.2 times the normal write voltage, reaching 1.38 times the normal write voltage. The coefficient 0.4 limits the maximum voltage increase, constraining the upper limit of the repair voltage to within 1.38 times the normal write voltage, thus preventing excessively high electric fields from causing irreversible dielectric breakdown of the memory cells. The saturation characteristics of the hyperbolic tangent function itself ensure that when the defect entropy further increases within the severe range, the repair voltage does not rise indefinitely but gradually approaches the upper limit, achieving adaptive saturation control of the voltage amplitude.

[0056] The generation module multiplies the reference voltage value by the calculated adjustment coefficient to obtain the final repair pulse voltage amplitude. for:

[0057] in, The reference voltage is represented by 0.4; the maximum adjustment range of the control voltage amplitude relative to the reference value is represented by 2.5; the amplification factor is represented by 2.5; and the preset repair trigger threshold value is represented by 0.75. The calculated repair pulse voltage amplitude is temporarily stored in the repair parameter register. When the voltage is slightly above 0.75, the voltage is increased slightly; when the defect is severe, the voltage saturates at 1.4 times, but the pulse width increases exponentially to provide more energy to drive the reconstruction of deep defects.

[0058] An exponential growth model is used when calculating the width of the repair pulse. First, the difference between the dynamic defect entropy and the 0.75 threshold is calculated. This difference is the same value used in the voltage calculation, ensuring that both parameters are adjusted synchronously based on the same defect severity. The module squares the difference and multiplies it by 0.8 to obtain the exponential part of the exponential function. The squaring operation ensures that when the defect entropy is slightly higher than the threshold, the difference is small, and even smaller after squaring, resulting in extremely slow width growth, almost remaining near the baseline width. When the defect entropy increases significantly, the square of the difference rapidly amplifies, accelerating the exponential growth. 0.8 is used as a scaling factor to control the sensitivity of the width to entropy growth.

[0059] The natural constant e is then raised to the power of this exponent to obtain the pulse width amplification factor. The exponential function of e exhibits unbounded growth, complementing the saturated growth of the voltage amplitude. For example, when the defect entropy is at the critical value of 0.75, the difference is 0, the exponent is 0, the amplification factor is 1, and the repair pulse width equals the base width of 100 nanoseconds. When the defect entropy reaches 1.0, the difference is 0.25, the square is 0.0625, multiplied by 0.8, approximately equal to 0.05, the exponential function output is approximately 1.05, and the width only slightly increases to 105 nanoseconds. When the defect entropy reaches a severe level of 1.25, the difference is 0.5, the square is 0.25, multiplied by 0.8, equal to 0.2, the exponential function output is approximately 1.22, and the width increases to 122 nanoseconds. It can be seen that even in the most severe case, the increase in pulse width is far less than the increase in voltage amplitude. This design follows the physical repair principle of ferroelectric field-effect transistors: excessively high voltages are more likely to induce arcing or breakdown, thus requiring saturation limiting; while appropriately extending the repair time can provide more migration time for ferroelectric domains without increasing the peak electric field, helping to overcome the constraints of deep-level pinning points. The choice of exponential rather than nonlinear growth ensures that the repair window hardly increases when the defect is not severe, avoiding unnecessary repair time, and only significantly extends the repair window when the defect is severe.

[0060] The generation module reads the baseline width value, multiplies it by the width amplification factor calculated above, and obtains the final repair pulse width. for:

[0061] in, This represents the base pulse width, 0.8 represents the exponential scaling factor, and 0.75 represents the trigger threshold. This calculation result is also written to the repair parameter register.

[0062] After calculating the repair pulse voltage amplitude and repair pulse width, the waveform shape selection stage begins. This step relies on the second parameter input from S130 (memory cell type identifier). The generation module reads this identifier and makes a judgment. For ferroelectric field-effect transistors, the waveform shape parameter is configured as a reverse polarization triangular wave. The reverse polarization triangular wave is characterized by a voltage that rises linearly from 0 to a positive peak, then linearly decreases to a negative peak and returns to 0. This symmetrical waveform can drive ferroelectric domains to reciprocate in different directions, which is beneficial for breaking the pinning effect and reforming an ordered polarization arrangement. For amorphous memory cells in magnetoelectric random access memory, the generation module is configured as a double exponential decaying oscillating magnetic field waveform. This waveform generates a rapidly decaying alternating positive and negative magnetic field in the excitation coil, gradually calibrating the magnetic moments in the amorphous state to the target direction. For two-dimensional material molybdenum disulfide memory cells, the generation module is configured as an incremental low-voltage pulse sequence, starting from a lower voltage and gradually increasing to the target amplitude, exciting lattice phonon modes to promote the migration and recombination of defect atoms.

[0063] Finally, the calculated repair pulse voltage amplitude, repair pulse width, and waveform shape parameters are combined into a complete set of repair excitation parameters and written into the repair control register. Simultaneously, the module sends a start signal to the repair execution unit, notifying S140 that the repair excitation can begin. At this point, the entire processing procedure of S130 is complete.

[0064] The above process enables adaptive repair excitation generation based on defect severity, avoiding the dilemma of accelerated fatigue failure due to excessive voltage or ineffective repair due to insufficient voltage in traditional fixed-strength repair methods. The voltage saturation characteristic brought by the hyperbolic tangent function ensures safety under high defect severity, while the unbounded width growth characteristic brought by the exponential function ensures the long repair window required for severe defects. The combined strategy of increasing and decreasing voltage and width, and saturation and divergence, maximizes the lattice reconstruction effect of the repair excitation within the safety boundary, significantly improving the repair success rate and the number of repair cycles for memory cells.

