Intelligent control method and system for probe card pressure distribution

Abstract

This invention relates to the field of probe card pressure testing technology, and more particularly to a method and system for intelligent control of probe card pressure distribution. The method involves acquiring pressure time-domain waveforms, performing feature extraction and pattern recognition to determine the contact state type, acquiring multi-modal data of pressure, temperature, and mechanical fields, and decoupling them based on coupling relationships to obtain independent physical field contribution components. A differentiated compensation strategy is generated by combining the contact state and contribution components, and multi-field compensation timing is coordinated and arranged. The probe card is divided into sub-regions, and control units are configured for local compensation, with cross-regional coordinated compensation when insufficient. Pressure response characteristics are monitored to identify faulty probes and select redundant probes for load redistribution. This invention achieves intelligent and balanced control of probe card pressure, improving test uniformity and reliability.

Description

Technical Field

[0001] This invention relates to the field of probe card pressure testing technology, and in particular to a method and system for intelligent control of probe card pressure distribution. Background Technology

[0002] In the field of semiconductor chip testing, probe cards are a key component in wafer testing, and the uniformity of their contact pressure distribution directly determines the reliability of test results and probe lifespan. Current conventional practices typically rely on mechanical leveling mechanisms and the inherent elasticity of spring probes for passive pressure distribution, maintaining the overall pressure level through periodic manual calibration or simple centralized pressure feedback systems.

[0003] Existing conventional practices have two significant drawbacks. First, the pressure monitoring and compensation mechanism does not adequately consider the effects of multi-physics coupling. During testing, the probe card is simultaneously subjected to mechanical contact forces, Joule heat generated by the test current, and changes in ambient temperature. These physical fields are coupled with each other, making it difficult for a single pressure signal to distinguish deviation components from different sources. For example, probe elongation caused by thermal expansion and pressure attenuation caused by mechanical wear are superimposed in the measurement signal, causing compensation actions based on global pressure deviation to frequently result in overcompensation or undercompensation, and even triggering new stress concentrations. Second, the independence of regional control and the static mode of fault handling lead to insufficient system robustness. Each control sub-region adjusts only based on its own local pressure. When pressure drift occurs in adjacent regions due to abnormal temperature differences or mechanical deformation, independent adjustment may exacerbate the pressure gradient at the interface, creating local overload. At the same time, the current method of handling faulty probes requires interrupting the testing process, which not only reduces testing efficiency but may also lead to batch yield fluctuations, failing to meet the real-time reliability requirements of high-density, multi-pin probe cards. Summary of the Invention

[0004] The present invention provides a method and system for intelligent control of probe card pressure distribution, which can solve the problems in the prior art.

[0005] A first aspect of the present invention provides a method for intelligent control of probe card pressure distribution, comprising: The pressure time-domain waveform of the probe contact process is collected at the pressure monitoring point of the probe card and segmented feature extraction is performed to obtain a multi-dimensional feature vector. The contact state type is obtained by pattern recognition of the multi-dimensional feature vector. Multimodal physical quantity data of pressure field, temperature field and mechanical field are collected at the monitoring points of the probe card. Based on the coupling relationship between the physical fields, the pressure deviation is decoupled from multiple sources to obtain independent physical field contribution components. Differentiated compensation strategies are generated based on the contact state type and the physical field contribution components. The compensation actions for different physical fields in the differentiated compensation strategies are then time-coordinated to obtain a multi-field coordinated compensation scheme. The probe card is divided into multiple control sub-regions and each control unit is configured to independently perform local compensation. When local compensation cannot meet the target, a cross-regional collaborative compensation scheme is generated by negotiating with adjacent control units. During the execution of the multi-field collaborative compensation scheme and the cross-regional collaborative compensation scheme, the pressure response characteristics of the probe are monitored. Based on the pressure response characteristics, faulty probes are identified and redundant probes are selected in their adjacent control sub-regions to perform load redistribution.

[0006] The pressure time-domain waveform of the probe contact process is acquired at the pressure monitoring point of the probe card, and segmented feature extraction is performed to obtain a multi-dimensional feature vector. Pattern recognition is then performed on the multi-dimensional feature vector to obtain the contact state type, including: During the probe contact with the wafer, the pressure time-domain waveform is acquired in real time. The pressure time-domain waveform is divided into the initial contact stage, the pressure stabilization stage and the overload stage according to the time process of the probe contact process. The pressure rise rate, pressure peak value and pressure fluctuation amplitude are extracted as stage features for each stage. The features of each stage are combined to form a multi-dimensional feature vector. The contact stiffness index is obtained by calculating the ratio of the pressure rise rate to the pressure peak value in the multidimensional feature vector, and the contact stability index is obtained by calculating the rate of change of the pressure fluctuation amplitude between each stage. The contact state type is determined based on the combination of the numerical ranges of the contact stiffness index and the contact stability index. When both the contact stiffness index and the contact stability index are within the normal range, it is determined to be normal contact. When the contact stiffness index decreases while the contact stability index is normal, it is determined to be poor contact caused by probe wear. When the contact stiffness index is normal while the contact stability index is abnormal, it is determined to be poor contact caused by probe position deviation. When both the contact stiffness index and the contact stability index are normal but the pressure peak value is lower than the target value, it is determined to be poor contact caused by probe deformation.

[0007] Multimodal physical quantity data of pressure field, temperature field and mechanical field are collected at the monitoring points of the probe card. Based on the coupling relationship between the physical fields, the pressure deviation is decoupled from multiple sources to obtain independent physical field contribution components, including: Pressure data, temperature data, and mechanical displacement data are simultaneously collected at the monitoring points of the probe card as multimodal physical quantity data; Based on the pressure data, the deviation between the actual pressure distribution and the target pressure distribution of the probe card is calculated to obtain the pressure deviation distribution; Based on the temperature data, a temperature-pressure coupling relationship is established for the influence of the temperature field on the thermal expansion of the probe card material, and a mechanical-pressure coupling relationship is established for the influence of the mechanical field on the deformation of the probe card structure based on the mechanical displacement data. Based on the temperature-pressure coupling relationship, the pressure deviation component caused by the temperature change represented by the temperature data is separated from the pressure deviation distribution to obtain the temperature field contribution component. Based on the mechanical-pressure coupling relationship, the pressure deviation component caused by the mechanical displacement represented by the mechanical displacement data is separated from the pressure deviation distribution to obtain the mechanical field contribution component. The remaining pressure deviation after deducting the temperature field contribution component and the mechanical field contribution component from the pressure deviation distribution is taken as the pressure field contribution component.

[0008] Based on the temperature data, a temperature-pressure coupling relationship is established for the influence of the temperature field on the thermal expansion of the probe card material; based on the mechanical displacement data, a mechanical-pressure coupling relationship is established for the influence of the mechanical field on the deformation of the probe card structure, including: A partial differential equation for heat conduction describing the spatiotemporal evolution of the temperature field inside the probe card and a partial differential equation for elasticity describing the spatiotemporal evolution of the stress field inside the probe card are established, and the temperature data and the mechanical displacement data are used as the boundary conditions of the partial differential equations at the monitoring point. The adjoint method is used to solve the partial differential equation of heat conduction in reverse to obtain the equivalent heat source distribution that leads to the current temperature field distribution. Based on the equivalent heat source distribution, the temperature field evolution trajectory at each spatial position on the probe card is calculated. According to the thermal expansion coefficient of the material, the temperature field evolution trajectory is mapped to the pressure change trajectory to establish a temperature-pressure coupling relationship. The adjoint method is used to solve the partial differential equation of elasticity in reverse to obtain the equivalent body force distribution that leads to the current stress field distribution. Based on the equivalent body force distribution, the stress field evolution trajectory at each spatial position on the probe card is calculated. The stress field evolution trajectory is converted into a pressure change trajectory to establish a mechanical-pressure coupling relationship.

[0009] A differentiated compensation strategy is generated based on the contact state type and the physical field contribution components. The compensation actions for different physical fields within the differentiated compensation strategy are then time-coordinated to obtain a multi-field coordinated compensation scheme, including: Mechanical compensation strategies to increase probe pressure for poor contact caused by probe wear in contact state types; mechanical compensation strategies to adjust probe position for poor contact caused by probe position offset in contact state types; and mechanical compensation strategies to reduce probe load for poor contact caused by probe deformation in contact state types. The adjustment ranges of the temperature field compensation action, mechanical field compensation action, and pressure field compensation action are determined based on the amplitudes of the temperature field contribution component, mechanical field contribution component, and pressure field contribution component, respectively. The mechanical compensation strategy is combined with the adjustment range of each physical field compensation action to form a differentiated compensation strategy; The influence time of temperature field compensation, mechanical field compensation, and pressure field compensation on the probe card is analyzed. Based on the influence time, the execution order and execution interval of each compensation action are determined. The compensation actions are then arranged in sequence according to the execution order and execution interval to obtain a multi-field collaborative compensation scheme.

