A real-time compensation method for chip test results based on PSO-KF algorithm

Abstract

A real-time compensation method for chip test results based on a PSO-KF algorithm belongs to the field of integrated circuit testing technology and addresses the problem of how to improve the accuracy and stability of integrated circuit testing. The method inputs collected raw measurement data into a Kalman filter for preliminary filtering processing; based on real-time changes in the measurement data and environmental noise, a particle swarm optimization algorithm is used to optimize the noise parameters in the Kalman filter; the particle swarm optimization algorithm updates the noise parameters after obtaining each measurement data and reapplies the optimal parameters to the Kalman filter, thereby achieving real-time compensation of the measurement data; the yield is evaluated based on the compensated data, and the data is fed back to the particle swarm optimization algorithm for the next iterative optimization, ensuring that the Kalman filter can adapt to noise changes and maintain the best compensation effect. The technical solution of the present invention improves the accuracy, yield and product quality of chip testing.

Description

Technical Field

[0001] The present invention belongs to the technical field of integrated circuit testing, and relates to a real-time compensation method for chip testing results based on a PSO-KF algorithm. Background Art

[0002] Integrated circuit (IC) test equipment plays a crucial role in IC R&D, manufacturing, and quality control. As IC integration and complexity increase, chip performance and quality requirements are also becoming increasingly stringent, making the testing process even more crucial. However, during IC testing, the impact of external noise on test results cannot be ignored. This noise interference comes from a wide range of sources, posing significant challenges to the accuracy and consistency of test data. This problem is particularly pronounced during high-precision, high-frequency IC testing.

[0003] In the IC production process, testing is a critical step in ensuring product quality. Testing equipment must perform extensive electrical performance and functional tests on chips under varying environmental conditions to verify compliance with design specifications and industry standards. However, because IC testing requires processing weak electrical signals, the inevitable noise interference in the test environment can easily affect the accuracy of test data, leading to biased test results and even misjudgment.

[0004] Noise interference can be categorized into two main types: internal noise and external noise. Internal noise primarily originates from the hardware structure of the test equipment itself and the internal structure of the IC under test, including thermal noise, shot noise, and power supply noise. These noises are typically caused by the material properties and operating conditions of the physical device and are difficult to completely eliminate. External noise primarily originates from external interference in the test environment, such as electromagnetic radiation, changes in ambient temperature and humidity, and power supply fluctuations. Other electronic devices surrounding the test equipment, such as computers, power modules, and radio equipment, all generate varying degrees of electromagnetic interference. This interference can enter the test system through coupling, causing deviations in the test signal.

[0005] With advances in semiconductor process technology, IC feature sizes continue to shrink and operating frequencies continue to increase, significantly increasing chips' sensitivity to noise. Especially at the nanoscale, where device operating voltages drop to 1V or even lower, signal amplitudes become extremely weak, significantly reducing noise tolerance. Consequently, when collecting and measuring chip signals, test equipment is highly susceptible to various noise interferences, resulting in reduced test accuracy, increased test failure rates and rework rates, and ultimately, impacting product yield.

[0006] Test data deviation caused by noise interference during IC testing is a major technical challenge in the field. This is particularly true as ICs move toward higher integration densities and higher frequencies, where the impact of noise interference on test accuracy is becoming increasingly prominent. Therefore, compensating for noise-impaired IC test data to ensure accuracy and consistency remains a pressing issue in IC test equipment development. Summary of the Invention

[0007] The technical solution of the present invention is used to solve the problem of how to improve the accuracy and stability of integrated circuit testing.

[0008] The present invention solves the above technical problems through the following technical solutions:

[0009] A real-time compensation method for chip test results based on the PSO-KF algorithm includes the following steps:

[0010] Step 1: Collect the original measurement data during the chip testing process and use the mode to fill in the missing values in the data;

[0011] Step 2: Input the collected raw measurement data into the Kalman filter for preliminary filtering;

[0012] Step 3: Based on the real-time changes of the measurement data and the environmental noise, the noise parameters in the Kalman filter are optimized using the particle swarm optimization algorithm;

[0013] Step 4: The particle swarm optimization algorithm updates the noise parameters after obtaining each measurement data, and reapplies the optimal parameters to the Kalman filter, thereby achieving real-time compensation of the measurement data;

[0014] Step 5: Evaluate the yield based on the compensated data, and the data will be fed back to the particle swarm optimization algorithm for the next iterative optimization to ensure that the Kalman filter can adapt to noise changes and maintain the best compensation effect.

