Flexible display screen bending angle and conductivity correlation test method and system

Abstract

The invention provides a flexible display screen bending angle and conductivity correlation test method and system, and relates to the technical field of display screen testing, and the method comprises the steps: synchronously recording an angle and a plurality of electrical parameters in a bending process, calculating a resistance hysteresis loop area and a residual increment, and distinguishing damage types through AC impedance. Microcracks are identified through resistance noise power, a phase-space analysis is established to identify a damage evolution stage and a turning point, a damage concentration section is determined, a position migration sequence is recorded, and a damage development trend is predicted, so that accurate evaluation and prediction of the reliability of the flexible display screen are realized.

Description

Technical Field

[0001] This invention relates to display screen testing technology, and more particularly to a method and system for testing the correlation between the bending angle and conductivity of flexible displays. Background Technology

[0002] Flexible display technology, as an important development direction of next-generation display technology, boasts advantages such as thinness, flexibility, and impact resistance, and has demonstrated enormous application potential in fields such as smart wearables, foldable phones, and rollable TVs. However, flexible displays need to withstand repeated bending during practical use, posing a severe challenge to their structural integrity and electrical performance. Flexible displays are typically composed of multiple layers of composite materials, including a substrate layer, conductive layer, light-emitting layer, and encapsulation layer. Repeated bending can lead to interface debonding, material fatigue, and microcrack formation, ultimately affecting the conductivity and display quality of the display. Therefore, establishing a testing method to correlate the bending angle of flexible displays with their conductivity is crucial for evaluating their reliability and predicting their lifespan.

[0003] Existing testing methods typically employ static bending at a fixed angle or simple periodic bending tests, which cannot monitor changes in the electrical parameters of the display screen in real time during the bending process. In particular, it is difficult to capture the dynamic relationship between changes in bending angle and degradation of electrical performance, resulting in test results that fail to reflect performance changes under actual usage conditions.

[0004] Existing technologies lack effective means of identifying damage mechanisms, making it impossible to accurately distinguish between different failure modes such as interface debonding, bulk material damage, and microcracks. It is also difficult to locate specific damage locations and predict damage evolution trends. This limits the guiding significance of test results and makes it difficult to provide targeted suggestions for material improvement and structural optimization. Summary of the Invention

[0005] This invention provides a method and system for testing the correlation between the bending angle and conductivity of flexible displays, which can solve the problems in the prior art.

[0006] A first aspect of the present invention provides a method for testing the correlation between the bending angle and conductivity of a flexible display screen, comprising: During the repeated bending of the flexible display screen, the bending angle is recorded simultaneously, and the resistance value, AC impedance and resistance noise power are collected in multiple monitoring sections set along the bending axis. The closed loop area formed by the change of resistance with angle in each bending cycle is calculated as the hysteresis loop area and the residual resistance increment of each monitoring section. The AC impedance is used to distinguish between interface debonding and bulk damage, and the resistance noise power is used to identify microcracks. Phase space analysis constructed by residual resistance increment, hysteresis toroidal accumulation, and low-frequency AC impedance identifies the damage evolution stage and locates the damage inflection point. The section with the largest residual resistance increment growth rate and microcracks is identified as the damage concentration section. Record the spatial location migration sequence of the damage concentration area, and determine the dominant failure mode based on the changing trends of AC impedance and resistive noise power; The spatial location migration sequence and historical electrical parameter data are input into a neural network to predict the migration direction and migration probability. The acquisition frequency and detection point position are dynamically adjusted in combination with the proximity of the current cycle distance to the damage inflection point.

[0007] The steps of synchronously recording the bending angle and acquiring resistance values, AC impedance, and resistance noise power in multiple monitoring sections set along the bending axis include: During the bending process, the bending angle change curve is acquired in real time and divided into a loading stage and an unloading stage. During the loading stage, the electrical parameters of each monitoring section are collected at a preset angle interval. The preset angle interval is determined according to the angle interval corresponding to the peak position of the resistance change rate in the previous bending cycle. During the unloading phase, the monitoring section whose resistance value has not recovered to the initial value is identified as the plastic deformation section, and the preset angle interval is reduced for the plastic deformation section in subsequent bending cycles; Establish a time-series correlation between the resistance values, AC impedance, and resistance noise power collected in each monitoring section during the loading and unloading phases and the corresponding bending angles.

[0008] The steps for calculating the closed loop area formed by the resistance change with angle in each bending cycle as the hysteresis loop area and the residual resistance increment of each monitoring section, where AC impedance is used to distinguish between interface debonding and bulk damage, and resistance noise power is used to identify microcracks, include: Extract the set of data points that form a closed trajectory from the collected angle-resistance data points, and calculate the area enclosed by the closed trajectory as the hysteresis loop area by integration; The difference between the resistance value when each bending cycle returns to the initial straight state and the initial reference value is calculated as the residual resistance increment. When the residual resistance increment exceeds the preset residual threshold, microcrack detection is triggered. During the peak angle holding stage of the bending cycle that triggers microcrack detection, a high-frequency pulse current is applied to the corresponding monitoring section as a perturbation excitation. The high-frequency fluctuation of the resistance signal is collected and the power spectral density is calculated. When the power spectral density exceeds the preset power threshold, it is determined that there is a microcrack in the monitoring section. The transient recovery curve of the resistance value of the monitoring section is collected after the perturbation excitation is removed. The rapid decline stage and the slow decline stage are identified, and the rapid decay time and the slow decay time are calculated respectively. The monitoring section is determined to have entered the critical instability state based on the change characteristics of the rapid decay time and the slow decay time. Different frequencies of AC excitation are applied to the monitoring section that has entered the critical instability state and the AC impedance is measured. When the low-frequency impedance increase exceeds the preset ratio, it is determined to be interface debonding. When the wide-band impedance fluctuation exceeds the preset dispersion, it is determined to be body damage.

[0009] The steps for acquiring the transient recovery curve of the resistance value of the monitored section after the perturbation excitation is removed, and determining the critical instability state, include: At the instant the perturbation excitation is removed, high-frequency sampling and monitoring of the resistance value of the section are started to form a transient recovery curve. The first derivative of the transient recovery curve is calculated to obtain the resistance decrease rate curve. The first peak point in the resistance decrease rate curve is identified as the dividing point between the rapid decrease stage and the slow decrease stage. The time interval from the removal of the perturbation excitation to the boundary point is defined as the rapid decay time, and the time interval from the boundary point to the stabilization of the resistance value is defined as the slow decay time. The rapid decay time constant is extracted by fitting an exponential function to the rapid decay phase, and the slow decay time constant is extracted by fitting a logarithmic function to the slow decay phase. The ratio of the two time constants is calculated as a dual-stage decay characteristic parameter. When the dual-stage decay characteristic parameter decreases monotonically and the deceleration rate exceeds a preset rate threshold, it is determined that the material's self-healing ability has decreased. Calculate the second derivative of the dual-stage decay characteristic parameter with respect to the number of cycles. When the second derivative changes from a negative value to a positive value, it is determined that the growth rate of the material stability index has reached an inflection point, and the monitoring section is marked to enter a critical instability state.

[0010] The steps of identifying damage evolution stages and locating damage inflection points through phase space analysis constructed using the residual resistance increment, hysteresis loop accumulation, and low-frequency AC impedance, and identifying the segment with the largest residual resistance increment growth rate and the presence of microcracks as the damage concentration segment, include: Extract the residual resistance increment, hysteresis loop accumulated value and low-frequency AC impedance of each bending cycle to construct three-dimensional phase space coordinate points. Calculate the Euclidean distance between adjacent cycle coordinate points as the state evolution step size. When the state evolution step size continues to increase, it is marked as entering the accelerated evolution stage. For cycles after entering the accelerated evolution stage, the coefficient of variation of the residual resistance increment and the accumulated value of the hysteresis loop within the sliding time window is calculated. When the rate of change of the coefficient of variation between cycles changes from less than the preset stability threshold to continuously increasing, it is determined to be an accelerated instability state. For a loop interval determined to be in an accelerated instability state, the mutual information between the residual resistance increment and the accumulated value of the hysteresis loop is calculated. When the decrease in the mutual information exceeds a preset decrease threshold, the loop interval is marked as having a change in damage mechanism. Within the cyclic interval where the damage mechanism changes, the information entropy of the distribution of three-dimensional phase space coordinate points is calculated, and the number of cycles in which the information entropy jump exceeds the preset entropy change threshold is identified as the damage inflection point. In the cycle corresponding to the damage inflection point and subsequent cycles, the section with the largest first derivative of the residual resistance increment and a positive second derivative is identified from the monitoring section with microcrack markings as the damage concentration section.