[0065] S140, based on the voltage amplitude and pulse width of the repair pulse, apply a repair excitation matching the type of the storage cell to the storage cell to drive the atoms, dipoles or magnetic moments in the storage medium to rearrange.

[0066] In this embodiment, after S130 completes the calculation of the repair pulse parameters and writes the parameters into the repair control register, S140 is executed. The inputs to S140 include the repair pulse voltage amplitude, repair pulse width, and waveform shape parameters generated by S130 and latched by the repair control register, as well as the target memory address that initially triggers the entire repair process. This address is determined when acquiring parameters in S110 and is passed step by step through S120 and S130, ultimately used in S140 to locate the physical memory cell where repair excitation needs to be applied. The complete processing procedure of S140 is described in detail below for three different types of memory cells.

[0067] In one embodiment of this application, applying a repair stimulus matching the memory cell type to the memory cell based on the voltage amplitude and pulse width of the repair pulse includes: When the memory cell is a ferroelectric field-effect transistor, a reverse polarization triangular wave is applied, and the falling edge slope of the triangular wave is dynamically adjusted according to the ferroelectric domain wall migration rate. When the storage unit is a magnetoelectric random access memory, a double exponentially decaying oscillating magnetic field pulse corresponding to the equivalent voltage amplitude is applied, and the number of oscillations is dynamically determined by the dynamic defect entropy. When the storage cell is a two-dimensional material molybdenum disulfide, a low-voltage pulse sequence is applied that linearly increases from a preset proportional voltage to the amplitude of the repair pulse voltage to excite the lattice phonon mode.

[0068] After S140 starts, the repair execution unit inside the chip first reads the waveform shape parameters from the repair control register to determine the type of the memory cell to be repaired. Based on the read result, the repair execution unit jumps to the corresponding repair sub-process. At the same time, it obtains the row address and column address of the target memory block from the address register, selects the corresponding word line through the row decoder, and selects the corresponding bit line through the column selector, ensuring that the repair stimulus is accurately applied to the target cell without interfering with adjacent memory cells.

[0069] For ferroelectric field-effect transistor type memory cells, a waveform generator is first configured to output a reverse-polarized triangular wave. The reverse-polarized triangular wave is characterized by a symmetrical triangular shape where the voltage waveform rises positively and then falls negatively relative to the time axis. Three key parameters are read from the repair control register: the repair pulse voltage amplitude, the repair pulse width, and implicit parameters related to the steepness factor. The repair pulse voltage amplitude determines the peak voltage of the triangular wave, which is dynamically calculated using the hyperbolic tangent function in S130. The repair pulse width determines the total duration of the triangular wave from start to finish, and is dynamically calculated using an exponential function. The repair execution unit sets the waveform period of the triangular wave to be equal to the repair pulse width, allocating half the width to each of the positive and negative half-cycles.

[0070] During waveform generation, the slope of the falling edge of the triangular wave is dynamically adjusted to match the migration rate of the ferroelectric domain walls. Specifically, after the upward phase of the triangular wave ends and the falling edge output begins, a fixed falling rate is not used; instead, the inversion current of the ferroelectric domain walls is monitored in real time. The transient current value flowing through the ferroelectric capacitor is continuously read by a sensitive amplifier and input to a proportional-integral-derivative (PID) controller. This controller fine-tunes the gate voltage falling rate in real time based on the error of the current deviating from the ideal inversion current curve. When the ferroelectric domain wall migration speed is fast, the inversion current is large, and the controller appropriately accelerates the falling rate to avoid electric field lag. When the ferroelectric domain wall is pinned by defects, causing slow migration, the inversion current is small, and the controller actively slows down the falling rate to maintain the electric field effect near the pinning point for a longer period, helping the domain wall overcome the potential barrier. This dynamic slope adjustment mechanism ensures that the ferroelectric domains obtain a driving electric field that matches their migration capability throughout the entire reverse polarization process. It avoids creating unflipped regions due to excessively fast falling rates, and also avoids wasting energy and time due to excessively slow falling rates. After the triangular wave ends, the gate voltage is reset to 0 potential, and a short relaxation time is allowed to stabilize the polarization charge distribution in the ferroelectric layer. At this point, the repair of the ferroelectric field-effect transistor is complete, and the repair completion flag is set.

[0071] Different execution logic is employed for amorphous memory cells of the magnetoelectric random access memory (MERAM) type. MERAM stores data based on the direction of magnetic moments in the amorphous alloy thin film; radiation-induced magnetic moment disturbance manifests as some magnetic moments deviating from the target direction. The repair pulse voltage amplitude is read from the repair control register. This amplitude is not directly applied as a voltage but is converted into a pulsed magnetic field of corresponding intensity through an on-chip integrated magnetic field generating coil drive circuit. The peak intensity of the magnetic field is proportional to the repair pulse voltage amplitude, and the conversion factor is determined by the number of coil turns and the drive circuit gain, and is calibrated at the chip factory.