[0015] Furthermore, the method of inputting the collected raw measurement data into the Kalman filter for preliminary filtering processing described in step 2 is specifically as follows:

[0016] 1) State variable and measurement model initialization: Define the state variables of the Kalman filter according to the specific requirements of the chip test; initialize the initial values and covariance matrix of the state variables to represent the uncertainty of the system state at the initial moment; build the measurement model, establish the mathematical relationship between the state variables and the actual measurement data, and define the state transfer matrix A, control matrix B, and observation matrix H;

[0017] 2) Prediction step: predict the state of the system based on the state transition equation, calculate the covariance matrix, and estimate the state value and uncertainty of the current data;

[0018] 3) Update step: Combined with the current measurement value, the state estimate is updated through the Kalman gain to reduce the difference between the predicted value and the measured value;

[0019] 4) Filtered data output: Output the updated state estimate as the estimate of the true state at the current moment and used for subsequent data analysis and compensation; the filtered data will be fed back as input to the particle swarm optimization algorithm to optimize the noise parameters.

[0020] Furthermore, the method of predicting the state of the system based on the state transition equation, calculating the covariance matrix, and estimating the state value and uncertainty of the current data is as follows:

[0021] Use the state transition equation to predict the estimated state value of the system at the current moment:

[0022]

[0023] in, represents the state prediction value at the current moment, A is the state transfer matrix, is the last state estimate, B is the control matrix, μ k-1 is the control input;

[0024] Compute the covariance matrix of the predicted states:

[0025] P k|k-1 =AP k-1 A T +Q

[0026] Among them, P k|k-1 represents the prediction covariance matrix, P k-1 is the covariance matrix of the previous moment, and Q is the process noise covariance matrix.

[0027] Furthermore, the method of combining the current measurement value and updating the state estimation value through the Kalman gain to reduce the difference between the predicted value and the measured value is as follows:

[0028] Calculate the Kalman gain K k as follows:

[0029] K k =P k|k-1 H T (HP k|k-1 H T +R) -1

[0030] Among them, K k represents the Kalman gain, H is the observation matrix, and R is the observation noise covariance matrix;

[0031] The updated state estimate is combined with the measurement data as follows:

[0032]

[0033] in, represents the updated state estimate, z k is the actual measured value at the current moment, is the measurement residual, that is, the difference between the predicted value and the measured value;

[0034] The updated state covariance matrix is as follows:

[0035] P k =(IK k H)P k|k-1

[0036] Among them, P k is the updated state covariance matrix, which represents the uncertainty of the updated state variables.

[0037] Furthermore, the method described in step 3 for optimizing the noise parameters in the Kalman filter using the particle swarm optimization algorithm based on the real-time changes in the measurement data and the environmental noise is as follows:

[0038] 1) Initialize the particle swarm: determine the size of the particle swarm, that is, the number of particles N. Each particle represents a noise parameter combination of the noise covariance matrix Q and the observation noise covariance matrix R. Randomly initialize the position and velocity of each particle and set the initial parameter range.

[0039] 2) Calculate the fitness function: For each particle, apply its current noise parameter combination to the Kalman filter to filter the measurement data; define the fitness function based on the filtering error to evaluate the filtering effect;

[0040] 3) Update individual and global extreme values: For each particle, calculate the current fitness value. If the current fitness value is better than the historical optimal value of the particle, update the current fitness value and corresponding position to the individual optimal value of the particle; find the particle with the best fitness value among the individual optimal values of all particles, and update its position and fitness value to the global optimal value;

[0041] 4) Update particle velocity and position;

[0042] 5) Boundary condition processing: Use the rebound boundary method or position reset method to process particles that exceed the boundary, so that the particle position is always within a reasonable range;

[0043] 6) Repeat steps 2) to 5) until a preset stopping condition is met;

[0044] 7) Application of optimal noise parameters: After particle swarm optimization, the optimal noise covariance matrix Q and observation noise covariance matrix R are found and applied to the Kalman filter for the next round of data filtering and real-time compensation.