[0010] The probe card is divided into multiple control sub-regions, and control units are configured to independently perform local compensation. When local compensation fails to meet the target, a cross-regional collaborative compensation scheme is generated through negotiation with adjacent control units, including: Based on the spatial distribution of probes on the probe card, the probe card is divided into multiple control sub-regions, and a control unit is configured for each control sub-region and a communication connection is established between adjacent control units. The control unit performs local compensation in the control sub-region according to the multi-field collaborative compensation scheme, collects pressure distribution data in the control sub-region and calculates the residual between the pressure distribution and the target pressure distribution. When the residual exceeds the allowable range, it is determined that the local compensation cannot meet the target. The control unit constructs a compensation demand vector containing residual spatial distribution characteristics and the pressure gradient at the boundary of the control sub-region, and sends it to the adjacent control unit; it identifies the pressure transmission direction based on the pressure gradient difference at the boundary between the control sub-region and the adjacent control sub-region, and calculates the pressure coupling coefficient when the control sub-region applies compensation to the adjacent control sub-region and the pressure coupling coefficient when the adjacent control sub-region applies compensation to the control sub-region. Based on the compensation demand vector and pressure coupling coefficient, a cross-regional compensation optimization objective is established to minimize the sum of residuals in each control sub-region. The compensation adjustment amount of each control sub-region is obtained by solving the cross-regional compensation optimization objective, which serves as a cross-regional collaborative compensation scheme.

[0011] During the execution of the multi-field collaborative compensation scheme and the cross-regional collaborative compensation scheme, the pressure response characteristics of the probes are monitored. Based on the pressure response characteristics, faulty probes are identified, and redundant probes are selected in their adjacent control sub-regions to perform load redistribution, including: During the execution of the multi-field collaborative compensation scheme and the cross-regional collaborative compensation scheme, the pressure change curves of each probe are recorded, and the pressure response speed, pressure overshoot amplitude and pressure stabilization time are extracted from the pressure change curves as pressure response characteristics. The pressure response characteristics of each probe are compared with the expected pressure response characteristics. When the pressure response speed of a probe is lower than the expected response threshold, the pressure overshoot exceeds the expected range, or the pressure stabilization time exceeds the expected stabilization threshold, the probe is identified as a faulty probe. Obtain candidate redundant probes in the adjacent control sub-region of the fault probe, and calculate the remaining load capacity of each candidate redundant probe and the amount of pressure field disturbance generated after bearing the load of the fault probe. The redundant probe selection optimization objective is established to minimize the total amount of pressure field disturbance under the constraint that the remaining load capacity meets the load transfer requirements. The target redundant probes and the load allocation of each target redundant probe are obtained by solving the redundant probe selection optimization objective. Load adjustment instructions are sent to each target redundant probe according to the load allocation to complete the load redistribution.

[0012] A second aspect of the present invention provides a probe card pressure distribution intelligent control system, comprising: The state recognition unit is used to collect the pressure time-domain waveform of the probe contact process at the pressure monitoring point of the probe card and perform segmented feature extraction to obtain a multi-dimensional feature vector. The multi-dimensional feature vector is then used for pattern recognition to obtain the contact state type. The multi-source decoupling unit is used to collect multi-modal physical quantity data of pressure field, temperature field and mechanical field at the monitoring point of the probe card, and to perform multi-source decoupling of pressure deviation based on the coupling relationship between physical fields to obtain independent physical field contribution components. The compensation orchestration unit is used to generate differentiated compensation strategies based on the contact state type and the physical field contribution components, and to perform time-series collaborative orchestration of the compensation actions for different physical fields in the differentiated compensation strategies to obtain a multi-field collaborative compensation scheme. The regional coordination unit is used to divide the probe card into multiple control sub-regions and configure the control units to independently perform local compensation. When the local compensation cannot meet the target, it negotiates with adjacent control units to generate a cross-regional collaborative compensation scheme. The fault reconfiguration unit is used to monitor the pressure response characteristics of the probes during the execution of the multi-field collaborative compensation scheme and the cross-regional collaborative compensation scheme, identify faulty probes based on the pressure response characteristics, and select redundant probes in their adjacent control sub-regions to perform load redistribution.

[0013] 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.

[0014] 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.

[0015] The system monitors the pressure time-domain waveform during probe contact in real time, extracts segmented features, and performs pattern recognition to accurately determine the contact state type, providing a precise basis for subsequent compensation. It collects multi-modal physical quantity data of pressure, temperature, and mechanical fields, performs multi-source decoupling based on the physical field coupling relationship, independently separates the contribution components of each physical field to pressure deviation, eliminates coupling interference, and improves the accuracy and reliability of pressure control.

[0016] Differentiated compensation strategies are generated based on different contact state types and physical field contribution components. Compensation actions for different physical fields are then time-coordinated and arranged to form a multi-field collaborative compensation scheme. This avoids conflicts between compensation actions of different physical fields, achieving synchronous and coordinated optimization, and significantly improving pressure distribution uniformity and contact stability. The probe card is divided into multiple control sub-regions, and control units are configured to independently execute local compensation, achieving rapid local response. When local compensation is insufficient, it negotiates with adjacent control units to generate a cross-regional collaborative compensation scheme, balancing local flexibility with global coordination and adapting to complex operating conditions.

[0017] During the execution of the compensation scheme, the probe pressure response characteristics are monitored in real time. Based on these characteristics, faulty probes are identified, and redundant probes are selected in adjacent control sub-regions to perform load redistribution. This avoids overall performance degradation due to single-point failures, improving system robustness and probe card lifespan. The entire control method achieves closed-loop control from state awareness to compensation execution to fault self-healing, enhancing the probe card's intelligent control capability of pressure distribution during wafer testing. Attached Figure Description

[0018] Figure 1 A flowchart illustrating the intelligent control method for probe card pressure distribution; Figure 2 Flowchart for decoupling and separating contribution components in multi-physics coupling. Detailed Implementation

[0019] 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.

[0020] 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.

[0021] Figure 1This is a flowchart illustrating the intelligent control method for probe card pressure distribution according to an embodiment of the present invention.

[0022] The intelligent control method for probe card pressure distribution includes: The pressure time-domain waveform of the probe contact process is collected at the pressure monitoring point of the probe card and segmented feature extraction is performed to obtain a multi-dimensional feature vector. The contact state type is obtained by pattern recognition of the multi-dimensional feature vector. Multimodal physical quantity data of pressure field, temperature field and mechanical field are collected at the monitoring points of the probe card. Based on the coupling relationship between the physical fields, the pressure deviation is decoupled from multiple sources to obtain independent physical field contribution components. Differentiated compensation strategies are generated based on the contact state type and the physical field contribution components. The compensation actions for different physical fields in the differentiated compensation strategies are then time-coordinated to obtain a multi-field coordinated compensation scheme. The probe card is divided into multiple control sub-regions and each control unit is configured to independently perform local compensation. When local compensation cannot meet the target, a cross-regional collaborative compensation scheme is generated by negotiating with adjacent control units. During the execution of the multi-field collaborative compensation scheme and the cross-regional collaborative compensation scheme, the pressure response characteristics of the probe are monitored. Based on the pressure response characteristics, faulty probes are identified and redundant probes are selected in their adjacent control sub-regions to perform load redistribution.

[0023] In one optional implementation, the pressure time-domain waveform of the probe contact process is acquired at the pressure monitoring point of the probe card, and segmented feature extraction is performed to obtain a multi-dimensional feature vector. Pattern recognition is then performed on the multi-dimensional feature vector to obtain the contact state type, including: During the probe contact with the wafer, the pressure time-domain waveform is acquired in real time. The pressure time-domain waveform is divided into the initial contact stage, the pressure stabilization stage and the overload stage according to the time process of the probe contact process. The pressure rise rate, pressure peak value and pressure fluctuation amplitude are extracted as stage features for each stage. The features of each stage are combined to form a multi-dimensional feature vector. The contact stiffness index is obtained by calculating the ratio of the pressure rise rate to the pressure peak value in the multidimensional feature vector, and the contact stability index is obtained by calculating the rate of change of the pressure fluctuation amplitude between each stage. The contact state type is determined based on the combination of the numerical ranges of the contact stiffness index and the contact stability index. When both the contact stiffness index and the contact stability index are within the normal range, it is determined to be normal contact. When the contact stiffness index decreases while the contact stability index is normal, it is determined to be poor contact caused by probe wear. When the contact stiffness index is normal while the contact stability index is abnormal, it is determined to be poor contact caused by probe position deviation. When both the contact stiffness index and the contact stability index are normal but the pressure peak value is lower than the target value, it is determined to be poor contact caused by probe deformation.