[0011] The steps of recording the spatial location migration sequence of the damage concentration section and determining the dominant failure mode based on the changing trends of AC impedance and resistive noise power include: The spatial location change of the damage concentration section is tracked in a continuous bending cycle. When the damage concentration section migrates from the first monitoring section to the second monitoring section, the migration event is recorded. The continuous migration events are arranged in chronological order to form a spatial location migration sequence. The frequency of occurrence of interface debonding judgment results and body damage judgment results corresponding to each monitoring segment in the spatial location migration sequence is statistically analyzed, and the failure mode corresponding to the judgment result with higher occurrence frequency is determined as the dominant failure mode. The dominant failure mode is cross-validated with the resistance noise power change trend of the corresponding monitoring section. When the dominant failure mode is interface debonding and there is no significant pulse change in resistance noise power, the judgment result is confirmed. When the dominant failure mode is body damage and there is a continuous pulse change in resistance noise power, the judgment result is confirmed.

[0012] The steps of inputting the spatial location migration sequence and historical electrical parameter data into a neural network to predict the migration direction and migration probability, and dynamically adjusting the acquisition frequency and detection point position based on the proximity of the current cycle to the damage inflection point, include: The migration direction label and the AC impedance frequency response curve and the time-domain fluctuation characteristics of the resistance noise power corresponding to the migration time in the spatial location migration sequence are extracted as input features. The migration probability distribution of each monitoring segment becoming the next damage concentration segment is output through the neural network, and the monitoring segment with the highest probability is identified from the migration probability distribution as the predicted migration direction. Calculate the difference between the accumulated value of the hysteresis torus corresponding to the current cycle and the accumulated value of the hysteresis torus corresponding to the damage inflection point. When the difference is less than a preset difference threshold and the growth rate sequence shows an upward trend, increase the sampling frequency to a preset multiple of the initial sampling frequency. The monitoring section corresponding to the predicted migration direction and its adjacent monitoring sections along the bending axis are marked as high-risk sections. In subsequent bending cycles, the resistance value, AC impedance, and resistance noise power acquisition density of the high-risk sections are increased to a preset multiple of those of other monitoring sections.

[0013] A second aspect of the present invention provides a testing system for the correlation between the bending angle and conductivity of a flexible display screen, comprising: The data acquisition module is used to simultaneously record the bending angle and collect resistance values, AC impedance, and resistance noise power in multiple monitoring sections set along the bending axis during the repeated bending of the flexible display screen. The feature extraction module is used to calculate the closed loop area formed by the change of resistance with angle in each bending cycle as the hysteresis loop area and the residual resistance increment of each monitoring section. Among them, the AC impedance is used to distinguish between interface debonding and bulk damage, and the resistance noise power is used to identify microcracks. The damage analysis module is used to identify the damage evolution stage and locate the damage inflection point through phase space analysis constructed by the residual resistance increment, the accumulated value of the hysteresis loop and the low-frequency AC impedance. The section with the largest growth rate of the residual resistance increment and the presence of microcracks is identified as the damage concentration section. The failure mode determination module is used to record the spatial location migration sequence of the damage concentration section and determine the dominant failure mode based on the changing trends of AC impedance and resistance noise power. The adaptive control module is used to input the spatial location migration sequence and historical electrical parameter data into the neural network to predict the migration direction and migration probability, and dynamically adjust the acquisition frequency and detection point position based on the proximity of the current cycle distance to the damage inflection point.

[0014] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0015] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0016] This invention innovatively introduces hysteresis loop area calculation and residual resistance increment analysis, combines the function of AC impedance to distinguish interface debonding and bulk damage, and the characteristic of resistance noise power to identify microcracks, to establish a complete damage feature extraction system; it also innovatively records the spatial location migration sequence of damage concentration sections, and combines the trend of AC impedance and resistance noise power changes to determine the dominant failure mode, thus achieving a deep understanding of the failure mechanism. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the method for testing the correlation between the bending angle and conductivity of a flexible display screen according to an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0020] Figure 1 This is a flowchart illustrating the testing method for the correlation between the bending angle and conductivity of a flexible display screen according to an embodiment of the present invention. Figure 1 As shown, the method includes: During the repeated bending of the flexible display screen, the bending angle is recorded simultaneously, and the resistance value, AC impedance and resistance noise power are collected in multiple monitoring sections set along the bending axis. The closed loop area formed by the change of resistance with angle in each bending cycle is calculated as the hysteresis loop area and the residual resistance increment of each monitoring section. The AC impedance is used to distinguish between interface debonding and bulk damage, and the resistance noise power is used to identify microcracks. Phase space analysis constructed by residual resistance increment, hysteresis toroidal accumulation, and low-frequency AC impedance identifies the damage evolution stage and locates the damage inflection point. The section with the largest residual resistance increment growth rate and microcracks is identified as the damage concentration section. Record the spatial location migration sequence of the damage concentration area, and determine the dominant failure mode based on the changing trends of AC impedance and resistive noise power; The spatial location migration sequence and historical electrical parameter data are input into a neural network to predict the migration direction and migration probability. The acquisition frequency and detection point position are dynamically adjusted in combination with the proximity of the current cycle distance to the damage inflection point.

[0021] In one optional implementation, the steps of simultaneously recording the bending angle and acquiring resistance values, AC impedance, and resistance noise power in multiple monitoring sections set along the bending axis include: During the bending process, the bending angle change curve is acquired in real time and divided into a loading stage and an unloading stage. During the loading stage, the electrical parameters of each monitoring section are collected at a preset angle interval. The preset angle interval is determined according to the angle interval corresponding to the peak position of the resistance change rate in the previous bending cycle. During the unloading phase, the monitoring section whose resistance value has not recovered to the initial value is identified as the plastic deformation section, and the preset angle interval is reduced for the plastic deformation section in subsequent bending cycles; Establish a time-series correlation between the resistance values, AC impedance, and resistance noise power collected in each monitoring section during the loading and unloading phases and the corresponding bending angles.

[0022] For example, a bending test platform is established, including a bending mechanism, an angle sensor, and an electrical parameter acquisition unit. The bending mechanism is used to achieve controllable bending of flexible materials, the angle sensor is used to acquire the bending angle in real time, and the electrical parameter acquisition unit is used to measure the resistance value, AC impedance, and resistance noise power of each monitoring section.

[0023] Multiple monitoring segments are formed on a flexible material along the bending axis. These monitoring segments are formed by depositing conductive material on the flexible substrate, and each segment is connected to the measuring device via an independent lead. The spacing between the monitoring segments is typically five to ten millimeters, which can be adjusted according to the material size and bending characteristics.

[0024] Before starting the bending test, initial electrical parameters for each monitoring section are collected as baseline values. These include the initial resistance value R0, AC impedance Z0, and resistive noise power P0. These initial values ​​will be used to calculate the relative rate of change and determine whether the material has undergone plastic deformation.

[0025] During the bending operation, the angle sensor collects bending angle data in real time and generates a bending angle change curve. This curve records the entire process from the start of bending to the maximum bending angle, and then to the return to a straight state. Based on the angle change trend, the bending process is divided into a loading stage (angle gradually increases) and an unloading stage (angle gradually decreases).

[0026] In the first bending cycle, electrical parameter acquisition is triggered at fixed angle intervals (e.g., every five degrees). This allows us to obtain the changes in resistance, AC impedance, and resistance noise power of each monitoring section at different bending angles. Simultaneously, the relative rate of change of resistance ΔR / R0 is calculated, and its variation with the bending angle is plotted.

[0027] By analyzing the resistance change rate curve during the first bending cycle, the angular range where the change rate reaches its peak can be identified. These peaks typically correspond to areas of stress concentration or drastic microstructural changes in the material, and are angular ranges that require close monitoring. For example, if the resistance change rate is found to be the highest between 60 and 70 degrees, the sampling angle interval for this range can be set smaller in subsequent bending cycles, such as recording data every one or two degrees, while larger sampling intervals can be maintained for other angular ranges.

[0028] When the bending process enters the unloading stage, the electrical parameters of each monitoring section continue to be collected at preset angle intervals. After unloading is completed, the resistance value of each monitoring section is compared to see if it has recovered to the initial value R0. If the resistance value of a certain monitoring section fails to recover to the initial value (usually the recovery rate is set to be less than 95% as the judgment criterion), then the section is marked as a plastic deformation section.

[0029] For identified plastic deformation sections, a more detailed monitoring strategy is employed in subsequent bending cycles. Specifically, the sampling angle interval within the corresponding angle range of this section is reduced, such as decreasing the original five-degree sampling interval to one-degree interval, to capture more detailed changes in electrical parameters. At the same time, the sampling angle range is appropriately expanded to cover areas where plastic deformation may extend.

[0030] While collecting data, a time-series correlation is established between the resistance value, AC impedance, and resistance noise power of each monitoring section and the corresponding bending angle. This correlation is achieved through timestamps or angle sequence numbers, ensuring that each set of electrical parameters accurately corresponds to a specific bending angle.

[0031] By conducting multiple bending cycle tests, the relationship between electrical parameters and bending angles is established to evaluate the fatigue characteristics and service life of the material. For example, by analyzing the trend of resistance noise power changing with the number of bends, potential material failures can be predicted in advance.

[0032] The method for determining the preset angle interval can be adjusted for different materials or structures. For materials with known bending characteristics, key angle intervals can be preset based on historical data or theoretical models; for new materials, the sampling strategy for subsequent cycles can be adaptively adjusted based on data from the initial few bending cycles.