[0072] Optionally, the number of oscillations of the magnetic field pulse is dynamically determined by the dynamic defect entropy. The defect entropy value is read from the defect entropy register, multiplied by 10, and rounded down to the nearest integer, discarding the decimal part, to obtain the number of oscillations in integer form. For example, when the defect entropy is a critical value of 0.75, the number of oscillations is 7; when the defect entropy is 0.9, the number of oscillations is 9; and when the defect entropy is 1.2, the number of oscillations is 12. The more severe the defect, the more oscillations occur, allowing the repair unit to repeatedly calibrate the magnetic moment by alternating magnetic field directions, gradually guiding the disordered magnetic moment back to the target direction.

[0073] After determining the peak magnetic field strength and the number of oscillations, the magnetic field waveform generator is activated, outputting a double exponentially decaying oscillation waveform. This waveform is characterized by the rapid decay of the amplitude of each oscillation according to a double exponential function; that is, both the peak values ​​of the positive and negative half-cycles decrease exponentially with time, but the rate of decrease is initially faster and then slower, forming an asymmetrical decay envelope. Compared to constant-amplitude oscillations or single exponential decay, this waveform has the advantage that the initial large-amplitude oscillations can effectively break the magnetic moment in a deeply pinned state, and the subsequent weak oscillations after decay can finely adjust the magnetic moment to a stable orientation, avoiding overshoot. The waveform is output cyclically according to the set number of oscillations. The count is incremented after each complete positive and negative oscillation until the required number of oscillations is reached. Then, the magnetic field coil drive circuit is turned off, and the magnetic moment system is allowed to relax and stabilize naturally before the repair completion flag is set.

[0074] For two-dimensional molybdenum disulfide (MoDS) memory cells, atomically thin layers of MoDS generate point defects such as sulfur vacancies after irradiation. These defects can be migrated and recombinated with the aid of lattice thermal vibrations, i.e., phonon modes. The repair pulse voltage amplitude is read from the repair control register and used as a basis to construct an incremental low-voltage pulse sequence. The starting voltage of this sequence is set to 0.8 times the repair pulse voltage amplitude, and the ending voltage is the repair pulse voltage amplitude itself. The total number of steps in the pulse sequence is preset to 8, with the voltage of each step increasing linearly from the previous step. The step size is equal to 0.2 times the repair pulse voltage amplitude divided by 7. The pulse generator is configured to output these 8 pulses sequentially, with each pulse width fixed at 100 nanoseconds and an interval of 50 nanoseconds between pulses.

[0075] During the application of the incremental pulse sequence, the repair execution unit simultaneously monitors the weak current flowing through the molybdenum disulfide channel. Changes in this current reflect the excitation state of the lattice phonon modes and the activity level of defect migration. As the pulse voltage gradually increases, the injected carrier energy increases, gradually activating the in-plane and out-of-plane phonon modes of molybdenum disulfide through electron-phonon coupling. The energy of these phonon modes is transferred to the area around sulfur vacancies, lowering the defect migration barrier and increasing the probability of sulfur atoms jumping to vacancy sites to achieve defect recombination. The incremental design avoids directly damaging the lattice with the initial high voltage; instead, it allows defect migration to proceed gradually under relatively mild conditions through progressive excitation. After the pulse sequence ends, the repair execution unit removes all bias voltage, allowing the lattice temperature to slowly drop to equilibrium. This cooling process helps stabilize the repaired lattice configuration, after which a repair completion marker is set.

[0076] Regardless of the memory cell type, a repair completion timestamp and preliminary status indicating successful repair execution are written to the repair status register. Once the repair completion flag is set, the repair is not immediately considered effective; instead, control is returned to the main control process. Upon recognizing the repair completion flag, the main control process re-invokes S110 to collect defect-sensitive electrical parameters of the memory block and recalculates the dynamic defect entropy after repair via S120 to verify the actual repair effect. If the defect entropy after repair decreases below 0.75, the repair is confirmed successful and the block's health status table is updated; if it remains above 0.75, a feedback adjustment process is initiated. Within the maximum number of repairs, the parameters from this repair are stored as historical records in the learning database for subsequent optimization of repair parameters for similar cells. The entire S140 design ensures that the repair stimulus possesses sufficient physical strength to drive defect reconstruction while being subject to strict dynamic constraints to avoid secondary damage to the storage medium, achieving safe and effective lattice-level self-repair.

[0077] The above process transforms abstract repair parameters into specific repair stimuli tailored to the physical properties of different storage media, achieving material-level lattice reconstruction. For ferroelectric field-effect transistors, a reverse polarization triangular wave combined with dynamic falling edge slope adjustment ensures that the electric field change rate matches the actual migration rate of the ferroelectric domain walls in real time. This avoids domain wall depinning failure due to excessively rapid electric field changes or energy waste due to excessively slow changes, thus achieving optimal efficiency in rearranging polarization domains. For magnetoelectric random access memory (MRAM), a double exponential decaying oscillating magnetic field pulse breaks the magnetic moment in a deeply pinned state through initial large-amplitude oscillations, followed by weak oscillations after decay to finely adjust the magnetic moment to a stable orientation. The number of oscillations dynamically increases with defect entropy, ensuring that magnetic moments with varying degrees of disorder can be fully calibrated. For the two-dimensional material molybdenum disulfide, an incremental low-voltage pulse sequence gradually excites lattice phonon modes starting from lower energies, avoiding direct destruction of the atomic lattice by high voltage. Simultaneously, progressive energy injection promotes the migration and recombination of point defects such as sulfur vacancies. These three types of targeted repair stimuli maximize defect elimination while minimizing additional damage to the intrinsic structure of the storage medium.