[0024] For example, during the probe's contact with the wafer, a pressure sensor continuously records the complete curve of the contact force between the probe and the wafer pads over time at a high sampling rate, i.e., the pressure time-domain waveform. This waveform contains information about the entire process from the probe's initial contact with the pad surface to stable pressure application and the potential for overload. To extract meaningful features from this continuous waveform, it is divided into three stages according to the physical time progression of the probe contact process: the initial contact stage, the pressure stabilization stage, and the overload stage. The initial contact stage corresponds to the period from when the probe tip just touches the pad to when the pressure begins to rise rapidly; the pressure signal in this stage typically exhibits a linear or near-linear upward trend. The pressure stabilization stage corresponds to the period when the probe pressure tends to approach the set target value and remains relatively constant; the waveform in this stage should exhibit low-fluctuation, stable characteristics. The overload stage corresponds to the period when the pressure exceeds the normal operating range; the occurrence of this stage is usually related to excessive probe pressure or localized protrusions on the wafer surface.

[0025] For each stage, three types of stage features are extracted. The pressure rise rate is defined as the ratio of the pressure change to the corresponding time length within that stage, reflecting the dynamic response speed of the probe contact. The pressure peak value is defined as the maximum pressure value collected within that stage, directly characterizing the maximum force exerted by the probe on the pad during that stage. The pressure fluctuation amplitude is defined as the difference between the maximum and minimum pressure signal values ​​within that stage, used to measure the stability of the pressure during that stage. The pressure rise rate, pressure peak value, and pressure fluctuation amplitude of the initial contact stage, pressure stabilization stage, and overload stage are concatenated in a fixed order to form a multi-dimensional feature vector covering nine components, which fully describes the dynamic behavior of a single probe contact process.

[0026] After obtaining the multidimensional eigenvectors, two comprehensive indices are further calculated to support the determination of the contact state. The contact stiffness index is... The calculation method is as follows: take the pressure rise rate during the initial contact phase. With the peak pressure of this stage The ratio, i.e. The physical meaning of this indicator is that when the probe tip material or geometry wears down, the displacement required for a unit pressure change increases, leading to a decrease in the rate of pressure rise relative to the pressure peak, thus... Decrease; while when the probe is in good condition Maintained within the normal range. Contact stability index The calculation method is as follows: take the rate of change of pressure fluctuation amplitude between the pressure stabilization stage and the initial contact stage, i.e. ,in This refers to the pressure fluctuation amplitude during the pressure stabilization phase. This represents the pressure fluctuation amplitude during the initial contact phase. When the probe position shifts, the contact area and contact angle between the probe tip and the pad change, causing abnormal jitter during the pressure stabilization phase. The values ​​will deviate significantly from the normal range; while under normal contact conditions, It should be within a small numerical range, indicating a smooth transition from initial contact to stable pressure.

[0027] based on and The combination of numerical ranges can determine the contact state type. By combining finite element simulation with factory calibration data of typical probe cards, a theoretical calculation model for contact stiffness and contact stability indices is established. Physical parameters such as the elastic modulus of the probe material, tip geometry, and pad material hardness are substituted into the model to calculate the contact state under normal contact conditions. and The theoretical value; based on this, considering material batch differences and manufacturing tolerances, a range of ±15% of the theoretical value is set as the upper and lower limits of the normal range, respectively denoted as... and . when and When both pressure peaks fall within their respective normal ranges and the peak pressures at each stage meet the target settings, the probe is considered to be in normal contact and requires no compensation adjustment. Below and If the pressure response per unit displacement during contact is still within the normal range, it indicates that the probe's pressure response is low. The corresponding physical reason is that the probe tip wears down due to repeated friction, resulting in a decrease in its effective contact stiffness. In this case, it is determined to be poor contact caused by probe wear. For this situation, subsequent compensation strategies should focus on compensating for the amount of probe tip wear, such as appropriately increasing the probe's downward pressure to compensate for the insufficient effective contact force caused by wear.

[0028] when Within the normal range but Exceeding If the pressure signal fluctuates abnormally during the stabilization phase, while the stiffness response is normal during the initial contact phase, the underlying physical cause is a lateral or longitudinal displacement of the probe during installation or use. This displacement causes the contact point between the probe tip and the pad to deviate from its designed position, resulting in uneven force distribution at the contact interface and causing periodic or random fluctuations in the pressure signal. This is considered a poor contact due to probe misalignment. This condition requires adjusting the probe's alignment or introducing lateral position correction into the compensation scheme to restore normal contact.

[0029] when and If all values ​​are within the normal range, but the peak pressure at each stage is lower than the preset target value, it indicates that the probe's stiffness response and stability during contact are normal, but the actual output contact force is insufficient. The corresponding physical reason is that the probe body has undergone elastic deformation or plastic bending during long-term use, resulting in a decrease in the elastic recovery force generated under the same downward pressure. In this case, it is determined to be poor contact caused by probe deformation. To address this situation, the compensation strategy needs to perform incremental compensation specifically for the downward pressure without changing the position and stiffness adjustment direction, to ensure that the actual contact force reaches the target value.

[0030] The above four contact state types form a complete multi-dimensional feature-driven pattern recognition framework. By decomposing the continuous pressure time-domain waveform into three physically meaningful stages, extracting representative statistical features at each stage, and further calculating comprehensive indicators, the framework can accurately distinguish contact anomalies caused by different physical factors. This staged, indicator-based feature extraction method avoids the computational complexity issues caused by excessively high signal dimensions when directly using the original waveform for pattern recognition, and also avoids the ambiguity of single feature indicators when facing complex faults. In actual probe card testing, the above feature extraction and state determination process is automatically triggered after each probe contact action. The determination result serves as the input for generating subsequent differentiated compensation strategies, ensuring that the compensation action accurately corresponds to the physical cause of the contact anomaly.

[0031] In one optional implementation, multimodal physical quantity data of pressure field, temperature field, and mechanical field are collected at the monitoring points of the probe card. Based on the coupling relationship between the physical fields, multi-source decoupling of the pressure deviation is performed to obtain independent physical field contribution components, including: Pressure data, temperature data, and mechanical displacement data are simultaneously collected at the monitoring points of the probe card as multimodal physical quantity data; Based on the pressure data, the deviation between the actual pressure distribution and the target pressure distribution of the probe card is calculated to obtain the pressure deviation distribution; Based on the temperature data, a temperature-pressure coupling relationship is established for the influence of the temperature field on the thermal expansion of the probe card material, and a mechanical-pressure coupling relationship is established for the influence of the mechanical field on the deformation of the probe card structure based on the mechanical displacement data. Based on the temperature-pressure coupling relationship, the pressure deviation component caused by the temperature change represented by the temperature data is separated from the pressure deviation distribution to obtain the temperature field contribution component. Based on the mechanical-pressure coupling relationship, the pressure deviation component caused by the mechanical displacement represented by the mechanical displacement data is separated from the pressure deviation distribution to obtain the mechanical field contribution component. The remaining pressure deviation after deducting the temperature field contribution component and the mechanical field contribution component from the pressure deviation distribution is taken as the pressure field contribution component.

[0032] Combination Figure 2 The flowchart illustrating multiphysics coupling decoupling and contribution component separation explains that at each monitoring point of the probe card, a synchronous acquisition network composed of pressure sensors, temperature sensors, and displacement sensors acquires pressure data, temperature data, and mechanical displacement data at the same sampling time, ensuring strict alignment of the three physical quantities in the time dimension. Pressure data reflects the actual contact force between the probe and the pads of the device under test at each monitoring point; temperature data reflects the thermal distribution of the probe card substrate, probe body, and surrounding structure; and mechanical displacement data reflects the structural deformation of the probe card at various locations under test loads and installation constraints. After synchronous acquisition, the three data streams are stored in a buffer for subsequent multi-source decoupling steps. The synchronous acquisition frequency must meet the pressure response bandwidth requirements, typically set to at least twice the cutoff frequency of the pressure sensor to avoid data distortion caused by frequency aliasing.

[0033] Based on the collected pressure data, the actual pressure distribution at each monitoring point is subtracted from the pre-calibrated target pressure distribution point by point to obtain the pressure deviation distribution. Let the total number of monitoring points be... , No. The actual pressure at each monitoring point is The target pressure is The pressure deviation at that point is The deviations at all monitoring points constitute the pressure deviation distribution vector. This deviation vector encompasses the combined effects of multiple sources, including temperature changes, mechanical deformation, and pure pressure field fluctuations. Further decoupling is required to develop targeted compensation strategies.