[0045] Furthermore, the formula for updating particle velocity is as follows:

[0046] v i t+1 =ωv i t +c1r1(pbest i -x i t )+c2r2(gbest i -x i t )

[0047] Among them, v i t+1 represents the velocity of particle i in the t+1th iteration, ω is the inertia weight, c1 and c2 are the individual and global learning factors respectively, r1 and r2 are randomly generated numbers in the interval [0,1], pbest i is the individual optimal position of the particle, gbest i is the global optimal position, x i t is the position of the particle at iteration t.

[0048] Furthermore, the formula for updating the particle position is as follows:

[0049] x i t+i =x i t +v i t+1

[0050] Among them, x i t+1 represents the position of the particle at the t+1th iteration.

[0051] Furthermore, the stopping condition is that the maximum number of iterations or the global optimal fitness value is less than a threshold.

[0052] The present invention also provides an electronic device, comprising a memory and a processor, characterized in that the memory is used to store a program that supports the processor to execute the above-mentioned real-time compensation method for chip test results based on the PSO-KF algorithm, and the processor is configured to execute the program stored in the memory.

[0053] The present invention also provides a storage medium storing a computer program. When the computer program is run by a processor, the steps of the above-mentioned method for real-time compensation of chip test results based on the PSO-KF algorithm are executed.

[0054] The advantages of the present invention are:

[0055] (1) During the data acquisition process, traditional compensation methods are mostly performed offline, making it difficult to achieve real-time measurement data correction, thereby limiting the timeliness and accuracy of data compensation. The present invention innovatively takes real-time as the core, and adopts a combination of Kalman filtering and PSO algorithm to instantly analyze and compensate each new measurement data. Through the real-time adjustment of the filter parameters by PSO, it is ensured that the measurement data can be immediately and accurately compensated when the noise environment changes. This real-time compensation mechanism ensures that the equipment can continuously adapt to changes in external noise during operation, thereby significantly improving the accuracy and consistency of the measurement data.

[0056] (2) In traditional Kalman filtering methods, noise parameters are often fixed, and the dynamic changes of noise during the test process are not fully considered, resulting in unsatisfactory filtering effects in practical applications. To solve this problem, the method of the present invention introduces a particle swarm optimization (PSO) algorithm, which can dynamically adjust the noise parameters in the Kalman filter according to real-time measurement data to achieve adaptive optimization of the noise level. Through this optimization method, the compensation process can better adapt to changes in the noise environment, improve the accuracy of compensation and the robustness of the system, and ensure that the test process always maintains a high-precision filtering effect.

[0057] (3) During the chip testing process, traditional methods are often not timely and effective in dealing with equipment noise and errors, which affects the accuracy of the test results and leads to a reduction in yield. The present invention significantly improves the accuracy of chip test results and reduces the deviation of measurement data by effectively compensating for equipment noise and errors in real time. In addition, through precise noise compensation and error correction, the proportion of defective products is effectively reduced, and the overall yield of the chip is improved. This improvement not only improves the reliability of test results, but also provides important technical support for quality control in the chip production process, promoting the overall improvement of production quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is a flow chart of a real-time compensation method for chip test results based on the PSO-KF algorithm;

[0059] Figure 2 This is the effect diagram of Kalman filter combined with PSO algorithm to compensate current test data;

[0060] Figure 3This is the effect diagram of Kalman filtering combined with PSO algorithm to compensate voltage test data. DETAILED DESCRIPTION

[0061] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0062] The technical solution of the present invention is further described below with reference to the accompanying drawings and specific embodiments:

[0063] Example 1

[0064] like Figure 1 As shown, a real-time compensation method for chip test results based on the PSO-KF algorithm according to an embodiment of the present invention includes the following steps:

[0065] 1. Data collection and preprocessing

[0066] During the chip testing process, original measurement data is collected. The embodiment of the present invention collects current and voltage test data. To ensure the reliability of the data, these original test data need to be preprocessed, that is, the mode is used to supplement the missing values in the data.

[0067] 2. Kalman filter processing

[0068] The collected raw measurement data is usually affected by different noise factors and needs to be filtered. The collected raw data is filtered using a Kalman filter. The specific steps are as follows:

[0069] 1) Initialization of state variables and measurement model

[0070] a. Define the state variables of the Kalman filter according to the specific requirements of the chip test. Usually, the state variables include the actual measurement quantities of the system (such as voltage, current, etc.);

[0071] b. Initialize the initial values of the state variables (usually zero or estimated values) and the covariance matrix to represent the uncertainty of the system state at the initial moment;

[0072] c. Construct a measurement model, establish the mathematical relationship between state variables and actual measurement data, and define the state transfer matrix A, control matrix B, observation matrix H, etc.