[0078] S150: Collect the defect-sensitive electrical parameters of the same address unit in the repaired storage unit, recalculate the dynamic defect entropy after repair, and judge the repair effect based on the comparison results of the dynamic defect entropy after repair and multiple preset thresholds. If the repair is ineffective, strengthen the repair parameters and repeat the repair until the maximum number of repairs is reached, and then mark it as permanently failed.

[0079] In this embodiment, after the lattice reconstruction repair operation is completed in S140 and the repair execution unit sets the repair completion flag, the closed-loop verification and decision-making process in S150 is executed. The input to S150 is the electrical parameter vector obtained by re-measuring the same memory address cell after repair. This vector is re-acquired by the main control flow in S110 and includes the repaired threshold voltage drift, subthreshold swing degradation, and impulse response attenuation coefficient. The calculation module in S120 recalculates the repaired dynamic defect entropy based on these parameters. The complete processing of S150 is described in detail below.

[0080] In one embodiment of this application, defect-sensitive electrical parameters of the same address unit in the repaired storage unit are collected, the dynamic defect entropy after repair is recalculated, and the repair effect is judged based on the comparison results of the dynamic defect entropy after repair with multiple preset thresholds, including: Collect the defect-sensitive electrical parameters of the same address unit in the repaired storage unit, and recalculate the dynamic defect entropy after repair; if the dynamic defect entropy after repair is less than or equal to the first threshold, the repair is determined to be successful and the health status table is updated. If the dynamic defect entropy after repair is between the first threshold and the preset repair threshold, the repair is deemed effective, and the current repair parameters are recorded for subsequent adaptive learning. If the dynamic defect entropy after repair is greater than or equal to the preset repair threshold, the repair is deemed invalid. The width of the repair pulse is increased by a predetermined step size, and the repair operation is repeated until the maximum number of repairs is reached. Then, the storage unit is marked as permanently failed, and backup block mapping is initiated.

[0081] In one embodiment of this application, S150 initiates the verification and decision module to read the register, obtain the repaired defect entropy value, and compare it with two preset thresholds, which are 0.65 and 0.75. 0.65 represents the excellent health recovery threshold; a value below this indicates that the repair effect is ideal, and the storage medium has basically returned to its factory-like state. 0.75 represents the critical threshold for repair triggering; a value above this indicates that the repair has failed to reduce the defect below a safe level. The interval between 0.65 and 0.75 represents a partially effective area, where the repair has played a certain role, but there are still observable residual defects.

[0082] First, it is determined whether the defect entropy after repair is less than or equal to 0.65. If this condition is met, the repair is considered completely successful. Then, the block health status table stored in the built-in flash memory is accessed. This table is a linear table indexed by the storage block address, with each entry occupying one byte and recording the latest health status of the corresponding storage block. The offset in the table is calculated based on the target storage block's address, and the value of the corresponding entry is updated to "excellent health status." Simultaneously, the repair success counter is incremented by 1. Optionally, the defect entropy before repair, the defect entropy after repair, and the repair voltage amplitude and pulse width used are appended to the repair history log as a successful repair record for subsequent statistical analysis. After these operations are completed, a successful repair status is returned to the main control process, the entire repair process ends, and the storage block resumes normal read / write use.

[0083] If the repaired defect entropy is greater than 0.65, the second-level judgment is initiated, checking whether the repaired defect entropy is less than or equal to 0.75. If this condition is met, the repair is considered partially effective, meaning the repair operation did improve the defect state of the storage medium, but the improvement was insufficient to reach an excellent level, and slight degradation still exists. The operation performed at this point differs from that of a fully successful repair. First, the block health status table is updated, marking the block's status as partially effective rather than excellent, implying that the block may require more frequent monitoring in the future. Then, all parameters of this repair, including the defect entropy before repair, the voltage amplitude and pulse width used in the repair, the defect entropy after repair, and the current cumulative radiation dose and temperature, are packaged and written into a dedicated partially effective repair record table. This record table is used for subsequent adaptive learning. The chip's internal learning algorithm analyzes these partially effective cases, uncovering the correlation between voltage, pulse width, and entropy changes, and gradually optimizing the repair parameter calculation formula in step three, enabling the selection of better repair parameters to improve the repair effect under similar defect levels in the future. The partially effective repair status is returned to the main control process, allowing the storage block to continue to be used but marked as requiring close monitoring.

[0084] Optionally, if the defect entropy after repair is greater than 0.75, the repair is considered invalid. This means that after one repair, the defect level of the storage medium is still above the critical threshold for triggering repair, and the enhanced repair sub-process begins. First, the pulse width value used in this repair is read from the repair history register and multiplied by 1.5 to obtain the new pulse width. Simultaneously, it is checked whether the new pulse width exceeds the 1-microsecond upper limit constraint. If it does, it is clamped to 1 microsecond. The 1-microsecond upper limit is designed based on the physical tolerance of ferroelectric and magnetic materials. Continuous pulses exceeding this width may induce electrode electromigration or dielectric thermal breakdown, causing irreversible damage. The pulse voltage amplitude remains unchanged at this stage because the voltage has already saturated from the hyperbolic tangent function to near the upper limit, and the risk of further increasing the voltage far outweighs the benefit.