[0033] This invention, by precisely controlling the sampling strategy and focusing on key angle ranges and plastic deformation sections, can effectively evaluate the performance changes and reliability characteristics of flexible materials under bending conditions.

[0034] In one optional implementation, the steps of calculating the closed loop area formed by the resistance change with angle in each bending cycle as the hysteresis loop area and the residual resistance increment of each monitoring section, wherein the AC impedance is used to distinguish between interface debonding and bulk damage, and the resistance noise power is used to identify microcracks, include: Extract the set of data points that form a closed trajectory from the collected angle-resistance data points, and calculate the area enclosed by the closed trajectory as the hysteresis loop area by integration; The difference between the resistance value when each bending cycle returns to the initial straight state and the initial reference value is calculated as the residual resistance increment. When the residual resistance increment exceeds the preset residual threshold, microcrack detection is triggered. During the peak angle holding stage of the bending cycle that triggers microcrack detection, a high-frequency pulse current is applied to the corresponding monitoring section as a perturbation excitation. The high-frequency fluctuation of the resistance signal is collected and the power spectral density is calculated. When the power spectral density exceeds the preset power threshold, it is determined that there is a microcrack in the monitoring section. The transient recovery curve of the resistance value of the monitoring section is collected after the perturbation excitation is removed. The rapid decline stage and the slow decline stage are identified, and the rapid decay time and the slow decay time are calculated respectively. The monitoring section is determined to have entered the critical instability state based on the change characteristics of the rapid decay time and the slow decay time. Different frequencies of AC excitation are applied to the monitoring section that has entered the critical instability state and the AC impedance is measured. When the low-frequency impedance increase exceeds the preset ratio, it is determined to be interface debonding. When the wide-band impedance fluctuation exceeds the preset dispersion, it is determined to be body damage.

[0035] For example, after completing the bending cycle test and acquiring angle and resistance data, a set of angle-resistance data points forming a closed trajectory is extracted from the data records of each bending cycle. These data points include a sequence of resistance values ​​from zero degrees to the maximum bending angle during the loading phase, and a sequence of resistance values ​​from the maximum bending angle back to zero degrees during the unloading phase. Due to the hysteresis effect of the material during bending, the loading and unloading paths do not coincide, thus forming a closed loop in the angle-resistance coordinate system. This closed loop is integrated, and the area of ​​the tiny trapezoids enclosed by adjacent data points is accumulated segment by segment using the trapezoidal rule. The sum of all tiny areas yields the total area enclosed by the closed trajectory, which is the hysteresis loop area. If the intervals between the collected data points are uneven, the areas of each trapezoid are weighted according to the angle difference to ensure that the integration result accurately reflects the actual loop area. The unit of the hysteresis loop area is the product of ohms and degrees, and its value directly reflects the degree of energy dissipation and plastic deformation accumulation of the material during bending.

[0036] After each bending cycle, wait for the flexible display screen to return to its initial flat state and stabilize for 5 to 10 seconds, then record the resistance value of each monitoring section at this time as the cycle-ending resistance value. Subtract this cycle-ending resistance value from the initial baseline value collected before the test; the difference is the residual resistance increment. A positive residual resistance increment indicates irreversible degradation of the material's conductivity; the larger the value, the more severe the cumulative damage. Set a preset residual threshold of 0.5% to 2% of the initial baseline value. When the residual resistance increment of a monitoring section exceeds this threshold, trigger the microcrack detection process for that section. The selection of the trigger threshold should be based on the material type and test sensitivity requirements. A lower threshold can be used for highly conductive transparent electrode materials, while the threshold can be appropriately relaxed for flexible substrate conductive layers.

[0037] After triggering microcrack detection, the bending action is paused when the peak angle is reached in the next bending cycle, maintaining this angle unchanged. At this time, the internal stress of the material is at its maximum, and the presence of microcracks will be more pronounced. A high-frequency pulsed current is applied as a perturbation excitation to the monitored section that triggered the detection. The pulse frequency is set in the range of 1000 Hz to 10000 Hz, and the pulse amplitude is controlled at 10% to 20% of the normal test current to avoid causing additional damage to the material. The pulse duration is 1 ms to 10 ms, and the resistance signal of the monitored section is continuously acquired at a sampling rate of no less than 100000 Hz during the pulse application period. The presence of microcracks will cause high-frequency oscillations in the current path at the crack, which is reflected in the high-frequency fluctuation component of the resistance signal. The acquired resistance signal is subjected to a fast Fourier transform to convert the time-domain fluctuations to the frequency domain, and the power spectral density distribution is calculated. The power spectral density reflects the energy distribution of the signal at each frequency component. When microcracks are present inside the material, the power spectral density in the high-frequency range will increase significantly. The preset power threshold is set to 3 to 5 times the average power spectral density under no-damage conditions. If the calculated power spectral density exceeds this threshold, it is determined that there are microcracks in the monitoring section.

[0038] After microcrack detection, the high-frequency pulsed current excitation was removed, and the transient recovery curve of the resistance value of the monitored section was immediately and continuously acquired at a sampling rate of no less than 10,000 Hz. At the instant the excitation was removed, the resistance value typically exhibited a brief peak or jump, followed by a gradual decrease and stabilization. This transient recovery process reflects the dynamic characteristics of carrier redistribution and local stress release within the material. By calculating the first derivative of the transient recovery curve, the relationship between the resistance decrease rate and time was obtained. The resistance decrease rate curve typically exhibits a two-stage characteristic of being fast initially and then slow. The decrease rate reaches its peak within the first few milliseconds after excitation removal, corresponding to the rapid decrease stage. Thereafter, the decrease rate gradually decreases and approaches zero, corresponding to the slow decrease stage. The first peak point in the resistance decrease rate curve was identified, and the time corresponding to this peak point was used as the boundary between the rapid and slow decrease stages. The time interval from the excitation removal moment to the boundary point was defined as the rapid decay time, and the time interval from the boundary point to the resistance value stabilizing (the rate of change being less than 0.01% of the initial value) was defined as the slow decay time.

[0039] The resistance recovery curve during the rapid descent phase is fitted with an exponential function. The fitted form is that the resistance value equals the product of the stable value plus the initial deviation and the natural exponential function, with the independent variable of the exponential function being the negative of the ratio of time to the rapid decay time constant. The rapid decay time constant is obtained by solving the least squares method. The resistance recovery curve during the slow descent phase is fitted with a logarithmic function. The fitted form is that the resistance value equals the product of the stable value plus the initial deviation and the natural logarithmic function, with the independent variable of the logarithmic function being the ratio of time to the slow decay time constant plus one. The slow decay time constant is also obtained by solving the least squares method. The ratio of the rapid decay time constant to the slow decay time constant is calculated, and this ratio is defined as the two-stage decay characteristic parameter. As the number of bending cycles increases, the material's self-healing ability gradually decreases, the proportion of the rapid descent phase decreases, and the proportion of the slow descent phase increases, resulting in a monotonically decreasing trend in the two-stage decay characteristic parameter. The rate of decrease of the two-stage decay characteristic parameter is calculated for several consecutive cycles (e.g., 10 to 20 cycles). When the rate of decrease exceeds a preset threshold (e.g., a decrease of 5% to 10% per cycle), the material's self-healing ability is determined to have significantly decreased. The second derivative of the two-stage decay characteristic parameter with respect to the number of cycles is further calculated. A positive second derivative indicates a slowing of the decay trend, while a negative second derivative indicates an accelerating decay trend. When the second derivative changes from negative to positive, it indicates an inflection point in the rate of deterioration of the material's stability index, marking the monitoring section as entering a critical instability state.

[0040] For monitoring sections entering the critical instability state, AC excitation at different frequencies is applied during subsequent bending cycles, and the AC impedance is measured. The AC excitation frequency covers a wide frequency band from 0.1 Hz to 10000 Hz. A sinusoidal voltage or current is applied at each frequency point, and the corresponding impedance magnitude and phase angle are measured. Low-frequency impedance (e.g., below 1 Hz) mainly reflects interfacial contact resistance and charge transfer processes, and is sensitive to interfacial debonding. When interfacial debonding occurs, the contact area decreases, the interfacial resistance increases, leading to a significant increase in low-frequency impedance. The increase in low-frequency impedance relative to the initial value is calculated, and a preset ratio is set from 50% to 100%. When the increase in low-frequency impedance exceeds this preset ratio, interfacial debonding is determined to have occurred in the monitoring section. Wide-band impedance fluctuations reflect the integrity of the conductive network inside the material. When the material is cracked, has pores, or the conductive path is broken, the impedance response at different frequencies will exhibit irregular fluctuations. The ratio of the standard deviation to the mean of the impedance values ​​at each frequency point within the wide-band is calculated, and this ratio is defined as the impedance dispersion. The preset dispersion is set to 20% to 50%. When the impedance dispersion exceeds the preset value, it is determined that the monitored section has suffered damage.