[0034] The temperature-pressure coupling relationship is established based on the thermal expansion characteristics of the probe card material. When the temperature of the probe card substrate or probe body changes, the thermal expansion of the material alters the effective length and contact angle of the probe, thus affecting the contact pressure. Based on temperature data, the change in temperature relative to the reference temperature within the neighborhood of each monitoring point is extracted. Combined with the linear expansion coefficient of the probe card material and the probe's geometric parameters (including probe length) and equivalent elastic modulus A model for estimating the pressure deviation component caused by temperature changes is established. Specifically, the change in the effective probe length caused by temperature changes is... This length change is achieved through the equivalent stiffness of the probe. This is converted into pressure change, i.e., the effect of the temperature field on the first... The pressure contribution component of each monitoring point is When a temperature gradient exists on the probe card, the thermal expansion of different regions is different, resulting in spatial differences in the temperature field contribution components of each monitoring point. This non-uniformity is the core content of temperature-pressure coupling relationship modeling.

[0035] The establishment of the mechanical-pressure coupling relationship is based on the deformation response of the probe card structure under external loads and mounting boundary conditions. Mechanical displacement data reflects the structural bending and warping deformation of the probe card substrate during testing, caused by factors such as the concentrated load of the probe array, the constraint reaction force of the mounting fixture, and the motion error of the testing machine. Let the first... The mechanical displacement measurements at each monitoring point are as follows: This displacement represents the normal deflection of the substrate at that point. The contact pressure of the probes is highly sensitive to mechanical displacement: when the substrate deflects, the actual interference fit of each probe changes, thus causing a change in contact pressure. Based on the equivalent mechanical model of the probe-substrate, the pressure deviation component caused by mechanical displacement is... ,in The equivalent stiffness used in the temperature-pressure coupling relationship is consistent, ensuring the consistency of physical parameters between the two types of coupled models. When the mechanical displacement distribution exhibits typical warping modes such as bowl-shaped or saddle-shaped, the mechanical field contribution components show corresponding spatial distribution patterns. The rationality of the decoupling results can be further verified through correlation analysis with the standard warping modes.

[0036] After obtaining the temperature field contribution component vectors respectively (Each component is) ) and mechanical field contribution vector (Each component is) After that, the pressure deviation distribution vector By subtracting the two types of contribution components mentioned above point by point, the remaining pressure deviation, i.e., the pressure field contribution component, is obtained. The pressure field contribution component represents the pressure deviation caused by changes in the state of the probe contact interface itself (such as probe tip wear, pad surface oxidation, contact medium residue, etc.) after excluding the effects of thermal effects and mechanical deformation. This component usually exhibits a locally concentrated distribution characteristic in space, which is significantly different from the large-scale gradual distribution of the temperature field contribution component and the structural modal distribution of the mechanical field contribution component. This difference in spatial characteristics can also serve as a basis for cross-validation of the correctness of the decoupling results.

[0037] To improve the accuracy of multi-source decoupling, offline calibration of the temperature-pressure coupling model and the mechanical-pressure coupling model is required in practical engineering applications. The calibration process involves applying temperature excitation separately under controlled conditions (while keeping the mechanical load constant) and applying mechanical load excitation separately (while keeping the temperature constant), measuring the pressure response at each monitoring point to obtain the measured values ​​of the temperature sensitivity coefficient and mechanical displacement sensitivity coefficient at each monitoring point. After calibration, the sensitivity coefficients are stored in matrix form and directly retrieved during online decoupling, replacing theoretical estimations based on material and geometric parameters, which significantly reduces model errors. When the probe card is replaced or repaired, the calibration process must be repeated to update the sensitivity coefficient matrix, ensuring that the decoupling model remains consistent with the actual physical characteristics of the current probe card.

[0038] The three independent contribution components obtained from decoupling—temperature field contribution component, mechanical field contribution component, and pressure field contribution component—correspond to different physical mechanisms, providing a clear input basis for the subsequent generation of differentiated compensation strategies. The temperature field contribution component guides thermal compensation actions (such as adjusting the local heating or cooling power of the probe card); the mechanical field contribution component guides structural compensation actions (such as adjusting the preload of the mounting fixture or driving the piezoelectric actuator to correct substrate warpage); and the pressure field contribution component guides direct pressure compensation actions at the contact interface (such as adjusting the downward stroke of the testing machine). The independent quantification of the three contribution components allows the compensation actions to accurately correspond to their respective physical sources, avoiding overcompensation or undercompensation problems caused by confusion of the action mechanisms due to the coupling of multiple physical fields in a single compensation strategy.

[0039] In one optional implementation, a temperature-pressure coupling relationship is established based on the temperature data to influence the thermal expansion of the probe card material, and a mechanical-pressure coupling relationship is established based on the mechanical displacement data to influence the deformation of the probe card structure, including: A partial differential equation for heat conduction describing the spatiotemporal evolution of the temperature field inside the probe card and a partial differential equation for elasticity describing the spatiotemporal evolution of the stress field inside the probe card are established, and the temperature data and the mechanical displacement data are used as the boundary conditions of the partial differential equations at the monitoring point. The adjoint method is used to solve the partial differential equation of heat conduction in reverse to obtain the equivalent heat source distribution that leads to the current temperature field distribution. Based on the equivalent heat source distribution, the temperature field evolution trajectory at each spatial position on the probe card is calculated. According to the thermal expansion coefficient of the material, the temperature field evolution trajectory is mapped to the pressure change trajectory to establish a temperature-pressure coupling relationship. The adjoint method is used to solve the partial differential equation of elasticity in reverse to obtain the equivalent body force distribution that leads to the current stress field distribution. Based on the equivalent body force distribution, the stress field evolution trajectory at each spatial position on the probe card is calculated. The stress field evolution trajectory is converted into a pressure change trajectory to establish a mechanical-pressure coupling relationship.

[0040] For example, before establishing the physical field coupling relationship, a unified spatial discrete model needs to be constructed based on the probe card's geometry and material properties. The probe card's physical domain is divided into a finite number of spatial units, each of which can carry multiple physical quantities such as temperature, displacement, and pressure. Temperature and mechanical displacement data are collected from each monitoring point and injected into the solution framework of partial differential equations in the form of discrete boundary conditions, thereby driving the reconstruction process of the entire physical field. This partial differential equation inversion strategy constrained by measured data can achieve high-precision estimation of the global physical field distribution under limited monitoring point coverage.

[0041] The partial differential equations describing the spatiotemporal evolution of the temperature field within the probe card, specifically the heat conduction equation, are in general form unsteady-state heat conduction equations, including parameters such as material thermal diffusivity, internal heat source terms, and boundary heat flux density. The partial differential equations describing the spatiotemporal evolution of the stress field within the probe card, based on the Navier displacement equation, include elements such as elastic modulus, Poisson's ratio, body force terms, and displacement boundary conditions. Both sets of equations are established independently within their respective physical domains, but are coupled unidirectionally or bidirectionally through thermal expansion strain terms: changes in the temperature field induce thermal strain, which in turn affects the distribution of the stress field, and the evolution of the stress field is directly related to changes in the probe contact pressure. Temperature measurements and mechanical displacement measurements collected at monitoring points are used as boundary conditions or constraints for the heat conduction equation and the elasticity equation at their corresponding spatial locations, ensuring that the solution results of both sets of equations are consistent with the actual physical state.

[0042] For the inverse solution of the partial differential equation of heat conduction, an adjoint method is employed to construct an adjoint equation that accompanies the original forward problem. The source terms of the adjoint equation are given by the partial derivatives of the objective function with respect to the state variables. The objective function is defined as the weighted sum of squares of the differences between the calculated and measured temperature values ​​at each monitoring point. By solving the adjoint equation, the gradient information of the objective function with respect to the equivalent heat source distribution can be obtained efficiently without needing to perform perturbation calculations for each heat source parameter separately, thus significantly reducing computational cost. Let the equivalent heat source distribution in the probe card spatial domain be... ,in For spatial coordinates, For time, the accompanying method optimizes through iteration to make the time... The driving forward heat conduction equation minimizes the residual between the predicted and measured temperatures at the monitoring point. The converged result is... This refers to the equivalent heat source distribution that causes the current temperature field distribution.

[0043] Based on the reconstructed equivalent heat source distribution Substituting this into the forward heat conduction equation and solving it across the entire field, the temperature field evolution trajectory at any spatial location on the probe card is obtained. Based on this, the coefficient of thermal expansion of the material is utilized. (This parameter, consistent with the aforementioned definition, is the linear expansion coefficient of the probe card material), mapping the spatiotemporal distribution of the temperature field to the local strain distribution caused by the thermal expansion of the material. For the probe body, the thermal expansion strain along the probe body axis directly causes a change in the effective contact length of the probe, thereby causing a shift in the contact pressure. Let the probe body be located in space... Due to temperature changes The resulting axial thermal strain is Combined with the equivalent elastic modulus of the probe Equivalent stiffness of probe (The definitions of both remain consistent with those described above), converting thermal strain into pressure change, thereby establishing a complete temperature-pressure coupling relationship across the entire spatial domain. This coupling relationship is stored in the form of a lookup table or function mapping, which can be directly invoked when generating subsequent differentiated compensation strategies.