[0085] Using the enhanced pulse width, combined with the original voltage amplitude and waveform shape parameters, a repair stimulus is applied again to the same memory address. After the second repair, the parameter acquisition in S110 and the entropy calculation in S120 are called again to obtain the defect entropy after the second repair. The entropy value after the second repair is compared with 0.75 again. If the entropy value after the second repair drops below 0.75, it is recorded as a successful repair but the enhancement parameters were used, and the health status of the block is marked as enhanced repair recovery. If the entropy value after the second repair is still greater than 0.75, a third repair is performed, with the pulse width multiplied by 1.5 again, with the upper limit still one microsecond. If the entropy value after the third repair still cannot drop below 0.75, the defect of the memory block is considered to have reached an irreversible level, and continuing repair will only waste time and energy and may affect surrounding memory cells.

[0086] At this point, the final failure handling process is executed. First, the address of the target memory block is written to a permanent failure block table. This table is stored in the chip's non-volatile memory area. During chip power-on initialization, this table is read and all addresses recorded in the table are skipped, ensuring that these failure blocks are never used by the operating system or file system. Then, the address of a free spare block is obtained from the spare block mapping table. Spare blocks are healthy memory blocks reserved during chip factory testing; typically, 2% to 5% of the total capacity is reserved as spares in each memory chip. The addresses of the failure block and the spare block are written to a remapping table, which is queried in real-time by a hardware address comparator. When any subsequent read or write operation accesses the failure block address, the address comparator automatically redirects the access to the corresponding spare block address. The entire process is completely transparent to the upper-layer software. Simultaneously, an unrecoverable block failure event is reported to the main control flow and recorded in the chip's fault log for ground analysis. At this point, the entire processing procedure of the S150 is complete.

[0087] The above process objectively quantifies the actual effect of the repair operation by re-collecting the electrical parameters after repair and calculating the defect entropy after repair, rather than blindly trusting the completion of the repair action alone. The three-level threshold judgment mechanism realizes a refined effect classification: repairs that are better than the excellent line are considered completely successful, and the block health status is confirmed; some effective repairs between the excellent line and the trigger line, although they have not reached the optimal state, have their repair parameters recorded for subsequent learning and optimization, so as to continuously accumulate experience and gradually improve the intelligence of the repair strategy; ineffective repairs that are higher than the trigger line trigger an enhanced repair process, which gradually increases the pulse width to give the defect a more sufficient opportunity for repair, avoiding the hasty abandonment of memory cells due to poor parameters in a single instance. Cells that are still ineffective after a maximum of three enhanced repairs are marked as permanently failed and a backup block remapping is initiated. This fallback mechanism ensures that even if a few cells are ultimately unrecoverable, the overall storage capacity and functional integrity of the chip will not be compromised, greatly improving the overall fault tolerance and lifespan of the system.

[0088] The technical solution of this application collects defect-sensitive electrical parameters of the storage cell, and calculates the dynamic defect entropy, which characterizes the degree of defect accumulation inside the storage medium, based on the defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature. When the dynamic defect entropy exceeds a preset repair threshold, the voltage amplitude and pulse width of the repair pulse are adaptively generated based on the dynamic defect entropy. Based on the voltage amplitude and pulse width of the repair pulse, a repair excitation matching the storage cell type is applied to the storage cell to drive the atoms, dipoles, or magnetic moments in the storage medium to rearrange. The defect-sensitive electrical parameters of the same address cell in the repaired storage cell are collected, the dynamic defect entropy after repair is recalculated, and the repair effect is judged based on the comparison results of the dynamic defect entropy after repair with multiple preset thresholds. If the repair is ineffective, the repair parameters are strengthened and the repair is repeated until the maximum number of repairs is reached, after which it is marked as permanently failed. By collecting defect-sensitive electrical parameters to quantify the degree of radiation damage, calculating dynamic defect entropy to trigger adaptive repair, applying repair stimuli that match the degree of defect, and verifying the repair effect in a closed loop, the system achieves proactive sensing and material-level lattice reconstruction repair of metastable defects inside the storage medium, thereby significantly extending the effective service life of the storage cell under radiation accumulation conditions.

[0089] The following describes embodiments of the automatic fault repair device based on a storage unit architecture according to this application, which can be used to execute the automatic fault repair method based on a storage unit architecture in the above embodiments of this application. It is understood that the automatic fault repair device based on a storage unit architecture can be a computer program (including program code) running on a computer device; for example, the automatic fault repair device based on a storage unit architecture can install industrial application software or industrial control management software; the automatic fault repair device based on a storage unit architecture can be used to execute the corresponding steps in the method provided in the embodiments of this application. For details not disclosed in the embodiments of the automatic fault repair device based on a storage unit architecture of this application, please refer to the embodiments of the automatic fault repair method based on a storage unit architecture described above in this application.

[0090] Figure 3 A block diagram of an automatic fault repair apparatus based on a storage cell architecture according to an embodiment of this application is shown.