[0073] 2) Prediction step

[0074] The state of the system is predicted based on the state transition equation, and the covariance matrix is calculated to estimate the state value and uncertainty of the current data.

[0075] a. Use the state transition equation to predict the estimated state value of the system at the current moment as follows:

[0076]

[0077] in, represents the state prediction value at the current moment, A is the state transfer matrix, is the last state estimate, B is the control matrix, μ k-1 is the control input.

[0078] b. Calculate the covariance matrix of the predicted state as follows:

[0079] P k|k-1 =AP k-1 A T +Q

[0080] Among them, P k|k-1 represents the prediction covariance matrix, P k-1 is the covariance matrix of the previous moment, and Q is the process noise covariance matrix, which is used to characterize the random disturbance of the system itself.

[0081] 3) Update steps

[0082] Combined with the current measurement value, the state estimate is updated through the Kalman gain to reduce the difference between the predicted value and the measured value.

[0083] a. Calculate the Kalman gain K k as follows:

[0084] K k =P k|k-1 H T (HP k|k-1 H T +R) -1

[0085] Among them, K k represents the Kalman gain, H is the observation matrix, and R is the observation noise covariance matrix. The Kalman gain is used to measure the influence of the measurement data on the state estimation.

[0086] b. Update the state estimate based on the measurement data as follows:

[0087]

[0088] in, represents the updated state estimate, z k is the actual measured value at the current moment, is the measurement residual, that is, the difference between the predicted value and the measured value.

[0089] c. Update the state covariance matrix as follows:

[0090] P k =(IK k H)P k|k-1

[0091] Among them, P k is the updated state covariance matrix, which represents the uncertainty of the updated state variables.

[0092] 4) Filter data output:

[0093] a. Output the updated state estimate It serves as an estimate of the actual state at the current moment and is used for subsequent data analysis and compensation.

[0094] b. The filtered data will be fed back as input to the particle swarm optimization (PSO) algorithm to optimize the noise parameters and ensure a more accurate filtering effect in the next iteration.

[0095] After the above steps, the Kalman filter performs preliminary processing on the original measurement data, minimizing the uncertainty caused by noise in the system, and providing a basic guarantee for the subsequent real-time compensation of the particle swarm optimization algorithm.

[0096] 3. Particle Swarm Optimization (PSO) adaptively adjusts filter parameters

[0097] Based on the real-time changes in measurement data and environmental noise, the particle swarm optimization (PSO) algorithm is used to optimize the noise parameters in the Kalman filter. The PSO algorithm randomly initializes a group of particles, each representing a different combination of noise parameters, and searches for the optimal noise parameters in the solution space through multiple iterations to improve the filtering effect. The specific steps are as follows:

[0098] (1) Initialize the particle swarm

[0099] First, determine the size of the particle swarm, that is, the number of particles N. Each particle represents a possible combination of noise parameters (such as process noise covariance matrix Q and observation noise covariance matrix R).

[0100] The position and velocity of each particle are randomly initialized, where the position represents the initial value of the noise parameter and the velocity represents the step size of the change of these parameters in the solution space.

[0101] The initial parameter range is set so that particles can search for the optimal solution in a larger solution space, ensuring the diversity of parameter settings and avoiding falling into local optimality.

[0102] (2) Calculate the fitness function

[0103] For each particle, its current noise parameter combination is applied to the Kalman filter to filter the measurement data.

[0104] Define a fitness function to evaluate the filtering effect. Typically, the fitness function can be based on the filtering error (such as the mean square error, MSE), which measures the difference between the filtered measurement data and the true value. The smaller the fitness function value, the better the noise parameter combination corresponding to the particle.

[0105] (3) Update individual extreme values and global extreme values

[0106] For each particle, calculate the current fitness value. If the current fitness value is better than the historical optimal value of the particle, update the current fitness value and corresponding position to the individual optimal value (pbest) of the particle.

[0107] Find the particle with the best fitness value among the individual optimal values of all particles, and update its position and fitness value to the global optimal value (gbest).