[0044] For the inverse solution of the partial differential equations of elasticity, the adjoint method is also used to construct the corresponding adjoint elasticity equations. The objective function is defined as the weighted sum of squares of the differences between the calculated displacement values ​​and the measured mechanical displacement values ​​at each monitoring point. The source terms of the adjoint equations are determined by the partial derivatives of this objective function with respect to the displacement state variables. Let the equivalent body force distribution within the physical domain of the probe card be... The accompanying method, through iterative optimization, makes the result... The residual between the predicted displacement and the measured mechanical displacement at the monitoring point is minimized by the driving positive elasticity equation. During the iteration process, the gradient information is directly given by the solution of the adjoint equation, avoiding the high computational cost of parameter-by-parameter finite difference. The converged result is... This is the equivalent body force distribution that leads to the current stress field distribution.

[0045] Based on the reconstructed equivalent body force distribution Substituting this into the positive elasticity equation and solving the whole field, the stress field evolution trajectory at any spatial location on the probe card is obtained. and strain field evolution trajectory By using a probe contact mechanics model, the normal stress component at the probe's location is converted into the trajectory of contact pressure variation between the probe and the pads of the chip under test. Specifically, for each probe's spatial location, the normal component of the stress tensor at that location is extracted. Combined with the probe contact area parameter, the contribution of pressure deviation caused by mechanical field deformation is calculated, thus establishing a complete mechanical-pressure coupling relationship across the entire spatial domain. This coupling relationship is also stored in the form of a function mapping, and together with the temperature-pressure coupling relationship, it constitutes the physical basis for multi-source decoupling.

[0046] In practical engineering applications, batch variations in the geometry and material properties of probe cards lead to uncertainties in some parameters of the heat conduction and elasticity equations. To improve the robustness of the coupling relationship, material parameters can be simultaneously corrected online during the accompanying inversion process: key material parameters such as thermal diffusivity and elastic modulus are incorporated into the set of optimization variables, along with the equivalent heat source distribution. and equivalent physical strength distribution Furthermore, iterative updates are performed using the accompanying gradient. This joint inversion strategy can continuously correct the physical field model during the probe card's service life, ensuring that the temperature-pressure coupling relationship and the mechanical-pressure coupling relationship always maintain a high degree of consistency with the actual physical state of the probe card. This provides accurate and reliable physical field contribution components for the generation of subsequent multi-field collaborative compensation schemes.

[0047] In one optional implementation, a differentiated compensation strategy is generated based on the contact state type and the physical field contribution components, and the compensation actions for different physical fields in the differentiated compensation strategy are time-series coordinated to obtain a multi-field coordinated compensation scheme, including: Mechanical compensation strategies to increase probe pressure for poor contact caused by probe wear in contact state types; mechanical compensation strategies to adjust probe position for poor contact caused by probe position offset in contact state types; and mechanical compensation strategies to reduce probe load for poor contact caused by probe deformation in contact state types. The adjustment ranges of the temperature field compensation action, mechanical field compensation action, and pressure field compensation action are determined based on the amplitudes of the temperature field contribution component, mechanical field contribution component, and pressure field contribution component, respectively. The mechanical compensation strategy is combined with the adjustment range of each physical field compensation action to form a differentiated compensation strategy; The influence time of temperature field compensation, mechanical field compensation, and pressure field compensation on the probe card is analyzed. Based on the influence time, the execution order and execution interval of each compensation action are determined. The compensation actions are then arranged in sequence according to the execution order and execution interval to obtain a multi-field collaborative compensation scheme.

[0048] For example, based on the comprehensive analysis results of contact state type and physical field contribution components, the generation of differentiated compensation strategies requires first distinguishing the mechanical compensation direction under different fault modes. When the pattern recognition result indicates that the current poor contact is due to probe wear, the effective contact length of the probe tip is shortened due to material loss, and the contact force is reduced accordingly. In this case, a mechanical compensation strategy to increase probe pressure needs to be generated, which restores the target contact force by increasing the overall or local depressurization of the probe card. When the recognition result indicates that the poor contact is due to probe position misalignment, there is a lateral misalignment between the probe tip and the center of the pad, resulting in a reduced contact area or contact point offset. In this case, a mechanical compensation strategy to adjust the probe position needs to be generated, which restores the alignment between the probe and the pad by applying a displacement correction to the lateral fine-tuning mechanism of the probe card. When the recognition result indicates that the poor contact is due to probe deformation, the probe body bends or buckles due to overload or fatigue. Continuing to apply the original load will aggravate the deformation and cause structural failure. In this case, a mechanical compensation strategy to reduce the probe load needs to be generated, which reduces the depressurization to keep the probe operating within the elastic recovery range and avoid permanent damage. The generation logic of the above three types of mechanical compensation strategies is independent of each other. At the same time, the same monitoring point corresponds to only one dominant fault mode. Therefore, the strategy selection is exclusive to ensure the unique determination of the compensation direction.

[0049] After determining the direction of the mechanical compensation strategy, it is necessary to quantify the adjustment amplitude of each compensation action based on the amplitude of each physical field contribution component. Temperature field contribution component. This reflects the effect of thermal expansion on the first The greater the amplitude of the influence of the pressure at each monitoring point, the more significant the impact of the temperature field deviation on the current pressure distribution. Correspondingly, a larger thermal control measure is needed for temperature field compensation, such as adjusting the temperature distribution of the probe card substrate through local heating or cooling units to suppress or offset thermally induced deformation. Mechanical field contribution component. This reflects the contribution of mechanical displacement deviation to pressure. Its amplitude determines the amount of corrective displacement that the displacement actuator needs to apply during mechanical field compensation. Smaller amplitudes require only sub-micrometer level fine adjustments, while larger amplitudes require a combination of coarse and fine adjustments at the millimeter level. Pressure field contribution component. This characterizes the remaining pressure deviation after excluding the influence of the temperature and mechanical fields, and its amplitude determines the output adjustment of the direct pressure control unit (such as a pneumatic actuator or an elastic preload mechanism). The control amplitude in each of the three directions is input with the amplitude of the corresponding contribution component, and is converted into the actual control quantity of the actuator through the transfer function of its respective physical field, thereby achieving precise matching at the amplitude level.

[0050] The mechanical compensation strategy is combined with the adjustment amplitudes of the compensation actions of the three physical fields to form a complete differentiated compensation strategy. The differentiation is reflected in two dimensions: first, different fault modes correspond to different mechanical compensation directions, meaning that the three directions of pressurization, displacement correction, or load reduction cannot be mixed; second, different monitoring points correspond to different adjustment amounts due to the different amplitudes of their physical field contribution components, avoiding overcompensation and undercompensation in some areas by applying a uniform compensation amplitude to all monitoring points. The differentiated compensation strategy records data at the monitoring point level. Each monitoring point corresponds to a quadruple containing the mechanical compensation direction, temperature adjustment amount, mechanical displacement correction amount, and direct pressure adjustment amount, forming a structured compensation parameter table that provides input for subsequent timing sequence arrangement.

[0051] The core of the multi-field collaborative compensation scheme lies in the timing-based collaborative orchestration, which is based on the temporal characteristics of the impact of each compensation action on the probe card. Temperature field compensation actions change the temperature distribution of the probe card substrate through heat conduction. The heat conduction process is constrained by the material's thermal diffusivity and structural dimensions. From the execution of a heating or cooling command to the probe body reaching a new thermal equilibrium state, a response time of several seconds to tens of seconds is typically required, denoted as . Mechanical field compensation is achieved by applying displacement correction through a displacement actuator (such as a piezoelectric ceramic driver or a stepper motor). The mechanical response time depends on the dynamic characteristics of the actuator and is typically on the order of milliseconds to hundreds of milliseconds, denoted as . The pressure field compensation action is achieved by directly adjusting the probe contact force, and its response time lies between that of the mechanical field and the temperature field, denoted as . The response times of the three components meet the requirements. The typical relationship between these time levels means that the various compensation actions cannot be triggered simultaneously. Otherwise, the rapid response mechanical field compensation will be completed before the temperature field has stabilized, leading to new pressure deviations introduced by subsequent temperature field compensation and disrupting the established compensation balance.