[0041] This invention achieves early identification of microcracks and accurate differentiation of failure modes through multidimensional electrical parameter and transient response analysis, significantly improving the accuracy of flexible display screen bending reliability assessment.

[0042] In one optional implementation, the step of acquiring the transient recovery curve of the resistance value of the monitoring section after the perturbation excitation is removed, and determining the critical instability state, includes: At the instant the perturbation excitation is removed, high-frequency sampling and monitoring of the resistance value of the section are started to form a transient recovery curve. The first derivative of the transient recovery curve is calculated to obtain the resistance decrease rate curve. The first peak point in the resistance decrease rate curve is identified as the dividing point between the rapid decrease stage and the slow decrease stage. The time interval from the removal of the perturbation excitation to the boundary point is defined as the rapid decay time, and the time interval from the boundary point to the stabilization of the resistance value is defined as the slow decay time. The rapid decay time constant is extracted by fitting an exponential function to the rapid decay phase, and the slow decay time constant is extracted by fitting a logarithmic function to the slow decay phase. The ratio of the two time constants is calculated as a dual-stage decay characteristic parameter. When the dual-stage decay characteristic parameter decreases monotonically and the deceleration rate exceeds a preset rate threshold, it is determined that the material's self-healing ability has decreased. Calculate the second derivative of the dual-stage decay characteristic parameter with respect to the number of cycles. When the second derivative changes from a negative value to a positive value, it is determined that the growth rate of the material stability index has reached an inflection point, and the monitoring section is marked to enter a critical instability state.

[0043] For example, the data acquisition system immediately switches to high-frequency sampling mode the instant the perturbation excitation is removed, continuously recording the resistance value of the monitored section at a sampling rate of no less than 10,000 Hz. The sampling start time is defined as the moment when the perturbation excitation current drops below 10% of the rated value, triggered by the falling edge detection of the current sensor. The sampling duration is set according to the material type: 50 to 100 milliseconds for fast-response materials and 200 to 500 milliseconds for slow-response materials. The acquired resistance value data points are stored in a ring buffer in timestamp order, with a buffer capacity set to accommodate at least 5,000 data points to ensure curve smoothness. The initial segment of the transient recovery curve often contains spike interference caused by the discharge of parasitic capacitance in the circuit. Median filtering is performed on the first 5 to 10 data points after the sampling start to remove outliers exceeding ±3 times the standard deviation of the mean of adjacent points. The filtered resistance value sequence constitutes the transient recovery curve, which records the complete dynamic process of the material transitioning from a perturbed state to a stable state.

[0044] When calculating the first derivative of the transient recovery curve, the central difference method is used to improve the accuracy of numerical differentiation. For the i-th data point on the curve, its first derivative value = (resistance value of the (i+1)-th data point - resistance value of the (i-1)-th data point) ÷ (2 × time interval). The derivatives at boundary points are calculated using forward or backward difference. The calculated derivative sequence is the resistance decrease rate curve, and its unit is ohms per second. The resistance decrease rate curve reflects how quickly the resistance value changes with time; a positive value indicates an increase in resistance, a negative value indicates a decrease in resistance, and a larger absolute value indicates a more drastic change. The resistance decrease rate curve is smoothed using a five-point moving average to eliminate the interference of high-frequency noise on peak identification. In the smoothed rate curve, the first local maximum point is searched backward from the moment the excitation is removed. The absolute value of the resistance decrease rate at this point is the largest, marking the most drastic state of the rapid release phase. This local maximum point is defined as the first peak point, and its corresponding time coordinate is used as the boundary between the rapid and slow decrease phases. If there are multiple local maxima in the curve, select the point with the earliest time and an absolute rate value exceeding 50% of the global peak value as the effective dividing point.

[0045] After determining the boundary point, the time interval from the removal of the perturbation excitation to the boundary point is defined as the fast decay time. This time interval reflects the characteristic timescale of the material's internal fast response mechanism and is typically related to carrier mobility and local stress release rate. The time interval from the boundary point to the stabilization of the resistance value is defined as the slow decay time. The resistance stability criterion is that the rate of change of resistance at 20 consecutive data points is less than 0.01% of the initial resistance value, or the absolute value of the resistance decrease rate is consistently less than 1% of the peak value of the initial decrease rate. The slow decay time reflects the timescale of the slow relaxation process within the material and is typically related to molecular chain rearrangement and interfacial charge redistribution. The ratio of the fast decay time to the slow decay time provides a quantitative characterization of the two-stage response properties.

[0046] An exponential function was used to fit the resistance recovery curve during the rapid descent phase, and the fitting was solved iteratively using the least squares method. The fitting function was in the form of resistance value R = steady-state resistance value R∞ + initial deviation ΔR0 × natural exponential function, where the independent variable of the exponential function was (-t ÷ τ1). The parameters R∞, ΔR0, and τ1 were iteratively adjusted by minimizing the sum of squares of the differences between the actual resistance value and the fitted value. The initial values ​​for iteration were set as follows: R∞ equal to the average resistance at the end of the rapid descent phase, ΔR0 = (resistance value at the moment of excitation removal - R∞), and τ1 = rapid decay time ÷ 3. The iteration terminated when the parameter change in two consecutive iterations was less than 0.1% of the parameter value, or when the number of iterations exceeded 100. The rapid decay time constant τ1 obtained after the fitting converged quantified the decay rate of the rapid response process.

[0047] The resistance recovery curve during the slow descent phase is fitted using a logarithmic function, employing the least squares method. The fitting function is in the form of resistance value R = steady-state resistance value R∞ + initial deviation ΔR1 × natural logarithm, where the independent variable of the logarithmic function is ((t-t0)÷τ2+1). The parameters R∞, ΔR1, and τ2 are iteratively adjusted by minimizing the sum of squares of the differences between the actual resistance value and the fitted value. The initial values ​​for iteration are set as follows: R∞ equals the mean resistance after stabilization, ΔR1 = (resistance at the boundary point - R∞), and τ2 = slow decay time ÷ 2. The iteration termination condition is the same as for the fast phase fitting. The slow decay time constant τ2 obtained after fitting convergence quantifies the decay rate of the slow relaxation process.

[0048] The ratio of the rapid decay time constant τ1 to the slow decay time constant τ2 is calculated and defined as the two-stage decay characteristic parameter λ, i.e., λ = τ1 ÷ τ2. This parameter is dimensionless; a larger value indicates a higher proportion of the rapid process relative to the slow process, and a stronger self-healing ability of the material. As the number of bending cycles increases, damage to the internal microstructure of the material accumulates, and the rapid response mechanism gradually fails, leading to a decrease in τ1 or an increase in τ2, thus causing λ to exhibit a monotonically decreasing trend. The sliding window method is used to calculate the deceleration rate of λ, with the window length set to 10 to 20 consecutive cycles. Linear fitting is performed on the λ values ​​within the window, and the slope of the fitted line is the deceleration rate, expressed as the change in λ per cycle. A preset rate threshold is set to decrease by 5% to 10% per cycle; when the calculated deceleration rate exceeds this threshold, the material's self-healing ability is considered to have significantly decreased.

[0049] Further calculations of the second derivative of the two-stage decay characteristic parameter λ with respect to the number of cycles were performed to identify changes in the decreasing trend. The second derivative was calculated using a three-point method: for the nth cycle, the second derivative value = (λ value of the (n+1)th cycle - 2 × λ ​​value of the nth cycle + λ value of the (n-1)th cycle) ÷ (cycle interval)². A positive second derivative indicates that the rate of decrease of λ is decreasing, i.e., the decreasing trend is leveling off; a negative second derivative indicates that the rate of decrease of λ is increasing, i.e., the decreasing trend is accelerating. The sign change of the second derivative was monitored. When the second derivative changed from negative to positive for three consecutive calculations, an inflection point was identified as the rate of deterioration of the material's stability index. This inflection point marks the material's transition from the accelerated degradation stage to the critical equilibrium stage; further bending cycles may lead to sudden failure. The monitoring section where the inflection point was detected was marked as entering a critical instability state, triggering subsequent failure mode diagnosis procedures.

[0050] This invention achieves quantitative identification of the decline in material self-healing ability and critical instability state through the analysis of two-stage decay characteristic parameters and their higher-order derivatives, providing a reliable basis for the prediction of the lifetime of flexible devices.