[0108] (4) Update particle speed and position

[0109] Update the velocity of each particle according to the following formula:

[0110] v i t+1 =ωv i t +c1r1(pbest i -x i t )+c2r2(gbest i -x i t )

[0111] Among them, v i t+1 represents the velocity of particle i in the t+1th iteration, ω is the inertia weight, which is used to control the influence of the current velocity of the particle on the next velocity, c1 and c2 are individual and global learning factors respectively, r1 and r2 are randomly generated numbers in the interval [0,1], (pbest i ) is the individual optimal position of the particle, (gbest i ) is the global optimal position, x i t is the position of the particle at iteration t.

[0112] Update the particle's position according to the following formula:

[0113] xi t+i =x i t +v i t+1

[0114] Among them, x i t+1 Represents the position of the particle in the t+1th iteration, that is, the updated noise parameter combination.

[0115] (5) Boundary condition processing

[0116] If the particle position exceeds the preset solution space boundary, it is necessary to limit its position within the boundary to avoid unphysical values of the noise parameters. The rebound boundary method or position reset method can be used to handle particles that exceed the boundary so that the particle position is always within a reasonable range.

[0117] (6) Iterate until the stopping condition is met

[0118] Repeat steps (2) to (5) until a preset stopping condition is met. The stopping condition can be the maximum number of iterations or the global optimal fitness value is less than a threshold.

[0119] Finally, the globally optimal noise parameter combination found is applied to the Kalman filter for real-time data compensation to ensure the accuracy and stability of the filtering process.

[0120] (7) Application of optimal noise parameters

[0121] After particle swarm optimization, the optimal noise covariance matrices (Q) and (R) are found. These optimal parameters are applied to the Kalman filter for the next round of data filtering and real-time compensation. With each new measurement input, the PSO algorithm reassesses the noise level and iteratively optimizes the noise parameters, ensuring the Kalman filter always operates optimally and provides accurate data compensation.

[0122] 4. Real-time filtering and data compensation

[0123] The PSO algorithm updates the noise parameters after each measurement and reapplies the optimal parameters to the Kalman filter, thereby achieving real-time compensation for the measurement data. Through this adaptive filtering process, the Kalman filter can respond to noise changes in the measurement environment in real time, ensuring the accuracy of the filtering effect.

[0124] Figure 2The figure shows a comparison between the original current test data and the filtered data. The blue curve represents the original data, which has large fluctuations and contains a large amount of high-frequency noise, resulting in unstable test results. After Kalman filtering, the red curve shows that the filtered data is much smoother than the original data. By combining the PSO algorithm, the filter can adaptively adjust the noise parameters, significantly suppressing noise, preserving the main characteristics of the original signal, and improving the measurement accuracy of the current data. This demonstrates that this method is effective in noise suppression and provides a higher-quality signal foundation for subsequent data analysis.

[0125] Figure 3 The figure shows the comparison between the original voltage test data and the filtered data, which shows the effect of filtering on the voltage data. The blue curve represents the original voltage data. Although the amplitude fluctuation is not large, the noise has a significant impact on the data. After the PSO-optimized Kalman filter processing, the filtered data represented by the red curve becomes more stable. The green and blue dotted lines represent the upper and lower limits of the system voltage, respectively. The filtered data remains within this range and significantly reduces irregular fluctuations, ensuring the stability and consistency of the data. This shows that in voltage testing, the present invention can effectively suppress noise interference and make the data more reliable and stable.

[0126] Experimental results show that combining Kalman filtering with PSO optimization can effectively suppress random noise in current and voltage tests, making the signals smoother and staying within the expected operating range. This approach significantly improves the quality and accuracy of test data while preserving the original signal characteristics, playing a significant role in improving the overall yield and reliability of chip production and electrical testing processes.

[0127] 5. Yield improvement and feedback

[0128] Yield is a metric that measures the proportion of products that meet quality requirements. The data in Table 1 compares the yield before and after compensation. After obtaining high-precision filtered data, we further analyzed the yield performance during chip testing. Table 1 demonstrates the yield improvement effect of the real-time compensation method for chip test results on current and voltage test data.