[0052] Based on the above analysis of the impact time, the execution order of each compensation action is arranged from longest to shortest response time, that is, the temperature field compensation action is triggered first, followed by the waiting time. (not less than) After that, the pressure field compensation action is triggered, followed by a waiting period. (not less than) This triggers the mechanical field compensation action. This "slow field first, fast field last" programming logic ensures that each compensation action only begins execution after the previous compensation action has fully stabilized, avoiding mutual interference when multiple fields operate simultaneously. Execution interval and The specific value is dynamically determined based on the actual response time of each field measured under the current working conditions, rather than a fixed constant, in order to adapt to the changes in response characteristics under different test environments (such as temperature, humidity, probe card model).

[0053] The compensation actions are performed in the order of temperature field compensation, pressure field compensation, and mechanical field compensation, with corresponding execution intervals. and The data, along with their respective control amplitudes, are organized into an ordered sequence of actions on a timeline, resulting in a multi-field collaborative compensation scheme. This scheme uses timestamps as indexes; each time node records the type of compensation action to be triggered, the set of monitoring points affected, and the corresponding control quantity. The execution layer can sequentially call each execution mechanism according to the timestamp order to accurately deploy the compensation actions. The multi-field collaborative compensation scheme also retains the expected pressure response after each compensation action is completed, for use in subsequent performance verification. If the actual pressure response deviates from the expected value by more than a threshold, a local correction of the compensation scheme can be triggered to ensure that the final pressure distribution converges to the target state.

[0054] In one optional implementation, the probe card is divided into multiple control sub-regions and each control unit is configured to independently perform local compensation. When local compensation fails to meet the target, a cross-regional collaborative compensation scheme is generated through negotiation with adjacent control units, including: Based on the spatial distribution of probes on the probe card, the probe card is divided into multiple control sub-regions, and a control unit is configured for each control sub-region and a communication connection is established between adjacent control units. The control unit performs local compensation in the control sub-region according to the multi-field collaborative compensation scheme, collects pressure distribution data in the control sub-region and calculates the residual between the pressure distribution and the target pressure distribution. When the residual exceeds the allowable range, it is determined that the local compensation cannot meet the target. The control unit constructs a compensation demand vector that includes the residual spatial distribution characteristics and the boundary pressure gradient of the control sub-region, and sends it to the adjacent control unit. The pressure transmission direction is identified based on the pressure gradient difference between the control sub-region and the adjacent control sub-region. The pressure coupling coefficient when the control sub-region applies a compensation effect to the adjacent control sub-region and the pressure coupling coefficient when the adjacent control sub-region applies a compensation effect to the control sub-region are calculated. Based on the compensation demand vector and pressure coupling coefficient, a cross-regional compensation optimization objective is established to minimize the sum of residuals in each control sub-region. The compensation adjustment amount of each control sub-region is obtained by solving the cross-regional compensation optimization objective, which serves as a cross-regional collaborative compensation scheme.

[0055] For example, based on the spatial distribution characteristics of the probes on the probe card, the entire probe card is divided into several control sub-regions. The division is based on the physical arrangement density of the probes, the clustering results of the probe tip coordinates, and the geometric boundaries of the probe card substrate. Specifically, based on the row and column distribution of the probe array, a uniform grid or adaptive grid method is used to divide the probe card plane into... The system comprises several control sub-regions, each containing a number of probes and their corresponding pressure sensor nodes. Each control sub-region is configured with a control unit responsible for reading sensor data within that region, executing compensation commands, and exchanging information with adjacent control units. Point-to-point communication connections are established between adjacent control units, with the communication interface supporting bidirectional transmission of real-time data frames to ensure that the compensation demand vector and coupling parameters can be exchanged within a specified time delay during cross-region negotiation.

[0056] After receiving multiple coordinated compensation schemes, each control unit independently executes local compensation actions within its designated control sub-region. After local compensation is completed, the actual pressure distribution data of each monitoring point within that region is collected, compared point-by-point with the target pressure distribution, and the residuals are calculated. Let the... A total of [number] sub-regions under regulation The monitoring point, the first The local residuals at each monitoring point are , defined as the difference between the actual pressure and the target pressure at the monitoring point. The root mean square value of the residuals at all monitoring points within the area is denoted as... ,Right now .when Exceeding the preset allowable threshold If the local compensation in the controlled sub-region is deemed insufficient to meet the objective, a cross-regional negotiation process is triggered. Allowable threshold. Pre-calibrated according to process requirements, it usually corresponds directly to the overall pressure uniformity index of the probe card.

[0057] After determining that local compensation is insufficient, the control unit constructs a compensation demand vector. This vector is composed of two parts: one part is the spatial distribution characteristics of the residuals, formed by arranging the residuals of each monitoring point within the region according to spatial coordinates to create a residual distribution sub-vector. The other part is the pressure gradient at the boundary of the control sub-region, which is discretized and sampled along the boundary to form a boundary gradient sub-vector. The boundary pressure gradient reflects the changing trend of the pressure field at the boundary with adjacent regions and is a key basis for determining the direction of pressure transmission. and The compensation demand vector is obtained by splicing. and send it to all those related to the first Each regulatory subregion is an adjacent regulatory unit.

[0058] After receiving the compensation demand vector, the pressure gradient difference at the boundary between the controlled sub-region and adjacent controlled sub-regions is analyzed to identify the pressure transmission direction. Let the first... The first regulatory subregion and the first On the shared boundary between adjacent control sub-regions, the first The boundary gradient components on the side are , No. The corresponding component on the side is .when At that time, the pressure was determined by the first The region to the first Pressure is transmitted in one region; otherwise, the transmission direction is determined to be opposite. Based on the pressure transmission direction, the pressure coupling coefficient in both directions is calculated. The first regulatory subregion to the first The pressure coupling coefficient when compensation is applied to each control sub-region is denoted as . Its physical meaning is: in the first When applying a unit compensation adjustment to a region, in the first... The ratio of the pressure response amplitude at the boundary of each region to the applied amount. Similarly, the first... The first regulatory subregion to the first The pressure coupling coefficient when compensation is applied to each control sub-region is denoted as . The coupling coefficient values ​​were obtained through offline calibration experiments. During calibration, a pressure disturbance of known amplitude was applied to a single region, and the pressure response of adjacent regions was recorded simultaneously. The coupling coefficient matrix was obtained by linear regression fitting and stored in the parameter library of the control unit.

[0059] Based on the compensation demand vector and pressure coupling coefficient of each control sub-region, a cross-regional compensation optimization objective is established. The optimization objective is to minimize the sum of residuals of all participating control sub-regions. Let the set of participating regions be denoted as . The compensation adjustment amounts for each region constitute the vector to be determined. ,in For the first The additional compensation adjustments required by each regulatory sub-region during the cross-regional coordination phase. Considering the pressure coupling effect between regions, the first... The equivalent residual for each region after collaborative compensation is: ,in For the first The set of neighboring regions of each control sub-region. The cross-region compensation optimization objective function is: At the same time, physical constraints are imposed on the compensation adjustment amount, requiring that the compensation adjustment amount in each region does not exceed the upper limit of the action range of its actuator. ,Right now This constrained optimization problem can be transformed into a constrained quadratic programming problem, which can be solved in real time on the embedded processor of the control unit using the interior point method or the effective set method to obtain the optimal compensation adjustment vector for each control sub-region. .

[0060] Solving for the results Subsequently, this was distributed as a cross-regional collaborative compensation scheme to all participating regulatory units. Each regulatory unit then added a corresponding compensation adjustment to its existing local compensation. The system executes synchronously according to the time sequence specified in the multi-field collaborative compensation scheme, ensuring that the cross-regional collaborative compensation actions match the response characteristics of each physical field in the time dimension. After execution, pressure distribution data for each region is collected again to verify whether the residuals after collaborative compensation have converged to within the allowable range. If there are still regions exceeding the threshold, the compensation demand vector is reconstructed using the results of this round of collaborative compensation as the initial state, and the next round of iterative negotiation is initiated until the global residuals meet the convergence condition or the maximum number of iterations is reached. This mechanism, which combines distributed negotiation with centralized optimization, enables the pressure regulation of each region of the probe card to maintain the speed of local response while eliminating the systemic pressure unevenness caused by insufficient regulation capacity in a single region through cross-regional coupled compensation.