[0051] In one optional implementation, the step of identifying damage evolution stages and locating damage inflection points through phase space analysis constructed by the residual resistance increment, hysteresis loop accumulation, and low-frequency AC impedance, and identifying the segment with the largest residual resistance increment growth rate and the presence of microcracks as the damage concentration segment includes: Extract the residual resistance increment, hysteresis loop accumulated value and low-frequency AC impedance of each bending cycle to construct three-dimensional phase space coordinate points. Calculate the Euclidean distance between adjacent cycle coordinate points as the state evolution step size. When the state evolution step size continues to increase, it is marked as entering the accelerated evolution stage. For cycles after entering the accelerated evolution stage, the coefficient of variation of the residual resistance increment and the accumulated value of the hysteresis loop within the sliding time window is calculated. When the rate of change of the coefficient of variation between cycles changes from less than the preset stability threshold to continuously increasing, it is determined to be an accelerated instability state. For a loop interval determined to be in an accelerated instability state, the mutual information between the residual resistance increment and the accumulated value of the hysteresis loop is calculated. When the decrease in the mutual information exceeds a preset decrease threshold, the loop interval is marked as having a change in damage mechanism. Within the cyclic interval where the damage mechanism changes, the information entropy of the distribution of three-dimensional phase space coordinate points is calculated, and the number of cycles in which the information entropy jump exceeds the preset entropy change threshold is identified as the damage inflection point. In the cycle corresponding to the damage inflection point and subsequent cycles, the section with the largest first derivative of the residual resistance increment and a positive second derivative is identified from the monitoring section with microcrack markings as the damage concentration section.

[0052] For example, three characteristic parameters are extracted from the completed bending cycle test data: the residual resistance increment, the hysteresis loop surface accumulation, and the low-frequency AC impedance for each bending cycle. The residual resistance increment is obtained by subtracting the initial reference value from the current resistance value at the end of a single cycle. The hysteresis loop surface accumulation is obtained by summing the hysteresis loop area of ​​the current cycle with the hysteresis loop areas of all previous cycles. The low-frequency AC impedance is the impedance magnitude measured when a 1 Hz AC excitation is applied. These three characteristic parameters are used as coordinate axes in a three-dimensional phase space. Each bending cycle corresponds to a coordinate point in the phase space. The coordinate point of the nth cycle is recorded as three components: the residual resistance increment, the hysteresis loop surface accumulation, and the low-frequency AC impedance for that cycle. The data structure of the phase space coordinate point includes a cycle number, three coordinate components, and a timestamp field, and is stored in a relational database or time-series database to support efficient querying and range scanning.

[0053] The Euclidean distance between adjacent cycle coordinate points is calculated as the state evolution step size. For the nth cycle and the (n+1th cycle), the state evolution step size is equal to the square root of the sum of the squares of the differences between the three coordinate components. Specifically, it is the square of the difference between the residual resistance increments of the (n+1th)th and nth cycles, the square of the difference between the accumulated hysteresis loops of the (n+1th)th and nth cycles, and the square root of the sum of these three squares. Since the dimensions and numerical ranges of the three coordinate axes differ significantly, normalization is required to avoid any one dimension dominating the distance calculation. The normalization method uses maximum-minimum scaling, mapping each coordinate component to the interval between 0 and 1. The normalized value is equal to the original value minus the minimum value of that coordinate component across all cycles, and the difference is divided by the difference between the maximum and minimum values ​​of that coordinate component. The Euclidean distance calculated after normalization reflects the magnitude of the material state's movement in phase space; a larger distance indicates a more drastic state change. A linear fit was performed on the state evolution step size for several consecutive cycles (e.g., 5 to 10 cycles). The slope of the fitted line reflects the increasing trend of the step size with the number of cycles. When the fitted slope is positive for three consecutive windows and the value exceeds 0.01, it is determined that the state evolution step size continues to increase, marking the material entering the accelerated evolution stage. The initial cycle number of the accelerated evolution stage is recorded as n_accel, and all data points after this cycle are included in subsequent analyses.

[0054] For cycles entering the accelerated evolution phase, the sliding time window length is set to 8 to 15 consecutive cycles, with the window sliding in single-cycle steps. Within each window, the coefficients of variation for the residual resistance increment and the accumulated hysteresis loop are calculated. The coefficient of variation is defined as the ratio of the standard deviation to the mean. The standard deviation is calculated by first summing the squares of the differences between each data point within the window and the mean, then dividing this sum by the number of data points and taking the square root. The standard deviation divided by the mean is the coefficient of variation. The coefficient of variation is dimensionless; a larger value indicates more significant data fluctuations relative to the mean. For each sliding window, the coefficients of variation for the residual resistance increment and the accumulated hysteresis loop are calculated, and the square root of their product is taken as the composite coefficient of variation. The difference between the composite coefficients of variation of adjacent windows is calculated and divided by the window sliding step size to obtain the inter-cycle rate of change of the coefficient of variation. A preset stability threshold is set to 0.005 to 0.01. When the inter-cycle rate of change of the coefficient of variation changes from being less than the preset stability threshold to being greater than this threshold for three consecutive windows and continuing to increase, the material is determined to have entered an accelerated instability state. The initial loop count for accelerating the instability state is recorded as n_unstable.

[0055] For cyclic intervals identified as accelerated instability states, the mutual information between the residual resistance increment and the accumulated hysteresis loop is calculated to quantify the statistical dependency between the two variables. The mutual information calculation requires first discretizing the continuous variables. An equal-frequency binning method is used to divide both the residual resistance increment and the accumulated hysteresis loop into 5 to 10 intervals, ensuring that each interval contains approximately the same number of data points. After discretization, a two-dimensional joint probability distribution matrix is ​​constructed. The matrix elements represent the probability that both the residual resistance increment and the accumulated hysteresis loop fall into a certain interval; this probability is equal to the number of data points that simultaneously satisfy both conditions divided by the total number of data points. Simultaneously, the marginal probability distributions of the residual resistance increment and the accumulated hysteresis loop are calculated. The marginal probability is equal to the number of data points corresponding to the variables falling into a certain interval divided by the total number of data points. Mutual information is calculated by traversing all interval combinations. For each interval combination, the following operations are performed: first, calculate the product of the two marginal probabilities of the combination; then, divide the joint probability of the combination by the aforementioned product of marginal probabilities to obtain the ratio; take the natural logarithm of this ratio; multiply the logarithmic result by the joint probability of the combination; finally, sum the products obtained from all interval combinations to obtain the mutual information value. A positive mutual information value indicates a positive correlation or nonlinear dependency between the two variables; the larger the value, the stronger the dependency. Mutual information is calculated using a sliding window for cycles within the accelerated instability interval, with a window length set to 10 to 20 cycles. The decrease in mutual information of the current window relative to the initial window mutual information is calculated; the decrease is equal to the initial window mutual information minus the current window mutual information, divided by the initial window mutual information, and then multiplied by 100%. A preset decrease threshold of 30% to 50% is set. When the decrease in mutual information exceeds this threshold, a damage mechanism shift is marked for that cycle interval. A damage mechanism shift means that the correlation between the residual resistance increment and the accumulated value of the hysteresis torus weakens, and material damage shifts from a single dominant mode to multi-mode coupling.

[0056] Within the cyclic intervals where damage mechanisms transition, the information entropy of the three-dimensional phase space coordinate point distribution is calculated to quantify the degree of disorder in the state distribution. The three-dimensional phase space is divided into a cubic grid, with the grid side length adaptively determined according to the data distribution range, resulting in 8 to 12 intervals in each dimension. The number of coordinate points falling into each grid cell is counted, and the probability of each cell is calculated as the number of points falling into that cell divided by the total number of cycles within the interval. The information entropy is obtained by traversing all non-empty grid cells, multiplying the probability of each cell by the negative value of the natural logarithm of that probability, and summing the product of all cells. A larger information entropy value indicates a more dispersed distribution of coordinate points and a more irregular state evolution path. Information entropy is calculated using a sliding window for cycles within the damage mechanism transition interval, with the window length set to 8 to 15 cycles. The difference in information entropy between adjacent windows is calculated as the change in information entropy, identifying the cycle positions where the change in information entropy jumps. A preset entropy change threshold is set to 15% to 25% of the average information entropy. When the increase in information entropy of a cycle relative to the previous cycle exceeds this threshold, the cycle number is identified as the damage inflection point. The damage inflection point marks a qualitative change in the material's damage mode, and subsequent evolution enters a new stage.

[0057] In the cycle corresponding to the damage inflection point and subsequent cycles, damage-concentrated sections are selected from the monitored sections marked with microcracks. Microcrack markers are obtained through previous power spectral density detection and stored in the monitored section status table. For each monitored section with a microcrack marker, the residual resistance increment sequence for that section at the damage inflection point and for the subsequent 10 to 20 cycles is extracted. The first derivative of the residual resistance increment is calculated. For the nth cycle, the first derivative equals the residual resistance increment of the (n+1)th cycle minus the residual resistance increment of the (n-1)th cycle, divided by twice the cycle interval. The first derivative reflects the growth rate of the residual resistance increment; a larger value indicates faster damage accumulation. The second derivative of the residual resistance increment is further calculated. For the nth cycle, the second derivative = (residual resistance increment of the (n+1)th cycle - 2 × residual resistance increment of the nth cycle + residual resistance increment of the (n-1)th cycle) ÷ (cycle interval)². A positive second derivative indicates an accelerating growth rate, while a negative value indicates a slowing growth rate. Among all monitoring sections with microcrack markings, the section with the largest first derivative and a positive second derivative is identified as the damage concentration section. If multiple sections meet the criteria, they are sorted in descending order of first derivative value, and the top three sections are selected as the main damage concentration sections. The identification results of the damage concentration sections include the section number, first derivative value, second derivative value, and the corresponding cycle number range, which are stored in the damage analysis results table for subsequent decision-making.