[0129] Table 1 Kalman filter combined with PSO compensation test data before and after yield comparison

[0130] yield Current test yield Voltage test yield Before compensation 98.365％ 99.598％ After compensation 99.754％ 99.983％ Improvement value 1.389％ 0.385％

[0131] For current testing, the yield before compensation was 98.365%. After Kalman filtering and PSO compensation, the yield significantly increased to 99.754%. This change demonstrates that the compensation process effectively reduced noise and errors in the measurement data, enabling more products to meet quality standards. For voltage testing, the yield before compensation was 99.598%, increasing to 99.983% after compensation. Despite the already high yield before compensation, the further improvement after compensation demonstrates that this method continues to be effective in improving accuracy.

[0132] These results show that the real-time compensation method for chip test results of the present invention performs well in improving the quality of test data, can effectively reduce the influence of noise, and significantly improve the reliability and yield of measurement data. For current testing, the yield is greatly improved after compensation, indicating that the noise influence in the original data is significant, and the quality is significantly improved after compensation. For voltage testing, although the initial yield is high, it is still further improved after compensation, indicating that this method also has room for improvement under high quality requirements and can ensure the consistency and stability of product quality. Therefore, the compensation method combining Kalman filtering and PSO is applicable to various noise environments and helps to improve the overall yield and production quality in the chip testing process.

[0133] By reducing noise and bias in the measurement data, the proportion of defective products in the production process has been significantly reduced. Throughout the process, measurement results are continuously fed back to the PSO algorithm for parameter optimization, ensuring that the filter can adapt in real time to the ever-changing noise environment and maintain the best compensation effect.

[0134] In summary, the real-time compensation method for chip test results of the present invention not only effectively improves the accuracy and yield of chip testing, but also provides a more scientific and reasonable idea in chip resource management, providing strong technical support for the development of the integrated circuit testing field.

[0135] The present invention first performs preliminary processing on the measurement data through Kalman filtering. However, unlike traditional methods, the noise parameters are no longer fixed, but are adaptively optimized and adjusted with the help of the PSO algorithm. Specifically, during the data acquisition process, each time a new batch of measurement data is obtained, the PSO algorithm will evaluate the noise level in the current measurement environment and dynamically adjust the noise parameters of the Kalman filter so that the filtering process can adapt to the real-time changing noise environment and ensure the accuracy of data processing. At the same time, the compensation process is carried out in real time, and the Kalman filter works in conjunction with the PSO algorithm to perform instant compensation for each measurement data. This process is repeated continuously to ensure that the measuring equipment can always maintain high-precision measurement results in unstable or noisy environments, thereby improving the accuracy, yield and product quality of chip testing.

[0136] Example 2

[0137] An electronic device includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the real-time compensation method for chip test results based on the PSO-KF algorithm in embodiment 1, and the processor is configured to execute the program stored in the memory.

[0138] Example 3

[0139] A storage medium stores a computer program, which, when executed by a processor, executes the steps of the method for real-time compensation of chip test results based on the PSO-KF algorithm in embodiment 1.

[0140] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A real-time compensation method for chip test results based on PSO-KF algorithm, characterized in that: The following steps are involved: Step 1: Collect the original measurement data during the chip testing process and use the mode to fill in the missing values in the data; Step 2: Input the collected raw measurement data into the Kalman filter for preliminary filtering. The specific method is as follows: 1) State variable and measurement model initialization: Define the state variables of the Kalman filter according to the specific requirements of the chip test; initialize the initial values and covariance matrix of the state variables to represent the uncertainty of the system state at the initial moment; build the measurement model, establish the mathematical relationship between the state variables and the actual measurement data, and define the state transfer matrix A, control matrix B, and observation matrix H; 2) Prediction step: predict the state of the system based on the state transition equation, calculate the covariance matrix, and estimate the state value and uncertainty of the current data; 3) Update step: Combined with the current measurement value, the state estimate is updated through the Kalman gain to reduce the difference between the predicted value and the measured value; 4) Filtered data output: Output the updated state estimate as the estimate of the current state and used for subsequent data analysis and compensation; the filtered data will be fed back as input to the particle swarm optimization algorithm to optimize the noise parameters; Step 3: Based on the real-time changes of the measurement data and the environmental noise, the noise parameters in the Kalman filter are optimized using the particle swarm optimization algorithm. The specific method is as follows: 1) Initialize the particle swarm: determine the size of the particle swarm, that is, the number of particles N. Each particle represents a noise parameter combination of the noise covariance matrix Q and the observation noise covariance matrix R. Randomly initialize the position and velocity of each particle and set the initial parameter range. 2) Calculate the fitness function: For each particle, apply its current noise parameter combination to the Kalman filter to filter the measurement data; Define a fitness function based on the filtering error to evaluate the filtering effect; 3) Update individual and global extreme values: For each particle, calculate the current fitness value. If the current fitness value is better than the historical optimal value of the particle, update the current fitness value and corresponding position to the individual optimal value of the particle; find the particle with the best fitness value among the individual optimal values of all particles, and update its position and fitness value to the global optimal value; 4) Update particle velocity and position; 5) Boundary condition processing: Use the rebound boundary method or position reset method to process particles that exceed the boundary, so that the particle position is always within a reasonable range; 6) Repeat steps 2) to 5) until a preset stopping condition is met; 7) Application of optimal noise parameters: After particle swarm optimization, the optimal noise covariance matrix Q and observation noise covariance matrix R are found and applied to the Kalman filter for the next round of data filtering and real-time compensation; Step 4: The particle swarm optimization algorithm updates the noise parameters after obtaining each measurement data, and reapplies the optimal parameters to the Kalman filter, thereby achieving real-time compensation of the measurement data; Step 5: Evaluate the yield based on the compensated data, and the data will be fed back to the particle swarm optimization algorithm for the next iterative optimization to ensure that the Kalman filter can adapt to noise changes and maintain the best compensation effect.