[0091] Reference Figure 3 As shown, an automatic fault repair apparatus based on a storage cell architecture according to an embodiment of this application includes: The acquisition unit 310 is used to acquire the defect-sensitive electrical parameters of the storage unit, including the threshold voltage drift, the subthreshold swing degradation, and the impulse response attenuation coefficient. The calculation unit 320 is used to calculate the dynamic defect entropy, which characterizes the degree of defect accumulation inside the storage medium, based on the defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature. The parameter tuning unit 330 is used to adaptively generate the voltage amplitude and pulse width of the repair pulse according to the dynamic defect entropy when the dynamic defect entropy exceeds the preset repair threshold. The driving unit 340 is used to apply a repair excitation matching the type of the storage cell to the storage cell according to the voltage amplitude and pulse width of the repair pulse, so as to drive the atoms, dipoles or magnetic moments in the storage medium to rearrange. Feedback unit 350 is used to collect defect-sensitive electrical parameters of the same address unit in the repaired storage unit, recalculate the repaired dynamic defect entropy, and judge the repair effect based on the comparison results of the repaired dynamic defect entropy and multiple preset thresholds. If the repair is ineffective, the repair parameters are strengthened and the repair is repeated until the maximum number of repairs is reached and then it is marked as permanently failed. The calculation of the dynamic defect entropy, which characterizes the degree of defect accumulation within the storage medium, based on the defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature, includes: normalizing the defect-sensitive electrical parameters to obtain a normalized parameter vector; constructing a hyperbolic tangent function based on the product of the normalized parameter vector, normalized cumulative radiation dose, and normalized chip junction temperature, and introducing a natural logarithm operation to generate a defect potential function; inputting the defect potential function into a sigmoid function and a sinusoidal square modulation function respectively, and multiplying the outputs of the two to obtain the dynamic defect entropy.

[0092] In this application, based on the aforementioned scheme, the acquisition of defect-sensitive electrical parameters of the memory cell includes: applying a linearly increasing gate voltage to the memory cell multiple times and reading the channel current abrupt change point; comparing the median of the multiple measurement results with the factory nominal value to obtain the threshold voltage drift; measuring the gate voltage change required for the leakage current to change from a first preset ratio to a second preset ratio in the subthreshold region; comparing the median of the multiple measurement results with the initial subthreshold swing value to obtain the subthreshold swing degradation; and measuring the channel conductance after applying a standard write pulse to the memory cell, and obtaining the pulse response attenuation coefficient based on the statistical data of the multiple measurement results.

[0093] In this application, based on the aforementioned scheme, the step of calculating the dynamic defect entropy, which characterizes the degree of defect accumulation inside the storage medium, according to the defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature, includes: normalizing the defect-sensitive electrical parameters to obtain a normalized parameter vector; constructing a hyperbolic tangent function based on the product of the normalized parameter vector, the normalized cumulative radiation dose, and the normalized chip junction temperature, and introducing a natural logarithm operation to generate a defect potential function; inputting the defect potential function into a sigmoid function and a sinusoidal square modulation function respectively, and multiplying the outputs of the two to obtain the dynamic defect entropy.

[0094] In this application, based on the aforementioned scheme, the step of adaptively generating the voltage amplitude and pulse width of the repair pulse according to the dynamic defect entropy when the dynamic defect entropy exceeds a preset repair threshold includes: when the dynamic defect entropy exceeds the preset repair threshold, reading the difference between the dynamic defect entropy and the preset repair threshold; dynamically adjusting the voltage multiplier using a hyperbolic tangent function based on the difference, and then multiplying the voltage multiplier by a reference voltage to obtain the voltage amplitude of the repair pulse, wherein the voltage multiplier saturates as the difference increases; dynamically adjusting the pulse width multiplier using a natural exponential function based on the square of the difference, and multiplying the pulse width multiplier by a reference pulse width to obtain the pulse width of the repair pulse, wherein the pulse width multiplier increases exponentially as the difference increases.

[0095] In this application, based on the aforementioned scheme, the step of applying a repair excitation to the memory cell that matches the memory cell type according to the voltage amplitude and pulse width of the repair pulse includes: when the memory cell is a ferroelectric field-effect transistor, applying a reverse polarized triangular wave and dynamically adjusting the falling edge slope of the triangular wave according to the ferroelectric domain wall migration rate; when the memory cell is a magnetoelectric random access memory, applying a double exponentially decaying oscillating magnetic field pulse corresponding to the equivalent voltage amplitude, and the number of oscillations being dynamically determined by the dynamic defect entropy; when the memory cell is a two-dimensional material molybdenum disulfide, applying a low-voltage pulse sequence that linearly increases from a preset proportional voltage to the voltage amplitude of the repair pulse to excite lattice phonon modes.

[0096] In this application, based on the aforementioned scheme, the defect-sensitive electrical parameters of the same address unit in the repaired storage unit are collected, the dynamic defect entropy after repair is recalculated, and the repair effect is judged based on the comparison results of the dynamic defect entropy after repair with multiple preset thresholds. This includes: collecting defect-sensitive electrical parameters of the same address unit in the repaired storage unit and recalculating the dynamic defect entropy after repair; if the dynamic defect entropy after repair is less than or equal to a first threshold, the repair is determined to be successful and the health status table is updated; if the dynamic defect entropy after repair is between the first threshold and the preset repair threshold, the repair is determined to be partially effective, and the current repair parameters are recorded for subsequent adaptive learning; if the dynamic defect entropy after repair is greater than or equal to the preset repair threshold, the repair is determined to be invalid, and the width of the repair pulse is increased by a predetermined step size and the repair operation is repeated until the maximum number of repairs is reached, after which the storage unit is marked as permanently failed and the backup block mapping is started.