[0058] This invention uses phase space trajectory analysis and information theory methods to achieve precise division of damage evolution stages and quantitative identification of turning points, providing theoretical and data support for early warning of failure in flexible devices.

[0059] In one optional implementation, the steps of recording the spatial location migration sequence of the damage concentration segment and determining the dominant failure mode based on the changing trends of AC impedance and resistive noise power include: The spatial location change of the damage concentration section is tracked in a continuous bending cycle. When the damage concentration section migrates from the first monitoring section to the second monitoring section, the migration event is recorded. The continuous migration events are arranged in chronological order to form a spatial location migration sequence. The frequency of occurrence of interface debonding judgment results and body damage judgment results corresponding to each monitoring segment in the spatial location migration sequence is statistically analyzed, and the failure mode corresponding to the judgment result with higher occurrence frequency is determined as the dominant failure mode. The dominant failure mode is cross-validated with the resistance noise power change trend of the corresponding monitoring section. When the dominant failure mode is interface debonding and there is no significant pulse change in resistance noise power, the judgment result is confirmed. When the dominant failure mode is body damage and there is a continuous pulse change in resistance noise power, the judgment result is confirmed.

[0060] For example, a damage concentration section tracking mechanism is established during continuous bending cycle testing. This mechanism dynamically updates the damage concentration section identifier based on the output of the damage concentration section identification module. The damage concentration section identification module performs a judgment after each bending cycle, outputting a record containing the damage concentration section number, cycle count, and judgment timestamp. The damage concentration section number output in each judgment is compared with the damage concentration section number from the previous cycle. If the two numbers are inconsistent, a migration event detection is triggered. The migration event record includes the source section number, target section number, the number of cycles in which the migration occurred, and the first derivative of the residual resistance increment at the time of migration. The data structure is designed as a relational table, containing a migration event sequence number, a source section field, a target section field, a cycle count field, a timestamp field, and a derivative value field, supporting indexed queries by cycle count and range scanning. When a damage concentration section migrates from the first monitoring section to the second monitoring section, the system automatically inserts a migration event record, filling the source section field with the first monitoring section number and the target section field with the second monitoring section number. All migration event records are arranged in ascending order of cycle number to form a spatial location migration sequence. This sequence fully records the transfer path and temporal evolution of damage concentration phenomena between different monitoring sections.

[0061] The analysis of spatial location migration sequences needs to be combined with the failure mode determination results of each monitoring segment. In the aforementioned AC impedance measurement stage, the system has provided interface debonding determination results or body damage determination results for each monitoring segment after a specific number of cycles. The interface debonding determination results are stored with the number of determination cycles, segment number, and low-frequency impedance increase value; the body damage determination results are stored with the number of determination cycles, segment number, and impedance dispersion value. Each migration event record in the spatial location migration sequence is traversed, and the target segment number and corresponding number of cycles for that migration event are extracted. The determination records for that segment within three to five cycles before and after that number of cycles are queried in the failure mode determination result table. If an interface debonding determination result is found, the event is marked as an interface debonding type; if a body damage determination result is found, it is marked as a body damage type. If both determination results exist simultaneously, the one with the most recent timestamp is selected. The number of occurrences of interface debonding type migration events and body damage type migration events in the spatial location migration sequence are counted, and the sum of the occurrences of the two types is the total number of migration events. The frequency of interface debonding is calculated as follows: (Number of interface debonding migration events ÷ Total number of migration events) × 100%. The frequency of body damage is calculated as follows: (Number of body damage migration events ÷ Total number of migration events) × 100%. The failure mode corresponding to the more frequent occurrence is determined as the dominant failure mode. When the frequency of interface debonding exceeds 50% and is greater than the frequency of body damage, the dominant failure mode is determined to be interface debonding. When the frequency of body damage exceeds 50% and is greater than the frequency of interface debonding, the dominant failure mode is determined to be body damage. If the difference in frequency between the two types is less than 10%, it is determined to be a mixed failure mode, requiring further analysis.

[0062] After identifying the dominant failure mode, cross-validation using the trend of resistance noise power variation is necessary to improve the reliability of the judgment. Resistance noise power data is derived from the power spectral density calculation results of the high-frequency pulse excitation response during the microcrack detection process. All monitoring segment numbers involved in the spatial migration sequence are extracted, and the resistance noise power time series for each segment is extracted for each cycle throughout the entire test period. The resistance noise power time series is a one-dimensional array, with the array index representing the cycle number and the array element representing the peak power spectral density measured for the corresponding cycle or the integrated power within a specified frequency band. Pulse abrupt change detection is performed on the time series. The detection method involves calculating the absolute value of the difference between adjacent cycle power values. When the absolute value of the difference exceeds 30% to 50% of the average power of the previous ten cycles, it is marked as a significant pulse abrupt change. The number of occurrences of significant pulse abrupt changes in the entire time series is counted. When the proportion of occurrences to the total number of cycles exceeds 5%, it is determined that there is a continuous pulse abrupt change in resistance noise power; when this proportion is less than 2%, it is determined that there is no significant pulse abrupt change in resistance noise power.

[0063] The cross-validation rule involves logically matching the dominant failure mode (DFM) determination result with the trend of resistance noise power variation. When the DFM is interface debonding, theoretically, the contact resistance increases slowly during debonding while the conductive path continuity is maintained, and the resistance noise power should not exhibit frequent high-amplitude pulses. The validation logic checks whether there are significant pulse abrupt changes in the resistance noise power. If there are indeed no significant pulse abrupt changes, the DFM determination result is confirmed as interface debonding, and the confirmation flag is written into the failure mode analysis result table. If the DFM is interface debonding but the resistance noise power exhibits continuous pulse abrupt changes, a conflict is identified, triggering a manual review process or lowering the decision confidence level to below 60%. When the DFM is body damage, theoretically, body cracking or conductive network breakage would lead to abrupt changes in the current path, and the resistance noise power should exhibit intermittent or continuous pulse characteristics. The validation logic checks whether there are continuous pulse abrupt changes in the resistance noise power. If there are indeed continuous pulse abrupt changes, the DFM determination result is confirmed as body damage, and the confirmation flag is written into the failure mode analysis result table. If the dominant failure mode is body damage but there is no significant pulse change in resistance noise power, a decision conflict is also determined and a review is triggered or the confidence level is reduced.

[0064] The failure mode analysis results table's data structure includes the test batch number, dominant failure mode field, frequency field, cross-validation result field, confidence level field, and recommended action field. The dominant failure mode field is categorized as interface debonding, bulk damage, or a hybrid failure. The frequency field records the percentage of this mode in the migration sequence. The cross-validation result field is categorized as confirmed, conflicting, or pending. The confidence level field records the degree of confidence in the judgment, ranging from 0% to 100%. The recommended action field automatically fills in the corresponding process improvement direction based on the dominant failure mode, such as optimizing the interface bonding process or improving the toughness of the substrate material. Analysis results are persistently stored, supporting historical queries and batch comparisons. The data retention period is no less than two years to support long-term trend analysis.

[0065] This invention achieves high-confidence identification of dominant failure modes through spatial location migration sequence analysis and multi-source signal cross-validation, providing precise targeting for improving the reliability of flexible devices.

[0066] In one optional implementation, the steps of inputting the spatial location migration sequence and historical electrical parameter data into a neural network to predict the migration direction and migration probability, and dynamically adjusting the acquisition frequency and detection point position based on the proximity of the current cycle distance to the damage inflection point, include: The migration direction label and the AC impedance frequency response curve and the time-domain fluctuation characteristics of the resistance noise power corresponding to the migration time in the spatial location migration sequence are extracted as input features. The migration probability distribution of each monitoring segment becoming the next damage concentration segment is output through the neural network, and the monitoring segment with the highest probability is identified from the migration probability distribution as the predicted migration direction. Calculate the difference between the accumulated value of the hysteresis torus corresponding to the current cycle and the accumulated value of the hysteresis torus corresponding to the damage inflection point. When the difference is less than a preset difference threshold and the growth rate sequence shows an upward trend, increase the sampling frequency to a preset multiple of the initial sampling frequency. The monitoring section corresponding to the predicted migration direction and its adjacent monitoring sections along the bending axis are marked as high-risk sections. In subsequent bending cycles, the resistance value, AC impedance, and resistance noise power acquisition density of the high-risk sections are increased to a preset multiple of those of other monitoring sections.