[0061] In one optional implementation, during the execution of the multi-field collaborative compensation scheme and the cross-regional collaborative compensation scheme, the pressure response characteristics of the probes are monitored. Based on the pressure response characteristics, faulty probes are identified, and redundant probes are selected in their adjacent control sub-regions to perform load redistribution, including: During the execution of the multi-field collaborative compensation scheme and the cross-regional collaborative compensation scheme, the pressure change curves of each probe are recorded, and the pressure response speed, pressure overshoot amplitude and pressure stabilization time are extracted from the pressure change curves as pressure response characteristics. The pressure response characteristics of each probe are compared with the expected pressure response characteristics. When the pressure response speed of a probe is lower than the expected response threshold, the pressure overshoot exceeds the expected range, or the pressure stabilization time exceeds the expected stabilization threshold, the probe is identified as a faulty probe. Obtain candidate redundant probes in the adjacent control sub-region of the fault probe, and calculate the remaining load capacity of each candidate redundant probe and the amount of pressure field disturbance generated after bearing the load of the fault probe. The redundant probe selection optimization objective is established to minimize the total amount of pressure field disturbance under the constraint that the remaining load capacity meets the load transfer requirements. The target redundant probes and the load allocation of each target redundant probe are obtained by solving the redundant probe selection optimization objective. Load adjustment instructions are sent to each target redundant probe according to the load allocation to complete the load redistribution.

[0062] For example, during the execution of multi-field and cross-regional collaborative compensation schemes, the pressure change curve of each probe is continuously recorded to collect complete dynamic process data from the issuance of the compensation command to pressure stabilization at a high sampling rate. Three types of pressure response characteristic parameters are extracted from the collected pressure change curves: pressure response speed... The pressure overshoot is defined as the average rate of change corresponding to the pressure rising from the initial value to 90% of the target value after the compensation action is triggered; The pressure settling time is defined as the ratio of the maximum deviation from the target pressure value during the pressure response process to the target pressure value. , defined as the time required for the pressure response curve to enter and remain within the ±2% error band of the target pressure value. These three types of characteristic parameters together constitute a quantitative description of the dynamic response behavior of each probe, comprehensively characterizing the actual working state of the probe from three dimensions: response speed, oscillation degree, and convergence speed.

[0063] The pressure response characteristics of each probe were compared with the expected pressure response characteristics item by item. The expected pressure response characteristics were obtained by statistical analysis of the probe card's factory calibration data and historical normal operation data, forming three thresholds: expected response threshold. Expected overshoot range and expected stability threshold For a given probe, if its pressure response speed... Below the expected response threshold This indicates that the probe's slow response after the compensation action is applied may indicate contact degradation, elastic fatigue, or a malfunction in the drive mechanism; if its pressure overshoot is significant... Exceeding expectations The upper bound indicates that the probe exhibits overshoot behavior due to abnormal stiffness or insufficient damping; if its pressure settling time... Exceeding the expected stability threshold This indicates that the pressure regulation process of the probe converges slowly, exhibiting continuous oscillations or systematic drift. Probes that meet any of the above criteria are identified as faulty probes, and their locations and corresponding fault characteristic types are marked in the control system so that the fault mode can be referenced for targeted load allocation during subsequent redundancy scheduling.

[0064] Obtain all probes within the control sub-region where the faulty probe is located and its adjacent control sub-regions as a candidate redundant probe set. For each candidate redundant probe, calculate its remaining load capacity. Defined as the upper limit of the current rated load-bearing pressure of the probe. Compared with its current actual pressure The difference, that is subscript This serves as the index for the candidate redundant probes. The remaining load capacity reflects the maximum additional pressure increment the probe can withstand without exceeding its own load-bearing limit. Simultaneously, the disturbance to the surrounding pressure field caused by each candidate redundant probe after absorbing the load of the failed probe is calculated. The calculation of the disturbance is based on the pressure field spatial transmission model: when the pressure increase of a candidate redundant probe is... At that time, it was on the surrounding area. Pressure impact at each monitoring point The spatial coupling coefficient between the probe and the monitoring point and The product is given, that is ,in The total pressure field disturbance of the candidate redundant probe is obtained by calibrating the probe card structure using a finite element model. The sum of the squares of the pressure disturbances at all affected monitoring points is then taken as the square root. ,Right now .

[0065] The optimization objective for selecting redundant probes is established as follows: Under the constraint that the remaining load capacity of the selected redundant probes meets the load transfer requirements, minimize the total pressure field disturbance caused by all redundant probes. Let the total load to be transferred from the faulty probe be... The target redundant probe set is The load distribution of each target redundant probe is as follows: The optimization objective is then expressed as minimizing The constraints include: the total load distribution is equal to the fault probe load, i.e. The number of redundant probes allocated to each target shall not exceed its remaining load capacity, i.e. ; and the allocation of redundant probes for each target is non-negative, i.e. This optimization problem is a quadratic programming problem with linear constraints. It can be solved using the interior-point method or the effective set method to obtain the optimal set of redundant target probes. and the optimal load allocation for each target redundant probe. .

[0066] After the solution is completed, the optimal load allocation for each target redundant probe is determined. A load adjustment command is generated, and a pressure increment control command is sent to the corresponding control unit to drive the target redundant probes to gradually take over the load released by the faulty probe. The execution of the load adjustment command adopts a gradual incremental approach to avoid secondary impacts on the overall pressure distribution of the probe card caused by sudden pressure changes. During the load redistribution process, the pressure response curves of each target redundant probe are continuously monitored to verify whether their actual bearing pressure converges to the target value as expected. If a target redundant probe exhibits an abnormal response during the execution process, it is removed from the target probe. The optimization solution is retried and replaced by eliminating redundant probes, and alternative solutions are added from the remaining candidate redundant probes to ensure the robustness and integrity of load redistribution. The entire fault probe identification and redundant load redistribution process runs in a closed loop within the compensation execution cycle, ensuring that the uniformity and stability of the global pressure distribution can still be maintained even when probes are stuck at single or multiple points of failure.

[0067] A second aspect of the present invention provides a probe card pressure distribution intelligent control system, comprising: The state recognition unit is used to collect the pressure time-domain waveform of the probe contact process at the pressure monitoring point of the probe card and perform segmented feature extraction to obtain a multi-dimensional feature vector. The multi-dimensional feature vector is then used for pattern recognition to obtain the contact state type. The multi-source decoupling unit is used to collect multi-modal physical quantity data of pressure field, temperature field and mechanical field at the monitoring point of the probe card, and to perform multi-source decoupling of pressure deviation based on the coupling relationship between physical fields to obtain independent physical field contribution components. The compensation orchestration unit is used to generate differentiated compensation strategies based on the contact state type and the physical field contribution components, and to perform time-series collaborative orchestration of the compensation actions for different physical fields in the differentiated compensation strategies to obtain a multi-field collaborative compensation scheme. The regional coordination unit is used to divide the probe card into multiple control sub-regions and configure the control units to independently perform local compensation. When the local compensation cannot meet the target, it negotiates with adjacent control units to generate a cross-regional collaborative compensation scheme. The fault reconfiguration unit is used to monitor the pressure response characteristics of the probes during the execution of the multi-field collaborative compensation scheme and the cross-regional collaborative compensation scheme, identify faulty probes based on the pressure response characteristics, and select redundant probes in their adjacent control sub-regions to perform load redistribution.

[0068] 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.

[0069] 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.

[0070] 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.

[0071] 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 method for intelligent control of probe card pressure distribution, characterized in that, include: The pressure time-domain waveform of the probe contact process is collected at the pressure monitoring point of the probe card and segmented feature extraction is performed to obtain a multi-dimensional feature vector. The contact state type is obtained by pattern recognition of the multi-dimensional feature vector. Multimodal physical quantity data of pressure field, temperature field and mechanical field are collected at the monitoring points of the probe card. Based on the coupling relationship between the physical fields, the pressure deviation is decoupled from multiple sources to obtain independent physical field contribution components. Differentiated compensation strategies are generated based on the contact state type and the physical field contribution components. The compensation actions for different physical fields in the differentiated compensation strategies are then time-coordinated to obtain a multi-field coordinated compensation scheme. The probe card is divided into multiple control sub-regions and each control unit is configured to independently perform local compensation. When local compensation cannot meet the target, a cross-regional collaborative compensation scheme is generated by negotiating with adjacent control units. During the execution of the multi-field collaborative compensation scheme and the cross-regional collaborative compensation scheme, the pressure response characteristics of the probe are monitored. Based on the pressure response characteristics, faulty probes are identified and redundant probes are selected in their adjacent control sub-regions to perform load redistribution.