[0097] In this application, based on the aforementioned scheme, the value range of the dynamic defect entropy is configured to be 0 to 1.25, the preset repair threshold is configured to be 0.75, and an automatic fault repair process is triggered when the dynamic defect entropy exceeds 0.75.

[0098] In this application, based on the aforementioned scheme, the statistical data of the multiple measurement results includes the median value of eight measurement results.

[0099] The technical solution of this application collects defect-sensitive electrical parameters of the storage cell, and calculates the dynamic defect entropy, which characterizes the degree of defect accumulation inside the storage medium, based on the defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature. When the dynamic defect entropy exceeds a preset repair threshold, the voltage amplitude and pulse width of the repair pulse are adaptively generated based on the dynamic defect entropy. Based on the voltage amplitude and pulse width of the repair pulse, a repair excitation matching the storage cell type is applied to the storage cell to drive the atoms, dipoles, or magnetic moments in the storage medium to rearrange. The defect-sensitive electrical parameters of the same address cell in the repaired storage cell are collected, the dynamic defect entropy after repair is recalculated, and the repair effect is judged based on the comparison results of the dynamic defect entropy after repair with multiple preset thresholds. If the repair is ineffective, the repair parameters are strengthened and the repair is repeated until the maximum number of repairs is reached, after which it is marked as permanently failed. By collecting defect-sensitive electrical parameters to quantify the degree of radiation damage, calculating dynamic defect entropy to trigger adaptive repair, applying repair stimuli that match the degree of defect, and verifying the repair effect in a closed loop, the system achieves proactive sensing and material-level lattice reconstruction repair of metastable defects inside the storage medium, thereby significantly extending the effective service life of the storage cell under radiation accumulation conditions.

[0100] Figure 4 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown.

[0101] It should be noted that the computer system of the electronic device in this embodiment is only an example and should not impose any limitations on the function and scope of use of the embodiments of this application.

[0102] In this embodiment, the computer system includes a central processing unit 401, which can perform various appropriate actions and processes based on programs stored in read-only memory 402 or programs loaded from storage section 408 into random access memory 403, such as executing the automatic fault repair method based on storage unit architecture described in the above embodiment. The random access memory 403 also stores various programs and data required for system operation, thereby realizing big data storage and big data management. The central processing unit 401, read-only memory 402, and random access memory 403 are interconnected via bus 404. Input / output interface 405 is also connected to bus 404.

[0103] The following components are connected to the input / output interface 405: an input section 406 including a keyboard, mouse, etc.; an output section 407 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the input / output interface 405 as needed. A removable medium 411, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 410 as needed so that computer programs read from it can be installed into the storage section 408 as needed.

[0104] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411. When the computer program is executed by central processing unit 401, it performs various functions defined in the system of this application.

[0105] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. The transmitted data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0106] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0107] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0108] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various alternative implementations described above.

[0109] In another aspect, this application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to implement the automatic fault repair method based on the storage unit architecture described in the above embodiments.

[0110] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0111] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the methods according to the embodiments of this application.

[0112] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0113] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. An automatic fault repair method based on a storage unit architecture, characterized in that, include: The defect-sensitive electrical parameters of the storage unit are collected, including threshold voltage drift, subthreshold swing degradation, and impulse response attenuation coefficient. Based on the defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature, calculate the dynamic defect entropy, which characterizes the degree of defect accumulation inside the storage medium. When the dynamic defect entropy exceeds the preset repair threshold, the voltage amplitude and pulse width of the repair pulse are adaptively generated based on the dynamic defect entropy. Based on the voltage amplitude and pulse width of the repair pulse, a repair excitation matching the type of the storage cell is applied to the storage cell to drive the atoms, dipoles or magnetic moments in the storage medium to rearrange. Collect the defect-sensitive electrical parameters of the same address unit in the repaired storage unit, recalculate the dynamic defect entropy after repair, and judge the repair effect based on the comparison results of the dynamic defect entropy after repair with multiple preset thresholds. If the repair is ineffective, strengthen the repair parameters and repeat the repair until the maximum number of repairs is reached, then mark it as permanently failed. Based on the aforementioned defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature, the dynamic defect entropy, characterizing the degree of defect accumulation within the storage medium, is calculated, including: The defect-sensitive electrical parameters are normalized to obtain a normalized parameter vector; Based on the product of the normalized parameter vector, the normalized cumulative radiation dose, and the normalized chip junction temperature, a hyperbolic tangent function is constructed, and the natural logarithm operation is introduced to generate the defect potential function. The defect potential function is input into the sigmoid function and the sinusoidal square modulation function respectively, and the outputs of the two are multiplied to obtain the dynamic defect entropy.

2. The automatic fault repair method based on storage unit architecture according to claim 1, characterized in that, The defect-sensitive electrical parameters of the acquired storage unit include: By applying linearly increasing gate voltage to the memory cell multiple times and reading the channel current abrupt change point, the median of the multiple measurement results is compared with the factory nominal value to obtain the threshold voltage drift. The gate voltage change required for the leakage current to change from the first preset ratio to the second preset ratio is measured in the subthreshold region. The median of multiple measurement results is then compared with the initial subthreshold swing value to obtain the subthreshold swing degradation amount. After applying a standard write pulse to the memory cell, the channel conductance is measured, and the pulse response attenuation coefficient is obtained based on the statistical data of multiple measurement results.