[0067] For example, migration direction labels are extracted from the spatial location migration sequence data table. These labels are defined as the spatial relationship between the target segment number and the source segment number during a migration event. If the flexible display monitoring segments are numbered sequentially from segment 1 to segment N along the bending axis, the migration direction is marked as positive when the target segment number is greater than the source segment number, negative when the target segment number is less than the source segment number, and skipped migration when the target segment and source segment are not adjacent. Historical electrical parameter data corresponding to each migration event is extracted, including the AC impedance frequency response curves and the time-domain fluctuation characteristics of resistance noise power for the three to five cycles prior to the migration event. The AC impedance frequency response curves are two-dimensional arrays, with row indices representing cycle numbers and column indices representing frequency points. Array elements represent the impedance magnitudes of the corresponding cycles and frequency points, covering 20 to 50 logarithmically uniformly distributed points within the range of 0.1 Hz to 10000 Hz. The time-domain fluctuation characteristics of resistor noise power are extracted by using a sliding window statistical method on the original resistor noise power time series. The window length is set to 5 cycles. The extracted features include the power mean, standard deviation, peak-to-peak value, skewness coefficient, and kurtosis coefficient within the window, forming a five-dimensional feature vector.

[0068] The neural network model employs a fully connected feedforward architecture, with the number of input layer nodes equal to the sum of the feature dimensions of a single migration event. The AC impedance frequency response curve is flattened into a one-dimensional vector. If 30 frequency points are used and the data from the first 3 cycles of migration are extracted, the vector length is 30 × 3 = 90. The length of the resistance noise power time-domain fluctuation feature vector is 5, and the migration direction label, after one-hot encoding, has a length of 3. The total number of input layer nodes is 90 + 5 + 3 = 98. Two hidden layers are set: the first hidden layer has 64 to 128 nodes, and the second hidden layer has 32 to 64 nodes. The activation function is a modified linear unit. The number of output layer nodes equals the total number of monitored segments N. Each node corresponds to the migration probability of a monitored segment becoming the next damage concentration segment. The softmax function is used for the output layer activation function to ensure that the sum of the output probabilities of all nodes is 1. The training dataset is constructed from the spatial location migration sequences of historical test batches. Each migration event record generates a training sample, and the sample label is the one-hot encoded vector of the target segment of the migration event. Training employs a cross-entropy loss function and an adaptive moment estimation optimizer. The initial learning rate is set to 0.001, the batch size is set to 16 to 32, and the number of training epochs is set to 50 to 100. An early stopping mechanism is triggered when the validation set loss does not decrease for 10 consecutive epochs. After model training, the weight parameters are saved to persistent storage. During inference, the weight parameters are loaded and the feature vector of the current loop is input. The output layer obtains the migration probability distribution of each monitoring segment. The node with the largest value is identified from the migration probability distribution, and the monitoring segment number corresponding to this node is the predicted migration direction.

[0069] The difference between the accumulated value of the hysteresis torus corresponding to the current cycle and the accumulated value of the hysteresis torus corresponding to the damage inflection point is calculated. The difference is calculated as: accumulated value of the hysteresis torus at the damage inflection point - accumulated value of the hysteresis torus in the current cycle. The accumulated value of the hysteresis torus corresponding to the damage inflection point has been recorded and stored in the damage analysis result table in the aforementioned damage inflection point identification step. The accumulated value of the hysteresis torus in the current cycle is obtained by real-time summation of the hysteresis torus areas of each cycle. A preset difference threshold is set to 10% to 20% of the accumulated value of the hysteresis torus at the damage inflection point. When the calculated difference is less than the preset difference threshold, a proximity determination is triggered. The growth rate sequence of the accumulated value of the hysteresis torus for the previous 5 to 10 cycles is further extracted. The growth rate is defined as the difference between the accumulated values ​​of the hysteresis torus in adjacent cycles divided by the cycle interval. A linear fit is performed on the growth rate sequence. When the slope of the fitted line is positive and the absolute value exceeds 0.01, the growth rate sequence is considered to show an upward trend. When the difference is less than the preset difference threshold and the growth rate sequence shows an upward trend, a sampling frequency increase mechanism is triggered. The initial sampling frequency is defined as sampling electrical parameters once after each bending cycle. A preset multiplier of 2 to 5 is set, and the increased sampling frequency = initial sampling frequency × preset multiplier. Specifically, intermediate sampling points are added during the loading and unloading phases within a single bending cycle. If the preset multiplier is 3, sampling is performed once each at 50% of the maximum angle after loading, at the maximum angle, and at 50% of the maximum angle after unloading, increasing the number of samplings per cycle from 1 to 3. The sampling frequency increase command is sent from the data acquisition control module to the resistance measurement units and impedance analyzers in each monitoring section. The control module maintains the sampling frequency status variable and reads this variable before the start of each cycle to determine the sampling strategy.

[0070] The monitoring segment number corresponding to the predicted migration direction is recorded as the predicted target segment. The adjacent monitoring segment numbers along the bending axis are queried in the monitoring segment spatial topology table. If the predicted target segment number is k, its adjacent segment number along the positive bending axis is k+1, and its adjacent segment number along the negative bending axis is k-1. The predicted target segment k and its adjacent segments k+1 and k-1 are marked as high-risk segments. This marking operation is performed by setting the risk flag field of the corresponding segment to a high-risk state in the monitoring segment status table. In subsequent bending cycles, the data acquisition control module adjusts the acquisition density of each segment based on the risk flag field. A preset multiplier is set to 3 to 8 times, and the resistance value acquisition density of high-risk segments = acquisition density of other monitoring segments × preset multiplier. If other monitoring segments acquire resistance values ​​once in a single cycle, high-risk segments acquire resistance values ​​3 to 8 times in a single cycle. Increased AC impedance acquisition density is achieved by increasing the number of frequency points measured in the sweep frequency measurement. If 20 frequency points are measured in other monitoring sections, 60 to 160 frequency points are measured in high-risk sections to obtain more refined frequency response characteristics. Increased resistance noise power acquisition density is achieved by extending the duration of the high-frequency pulse excitation and increasing the sampling rate. If the pulse duration is 5 milliseconds and the sampling rate is 100 kHz in other monitoring sections, the pulse duration is 15 to 40 milliseconds and the sampling rate is 300,000 to 800,000 Hz in high-risk sections, making the number of acquired data points 3 to 8 times that of other sections.

[0071] The triggering conditions for high-risk segment marking and acquisition density adjustment include a neural network prediction confidence threshold. High-risk marking is not triggered when the highest migration probability corresponding to the predicted migration direction is below 60%, avoiding resource waste caused by low-confidence predictions. Data storage after acquisition density adjustment needs to be expanded, allocating an independent high-speed cache for high-risk segments. The cache capacity is set to 8 to 10 times that of the standard segment to accommodate incremental data. Data transmission bandwidth needs to be increased accordingly, and the communication interface between the high-risk segment and the data acquisition control module is set to the highest priority, allocating an independent transmission channel to avoid competition for bus resources with other segments. The acquisition density recovery mechanism is triggered when no damage-concentrated segment migration occurs in the high-risk segment for 10 to 20 consecutive cycles, restoring the risk marker to a normal state and reducing the acquisition density back to the standard level.

[0072] This invention combines neural network prediction with an adaptive acquisition strategy to achieve forward-looking monitoring and optimized resource allocation of damage evolution paths, significantly improving the timeliness and accuracy of failure early warning for flexible devices.

[0073] A second aspect of the present invention provides a testing system for the correlation between the bending angle and conductivity of a flexible display screen, comprising: The data acquisition module is used to simultaneously record the bending angle and collect resistance values, AC impedance, and resistance noise power in multiple monitoring sections set along the bending axis during the repeated bending of the flexible display screen. The feature extraction module is used to calculate the closed loop area formed by the change of resistance with angle in each bending cycle as the hysteresis loop area and the residual resistance increment of each monitoring section. Among them, the AC impedance is used to distinguish between interface debonding and bulk damage, and the resistance noise power is used to identify microcracks. The damage analysis module is used to identify the damage evolution stage and locate the damage inflection point through phase space analysis constructed by the residual resistance increment, the accumulated value of the hysteresis loop and the low-frequency AC impedance. The section with the largest growth rate of the residual resistance increment and the presence of microcracks is identified as the damage concentration section. The failure mode determination module is used to record the spatial location migration sequence of the damage concentration section and determine the dominant failure mode based on the changing trends of AC impedance and resistance noise power. The adaptive control module is used to input the spatial location migration sequence and historical electrical parameter data into the neural network to predict the migration direction and migration probability, and dynamically adjust the acquisition frequency and detection point position based on the proximity of the current cycle distance to the damage inflection point.

[0074] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0075] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0076] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A test method for the correlation between bending angle and conductivity of flexible display screens, characterized in that, include: During the repeated bending of the flexible display screen, the bending angle is recorded simultaneously, and the resistance value, AC impedance and resistance noise power are collected in multiple monitoring sections set along the bending axis. The closed loop area formed by the change of resistance with angle in each bending cycle is calculated as the hysteresis loop area and the residual resistance increment of each monitoring section. The AC impedance is used to distinguish between interface debonding and bulk damage, and the resistance noise power is used to identify microcracks. Phase space analysis constructed by residual resistance increment, hysteresis toroidal accumulation, and low-frequency AC impedance identifies the damage evolution stage and locates the damage inflection point. The section with the largest residual resistance increment growth rate and microcracks is identified as the damage concentration section. Record the spatial location migration sequence of the damage concentration area, and determine the dominant failure mode based on the changing trends of AC impedance and resistive noise power; The spatial location migration sequence and historical electrical parameter data are input into a neural network to predict the migration direction and migration probability. The acquisition frequency and detection point position are dynamically adjusted in combination with the proximity of the current cycle distance to the damage inflection point.