2. The method of claim 1, wherein, The pressure time-domain waveform of the probe contact process is acquired at the pressure monitoring point of the probe card, and segmented feature extraction is performed to obtain a multi-dimensional feature vector. Pattern recognition is then performed on the multi-dimensional feature vector to obtain the contact state type, including: During the probe contact with the wafer, the pressure time-domain waveform is acquired in real time. The pressure time-domain waveform is divided into the initial contact stage, the pressure stabilization stage and the overload stage according to the time process of the probe contact process. The pressure rise rate, pressure peak value and pressure fluctuation amplitude are extracted as stage features for each stage. The features of each stage are combined to form a multi-dimensional feature vector. The contact stiffness index is obtained by calculating the ratio of the pressure rise rate to the pressure peak value in the multidimensional feature vector, and the contact stability index is obtained by calculating the rate of change of the pressure fluctuation amplitude between each stage. The contact state type is determined based on the combination of the numerical ranges of the contact stiffness index and the contact stability index. When the contact stiffness index decreases while the contact stability index is normal, it is determined to be poor contact caused by probe wear. When the contact stiffness index is normal while the contact stability index is abnormal, it is determined to be poor contact caused by probe position deviation. When both the contact stiffness index and the contact stability index are normal but the pressure peak value is lower than the target value, it is determined to be poor contact caused by probe deformation.

3. The method of claim 1, wherein, Multimodal physical quantity data of pressure field, temperature field and mechanical field are collected at the monitoring points of the probe card. Based on the coupling relationship between the physical fields, the pressure deviation is decoupled from multiple sources to obtain independent physical field contribution components, including: Pressure data, temperature data, and mechanical displacement data are simultaneously collected at the monitoring points of the probe card as multimodal physical quantity data; Based on the pressure data, the deviation between the actual pressure distribution and the target pressure distribution of the probe card is calculated to obtain the pressure deviation distribution; Based on the temperature data, a temperature-pressure coupling relationship is established for the influence of the temperature field on the thermal expansion of the probe card material, and a mechanical-pressure coupling relationship is established for the influence of the mechanical field on the deformation of the probe card structure based on the mechanical displacement data. Based on the temperature-pressure coupling relationship, the pressure deviation component caused by the temperature change represented by the temperature data is separated from the pressure deviation distribution to obtain the temperature field contribution component. Based on the mechanical-pressure coupling relationship, the pressure deviation component caused by the mechanical displacement represented by the mechanical displacement data is separated from the pressure deviation distribution to obtain the mechanical field contribution component. The remaining pressure deviation after deducting the temperature field contribution component and the mechanical field contribution component from the pressure deviation distribution is taken as the pressure field contribution component.

4. The method of claim 3, wherein, Based on the temperature data, a temperature-pressure coupling relationship is established for the influence of the temperature field on the thermal expansion of the probe card material; based on the mechanical displacement data, a mechanical-pressure coupling relationship is established for the influence of the mechanical field on the deformation of the probe card structure, including: A partial differential equation for heat conduction describing the spatiotemporal evolution of the temperature field inside the probe card and a partial differential equation for elasticity describing the spatiotemporal evolution of the stress field inside the probe card are established, and the temperature data and the mechanical displacement data are used as the boundary conditions of the partial differential equations at the monitoring point. The adjoint method is used to solve the partial differential equation of heat conduction in reverse to obtain the equivalent heat source distribution that leads to the current temperature field distribution. Based on the equivalent heat source distribution, the temperature field evolution trajectory at each spatial position on the probe card is calculated. According to the thermal expansion coefficient of the material, the temperature field evolution trajectory is mapped to the pressure change trajectory to establish a temperature-pressure coupling relationship. The adjoint method is used to solve the partial differential equation of elasticity in reverse to obtain the equivalent body force distribution that leads to the current stress field distribution. Based on the equivalent body force distribution, the stress field evolution trajectory at each spatial position on the probe card is calculated. The stress field evolution trajectory is converted into a pressure change trajectory to establish a mechanical-pressure coupling relationship.

5. The method of claim 3, wherein, A differentiated compensation strategy is generated based on the contact state type and the physical field contribution components. The compensation actions for different physical fields within the differentiated compensation strategy are then time-coordinated to obtain a multi-field coordinated compensation scheme, including: Mechanical compensation strategies to increase probe pressure for poor contact caused by probe wear in contact state types; mechanical compensation strategies to adjust probe position for poor contact caused by probe position offset in contact state types; and mechanical compensation strategies to reduce probe load for poor contact caused by probe deformation in contact state types. The adjustment ranges of the temperature field compensation action, mechanical field compensation action, and pressure field compensation action are determined based on the amplitudes of the temperature field contribution component, mechanical field contribution component, and pressure field contribution component, respectively. The mechanical compensation strategy is combined with the adjustment range of each physical field compensation action to form a differentiated compensation strategy; The influence time of temperature field compensation, mechanical field compensation, and pressure field compensation on the probe card is analyzed. Based on the influence time, the execution order and execution interval of each compensation action are determined. The compensation actions are then arranged in sequence according to the execution order and execution interval to obtain a multi-field collaborative compensation scheme.

6. The method of claim 1, wherein, The probe card is divided into multiple control sub-regions, and control units are configured to independently perform local compensation. When local compensation fails to meet the target, a cross-regional collaborative compensation scheme is generated through negotiation with adjacent control units, including: Based on the spatial distribution of probes on the probe card, the probe card is divided into multiple control sub-regions, and a control unit is configured for each control sub-region and a communication connection is established between adjacent control units. The control unit performs local compensation in the control sub-region according to the multi-field collaborative compensation scheme, collects pressure distribution data in the control sub-region and calculates the residual between the pressure distribution and the target pressure distribution. When the residual exceeds the allowable range, it is determined that the local compensation cannot meet the target. The control unit constructs a compensation demand vector containing residual spatial distribution characteristics and the pressure gradient at the boundary of the control sub-region, and sends it to the adjacent control unit; it identifies the pressure transmission direction based on the pressure gradient difference at the boundary between the control sub-region and the adjacent control sub-region, and calculates the pressure coupling coefficient when the control sub-region applies compensation to the adjacent control sub-region and the pressure coupling coefficient when the adjacent control sub-region applies compensation to the control sub-region. Based on the compensation demand vector and pressure coupling coefficient, a cross-regional compensation optimization objective is established to minimize the sum of residuals in each control sub-region. The compensation adjustment amount of each control sub-region is obtained by solving the cross-regional compensation optimization objective, which serves as a cross-regional collaborative compensation scheme.

7. The method of claim 1, wherein, During the execution of the multi-field collaborative compensation scheme and the cross-regional collaborative compensation scheme, the pressure response characteristics of the probes are monitored. Based on the pressure response characteristics, faulty probes are identified, and redundant probes are selected in their adjacent control sub-regions to perform load redistribution, including: During the execution of the multi-field collaborative compensation scheme and the cross-regional collaborative compensation scheme, the pressure change curves of each probe are recorded, and the pressure response speed, pressure overshoot amplitude and pressure stabilization time are extracted from the pressure change curves as pressure response characteristics. The pressure response characteristics of each probe are compared with the expected pressure response characteristics. When the pressure response speed of a probe is lower than the expected response threshold, the pressure overshoot exceeds the expected range, or the pressure stabilization time exceeds the expected stabilization threshold, the probe is identified as a faulty probe. Obtain candidate redundant probes in the adjacent control sub-region of the fault probe, and calculate the remaining load capacity of each candidate redundant probe and the amount of pressure field disturbance generated after bearing the load of the fault probe. The redundant probe selection optimization objective is established to minimize the total amount of pressure field disturbance under the constraint that the remaining load capacity meets the load transfer requirements. The target redundant probes and the load allocation of each target redundant probe are obtained by solving the redundant probe selection optimization objective. Load adjustment instructions are sent to each target redundant probe according to the load allocation to complete the load redistribution.

8. Probe card pressure distribution intelligent regulation system for implementing the method according to any one of claims 1 to 7, characterized in that, include: The state recognition unit is used to collect the pressure time-domain waveform of the probe contact process at the pressure monitoring point of the probe card and perform segmented feature extraction to obtain a multi-dimensional feature vector. The multi-dimensional feature vector is then used for pattern recognition to obtain the contact state type. The multi-source decoupling unit is used to collect multi-modal physical quantity data of pressure field, temperature field and mechanical field at the monitoring point of the probe card, and to perform multi-source decoupling of pressure deviation based on the coupling relationship between physical fields to obtain independent physical field contribution components. The compensation orchestration unit is used to generate differentiated compensation strategies based on the contact state type and the physical field contribution components, and to perform time-series collaborative orchestration of the compensation actions for different physical fields in the differentiated compensation strategies to obtain a multi-field collaborative compensation scheme. The regional coordination unit is used to divide the probe card into multiple control sub-regions and configure the control units to independently perform local compensation. When the local compensation cannot meet the target, it negotiates with adjacent control units to generate a cross-regional collaborative compensation scheme. The fault reconfiguration unit is used to monitor the pressure response characteristics of the probes during the execution of the multi-field collaborative compensation scheme and the cross-regional collaborative compensation scheme, identify faulty probes based on the pressure response characteristics, and select redundant probes in their adjacent control sub-regions to perform load redistribution.

9. An electronic device, comprising: 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 stored thereon computer program instructions, wherein, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images