2. The chip test result real-time compensation method based on the PSO-KF algorithm according to claim 1 is characterized in that: The method of predicting the state of the system based on the state transition equation, calculating the covariance matrix, and estimating the state value and uncertainty of the current data is as follows: Use the state transition equation to predict the estimated state value of the system at the current moment: in, represents the state prediction value at the current moment, A is the state transfer matrix, is the last state estimate, B is the control matrix, μ k-1 is the control input; Compute the covariance matrix of the predicted states: P k | k-1 =AP k-1 From T +Q Among them, P k|k-1 represents the prediction covariance matrix, P k-1 is the covariance matrix of the previous moment, and Q is the process noise covariance matrix.

3. The chip test result real-time compensation method based on the PSO-KF algorithm according to claim 2 is characterized in that: The method of combining the current measurement value and updating the state estimate value through the Kalman gain to reduce the difference between the predicted value and the measured value is as follows: Calculate the Kalman gain K k as follows: K k =P k | k-1 H T (HP k | k-1 H T +R) -1 Among them, K k represents the Kalman gain, H is the observation matrix, and R is the observation noise covariance matrix; The updated state estimate is combined with the measurement data as follows: in, represents the updated state estimate, z k is the actual measured value at the current moment, is the measurement residual, that is, the difference between the predicted value and the measured value; The updated state covariance matrix is as follows: P k =(I-K k H)P k | k-1 Among them, P k is the updated state covariance matrix, which represents the uncertainty of the updated state variables.

4. The chip test result real-time compensation method based on the PSO-KF algorithm according to claim 1 is characterized in that: The formula for updating particle velocity is as follows: v i t+1 =ωv i t +c1r1(pbest i -x i t )+c2r2(gbest i -x i t ) Among them, v i t+1 represents the velocity of particle i in the t+1th iteration, ω is the inertia weight, c1 and c2 are the individual and global learning factors respectively, r1 and r2 are randomly generated numbers in the interval [0,1], pbest i is the individual optimal position of the particle, gbest i is the global optimal position, x i t is the position of the particle at iteration t.

5. The chip test result real-time compensation method based on the PSO-KF algorithm according to claim 4 is characterized in that: The formula for updating the particle position is as follows: x i t+i =x i t +v i t+1 Among them, x i t+1 represents the position of the particle at the t+1th iteration.

6. The chip test result real-time compensation method based on the PSO-KF algorithm according to claim 1 is characterized in that: The stopping condition is that the maximum number of iterations or the global optimal fitness value is less than a threshold.

7. An electronic device comprising a memory and a processor, characterized in that: The memory is used to store a program that supports the processor to execute the real-time compensation method for chip test results based on the PSO-KF algorithm as described in any one of claims 1 to 6, and the processor is configured to execute the program stored in the memory.

8. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the real-time compensation method for chip test results based on the PSO-KF algorithm according to any one of claims 1 to 6 are executed.

Images