2. The method according to claim 1, characterized in that, The steps of synchronously recording the bending angle and acquiring resistance values, AC impedance, and resistance noise power in multiple monitoring sections set along the bending axis include: During the bending process, the bending angle change curve is acquired in real time and divided into a loading stage and an unloading stage. During the loading stage, the electrical parameters of each monitoring section are collected at a preset angle interval. The preset angle interval is determined according to the angle interval corresponding to the peak position of the resistance change rate in the previous bending cycle. During the unloading phase, the monitoring section whose resistance value has not recovered to the initial value is identified as the plastic deformation section, and the preset angle interval is reduced for the plastic deformation section in subsequent bending cycles; Establish a time-series correlation between the resistance values, AC impedance, and resistance noise power collected in each monitoring section during the loading and unloading phases and the corresponding bending angles.

3. The method according to claim 1, characterized in that, The steps for calculating the closed loop area formed by the resistance change with angle in each bending cycle as the hysteresis loop area and the residual resistance increment of each monitoring section, where AC impedance is used to distinguish between interface debonding and bulk damage, and resistance noise power is used to identify microcracks, include: Extract the set of data points that form a closed trajectory from the collected angle-resistance data points, and calculate the area enclosed by the closed trajectory as the hysteresis loop area by integration; The difference between the resistance value when each bending cycle returns to the initial straight state and the initial reference value is calculated as the residual resistance increment. When the residual resistance increment exceeds the preset residual threshold, microcrack detection is triggered. During the peak angle holding stage of the bending cycle that triggers microcrack detection, a high-frequency pulse current is applied to the corresponding monitoring section as a perturbation excitation. The high-frequency fluctuation of the resistance signal is collected and the power spectral density is calculated. When the power spectral density exceeds the preset power threshold, it is determined that there is a microcrack in the monitoring section. The transient recovery curve of the resistance value of the monitoring section is collected after the perturbation excitation is removed. The rapid decline stage and the slow decline stage are identified, and the rapid decay time and the slow decay time are calculated respectively. The monitoring section is determined to have entered the critical instability state based on the change characteristics of the rapid decay time and the slow decay time. Different frequencies of AC excitation are applied to the monitoring section that has entered the critical instability state and the AC impedance is measured. When the low-frequency impedance increase exceeds the preset ratio, it is determined to be interface debonding. When the wide-band impedance fluctuation exceeds the preset dispersion, it is determined to be body damage.

4. The method according to claim 3, characterized in that, The steps for acquiring the transient recovery curve of the resistance value of the monitored section after the perturbation excitation is removed, and determining the critical instability state, include: At the instant the perturbation excitation is removed, high-frequency sampling and monitoring of the resistance value of the section are started to form a transient recovery curve. The first derivative of the transient recovery curve is calculated to obtain the resistance decrease rate curve. The first peak point in the resistance decrease rate curve is identified as the dividing point between the rapid decrease stage and the slow decrease stage. The time interval from the removal of the perturbation excitation to the boundary point is defined as the rapid decay time, and the time interval from the boundary point to the stabilization of the resistance value is defined as the slow decay time. The rapid decay time constant is extracted by fitting an exponential function to the rapid decay phase, and the slow decay time constant is extracted by fitting a logarithmic function to the slow decay phase. The ratio of the two time constants is calculated as a dual-stage decay characteristic parameter. When the dual-stage decay characteristic parameter decreases monotonically and the deceleration rate exceeds a preset rate threshold, it is determined that the material's self-healing ability has decreased. Calculate the second derivative of the dual-stage decay characteristic parameter with respect to the number of cycles. When the second derivative changes from a negative value to a positive value, it is determined that the growth rate of the material stability index has reached an inflection point, and the monitoring section is marked to enter a critical instability state.

5. The method according to claim 1, characterized in that, The steps of identifying damage evolution stages and locating damage inflection points through phase space analysis constructed using the residual resistance increment, hysteresis loop accumulation, and low-frequency AC impedance, and identifying the segment with the largest residual resistance increment growth rate and the presence of microcracks as the damage concentration segment, include: Extract the residual resistance increment, hysteresis loop accumulated value and low-frequency AC impedance of each bending cycle to construct three-dimensional phase space coordinate points. Calculate the Euclidean distance between adjacent cycle coordinate points as the state evolution step size. When the state evolution step size continues to increase, it is marked as entering the accelerated evolution stage. For cycles after entering the accelerated evolution stage, the coefficient of variation of the residual resistance increment and the accumulated value of the hysteresis loop within the sliding time window is calculated. When the rate of change of the coefficient of variation between cycles changes from less than the preset stability threshold to continuously increasing, it is determined to be an accelerated instability state. For a loop interval determined to be in an accelerated instability state, the mutual information between the residual resistance increment and the accumulated value of the hysteresis loop is calculated. When the decrease in the mutual information exceeds a preset decrease threshold, the loop interval is marked as having a change in damage mechanism. Within the cyclic interval where the damage mechanism changes, the information entropy of the distribution of three-dimensional phase space coordinate points is calculated, and the number of cycles in which the information entropy jump exceeds the preset entropy change threshold is identified as the damage inflection point. In the cycle corresponding to the damage inflection point and subsequent cycles, the section with the largest first derivative of the residual resistance increment and a positive second derivative is identified from the monitoring section with microcrack markings as the damage concentration section.

6. The method according to claim 1, characterized in that, The steps of recording the spatial location migration sequence of the damage concentration section and determining the dominant failure mode based on the changing trends of AC impedance and resistive noise power include: The spatial location change of the damage concentration section is tracked in a continuous bending cycle. When the damage concentration section migrates from the first monitoring section to the second monitoring section, the migration event is recorded. The continuous migration events are arranged in chronological order to form a spatial location migration sequence. The frequency of occurrence of interface debonding judgment results and body damage judgment results corresponding to each monitoring segment in the spatial location migration sequence is statistically analyzed, and the failure mode corresponding to the judgment result with higher occurrence frequency is determined as the dominant failure mode. The dominant failure mode is cross-validated with the resistance noise power change trend of the corresponding monitoring section. When the dominant failure mode is interface debonding and there is no significant pulse change in resistance noise power, the judgment result is confirmed. When the dominant failure mode is body damage and there is a continuous pulse change in resistance noise power, the judgment result is confirmed.

7. The method according to claim 6, characterized in that, The steps of inputting the spatial location migration sequence and historical electrical parameter data into a neural network to predict the migration direction and migration probability, and dynamically adjusting the acquisition frequency and detection point position based on the proximity of the current cycle to the damage inflection point, include: The migration direction label and the AC impedance frequency response curve and the time-domain fluctuation characteristics of the resistance noise power corresponding to the migration time in the spatial location migration sequence are extracted as input features. The migration probability distribution of each monitoring segment becoming the next damage concentration segment is output through the neural network, and the monitoring segment with the highest probability is identified from the migration probability distribution as the predicted migration direction. Calculate the difference between the accumulated value of the hysteresis torus corresponding to the current cycle and the accumulated value of the hysteresis torus corresponding to the damage inflection point. When the difference is less than a preset difference threshold and the growth rate sequence shows an upward trend, increase the sampling frequency to a preset multiple of the initial sampling frequency. The monitoring section corresponding to the predicted migration direction and its adjacent monitoring sections along the bending axis are marked as high-risk sections. In subsequent bending cycles, the resistance value, AC impedance, and resistance noise power acquisition density of the high-risk sections are increased to a preset multiple of those of other monitoring sections.

8. A testing system for the correlation between bending angle and conductivity of flexible display screens, used to implement the method described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to simultaneously record the bending angle and collect resistance values, AC impedance, and resistance noise power in multiple monitoring sections set along the bending axis during the repeated bending of the flexible display screen. The feature extraction module is used to calculate the closed loop area formed by the change of resistance with angle in each bending cycle as the hysteresis loop area and the residual resistance increment of each monitoring section. Among them, the AC impedance is used to distinguish between interface debonding and bulk damage, and the resistance noise power is used to identify microcracks. The damage analysis module is used to identify the damage evolution stage and locate the damage inflection point through phase space analysis constructed by the residual resistance increment, the accumulated value of the hysteresis loop and the low-frequency AC impedance. The section with the largest growth rate of the residual resistance increment and the presence of microcracks is identified as the damage concentration section. The failure mode determination module is used to record the spatial location migration sequence of the damage concentration section and determine the dominant failure mode based on the changing trends of AC impedance and resistance noise power. The adaptive control module is used to input the spatial location migration sequence and historical electrical parameter data into the neural network to predict the migration direction and migration probability, and dynamically adjust the acquisition frequency and detection point position based on the proximity of the current cycle distance to the damage inflection point.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images