3. The automatic fault repair method based on storage unit architecture according to claim 1, characterized in that, When the dynamic defect entropy exceeds a preset repair threshold, the voltage amplitude and pulse width of the repair pulse are adaptively generated based on the dynamic defect entropy, including: When the dynamic defect entropy exceeds a preset repair threshold, the difference between the dynamic defect entropy and the preset repair threshold is read. Based on the difference, the voltage multiplier is dynamically adjusted using a hyperbolic tangent function, and then the voltage multiplier is multiplied by the reference voltage to obtain the voltage amplitude of the repair pulse, wherein the voltage multiplier saturates as the difference increases; Based on the square of the difference, the pulse width ratio is dynamically adjusted using a natural exponential function, and the pulse width ratio is multiplied by the reference pulse width to obtain the pulse width of the repair pulse, wherein the pulse width ratio increases exponentially with the increase of the difference.

4. The automatic fault repair method based on storage unit architecture according to claim 1, characterized in that, Based on the voltage amplitude and pulse width of the repair pulse, a repair stimulus matching the memory cell type is applied to the memory cell, including: When the memory cell is a ferroelectric field-effect transistor, a reverse polarization triangular wave is applied, and the falling edge slope of the triangular wave is dynamically adjusted according to the ferroelectric domain wall migration rate. When the storage unit is a magnetoelectric random access memory, a double exponentially decaying oscillating magnetic field pulse corresponding to the equivalent voltage amplitude is applied, and the number of oscillations is dynamically determined by the dynamic defect entropy. When the storage cell is a two-dimensional material molybdenum disulfide, a low-voltage pulse sequence is applied that linearly increases from a preset proportional voltage to the amplitude of the repair pulse voltage to excite the lattice phonon mode.

5. The automatic fault repair method based on storage unit architecture according to claim 1, characterized in that, Collect defect-sensitive electrical parameters of the same address unit in the repaired storage unit, recalculate the repaired dynamic defect entropy, and determine the repair effect based on the comparison results of the repaired dynamic defect entropy with multiple preset thresholds, including: Collect the defect-sensitive electrical parameters of the same address unit in the repaired storage unit, and recalculate the dynamic defect entropy after repair; If the dynamic defect entropy after repair is less than or equal to the first threshold, the repair is determined to be successful and the health status table is updated. If the dynamic defect entropy after repair is between the first threshold and the preset repair threshold, the repair is deemed effective, and the current repair parameters are recorded for subsequent adaptive learning. If the dynamic defect entropy after repair is greater than or equal to the preset repair threshold, the repair is deemed invalid. The width of the repair pulse is increased by a predetermined step size, and the repair operation is repeated until the maximum number of repairs is reached. Then, the storage unit is marked as permanently failed, and backup block mapping is initiated.

6. The automatic fault repair method based on a storage cell architecture according to any one of claims 1-5, characterized in that, The dynamic defect entropy is configured to range from 0 to 1.25, the preset repair threshold is configured to be 0.75, and an automatic fault repair process is triggered when the dynamic defect entropy exceeds 0.

75.

7. The automatic fault repair method based on storage unit architecture according to claim 2, characterized in that, The statistical data of the multiple measurements includes the median value of the eight measurements.

8. An automatic fault repair device based on a storage unit architecture, characterized in that, include: The acquisition unit is used to acquire defect-sensitive electrical parameters of the storage unit, including threshold voltage drift, subthreshold swing degradation, and impulse response attenuation coefficient. The calculation unit is used to calculate the dynamic defect entropy, which characterizes the degree of defect accumulation inside the storage medium, based on the defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature. The parameter tuning unit is used to adaptively generate the voltage amplitude and pulse width of the repair pulse based on the dynamic defect entropy when the dynamic defect entropy exceeds the preset repair threshold. A driving unit is used to apply a repair excitation matching the type of the storage cell to the storage cell according to the voltage amplitude and pulse width of the repair pulse, so as to drive the atoms, dipoles or magnetic moments in the storage medium to rearrange. The feedback unit is used to collect the defect-sensitive electrical parameters of the same address unit in the repaired storage unit, recalculate the dynamic defect entropy after repair, and judge the repair effect based on the comparison results of the dynamic defect entropy after repair and multiple preset thresholds. If the repair is ineffective, the repair parameters are strengthened and the repair is repeated until the maximum number of repairs is reached and then it is marked as permanently failed. Based on the aforementioned defect-sensitive electrical parameters, cumulative radiation dose, and chip junction temperature, the dynamic defect entropy, characterizing the degree of defect accumulation within the storage medium, is calculated, including: The defect-sensitive electrical parameters are normalized to obtain a normalized parameter vector; Based on the product of the normalized parameter vector, the normalized cumulative radiation dose, and the normalized chip junction temperature, a hyperbolic tangent function is constructed, and the natural logarithm operation is introduced to generate the defect potential function. The defect potential function is input into the sigmoid function and the sinusoidal square modulation function respectively, and the outputs of the two are multiplied to obtain the dynamic defect entropy.

9. A computer-readable medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the automatic fault repair method based on the storage unit architecture as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the automatic fault repair method based on a storage cell architecture as described in any one of claims 1 to 7.

Images