Semiconductor aging test method and system

Abstract

The invention discloses a semiconductor aging test method and system, and relates to the field of artificial intelligence and automatic testing, and the method comprises the steps: collecting the performance parameters of a semiconductor, carrying out the preprocessing, building a quantum platform, deploying a nanometer sensor network to monitor the semiconductor, obtaining the monitoring data of the semiconductor, and building an aging prediction model. Obtaining the aging degree of the semiconductor according to the collected performance parameters, analyzing the semiconductor monitoring data to obtain the state information of the semiconductor, and combining the aging degree and the state information of the semiconductor to form a complete semiconductor test result; besides, a time-frequency diagram of semiconductor monitoring data in a time-frequency domain is generated by using short-time Fourier transform, and the method captures a microstructure change trend and is beneficial to early discovery of potential problems.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence and automated testing, and particularly to a semiconductor aging test method and system. Background Art

[0002] The aging test of semiconductor devices is an important link to ensure their long-term reliability. Especially in modern electronic devices, the performance and lifespan of semiconductor devices directly affect the overall stability and safety of the devices. In recent years, with the rapid development of semiconductor technology, the requirements for aging test methods have also been increasing.

[0003] Common semiconductor aging test methods mainly include electrical parameter testing and thermal analysis, etc. Electrical parameter testing can only provide information at the macroscopic level and is difficult to reveal changes in the microscopic structure. Thermal analysis indirectly infers the stress distribution inside the device by monitoring the temperature distribution. This method has low sensitivity for detecting early defects and cannot monitor in real time. In addition, most existing machine learning methods are limited to classical computing platforms and do not fully utilize the powerful processing capabilities of quantum computing. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a semiconductor aging test method to solve the problems of the lack of microscopic structure changes and incompleteness in traditional semiconductor aging test methods.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a semiconductor aging test method, which includes

[0008] Collecting the performance parameters of the semiconductor and performing preprocessing;

[0009] Building a quantum platform, deploying a nano-sensor network to monitor the semiconductor, and obtaining semiconductor monitoring data;

[0010] Constructing an aging prediction model to obtain the aging degree of the semiconductor based on the collected performance parameters;

[0011] Analyzing the semiconductor monitoring data to obtain the state information of the semiconductor;

[0012] Combining the aging degree and state information of the semiconductor to form a complete semiconductor test result;

[0013] Providing maintenance suggestions and visualization icons according to the semiconductor test result.

[0014] As a preferred embodiment of the semiconductor aging test method of the present invention, it includes: collecting the performance parameters of the semiconductor and performing preprocessing, specifically including the following steps,

[0015] Collect the working voltage, current, temperature, and humidity of the semiconductor in real time as performance parameters;

[0016] Clean, convert the format, remove outliers, and fill in missing values for the collected performance parameters;

[0017] Standardize the cleaned data.

[0018] As a preferred embodiment of the semiconductor aging test method of the present invention, it includes: building a quantum platform, deploying a nano-sensor network to monitor the semiconductor, and obtaining semiconductor monitoring data, specifically including the following steps,

[0019] Select IBM Quantum as the quantum cloud service platform;

[0020] Configure the hardware interface between the classical computer and the quantum processor, and write an initialization script to set the initial state vector;

[0021] Arrange a nano-sensor array at the critical positions of the semiconductor to form a densely distributed sensing network;

[0022] The nano-sensor network will monitor the state of the semiconductor in real time to obtain monitoring data.

[0023] As a preferred embodiment of the semiconductor aging test method of the present invention, it includes: building an aging prediction model, and obtaining the aging degree of the semiconductor according to the collected performance parameters, specifically including the following steps,

[0024] Define a feature extraction function to capture the implicit patterns in the collected performance parameters through integral transformation, and its expression is:

[0025]

[0026] Among them, F(x) represents the result of the captured implicit pattern, x represents the performance parameter vector, α represents the regularization coefficient, t represents time, n represents the number of performance parameters, x' i represents the i-th standardized performance parameter, and dt represents the differential symbol;

[0027] Map the result of the implicit pattern to a quantum state;

[0028] According to the result of the hidden pattern mapped to the quantum state, introduce the results of the hidden patterns of multiple quantum states. Use the quantum kernel function to calculate the inner product similarity of the results of the hidden patterns of any two quantum states in the quantum state space, and form a quantum kernel matrix with the inner product similarities between the results of all quantum state hidden patterns;

[0029] Define an information filtering function to extract the feature similarity information related to the aging characteristics from the quantum kernel matrix, and emphasize the high similarity feature pairs. Its expression is:

[0030]

[0031] where \(G(K)\) represents extracting the feature similarity information related to the aging characteristics from the quantum kernel matrix \(K\), \(K\) represents the quantum kernel matrix, \(K_{ij}\) ij represents the inner product similarity of the results of the \(i\)-th and \(j\)-th quantum state hidden patterns in the quantum state space, \(\beta\) represents the temperature parameter, \(K_{ii}\) ii represents the similarity of the result of the \(i\)-th quantum state hidden pattern with itself, \(K_{jj}\) jj represents the similarity of the result of the \(j\)-th quantum state hidden pattern with itself;

[0032] Combine the feature similarity information extracted from the quantum kernel matrix related to the aging characteristics into a feature similarity information vector, and input it into the fully connected layer. Apply the softmax function to obtain the probabilities of the semiconductor in different aging states. Its expression is:

[0033]

[0034] where \(P(z)_a\) a represents the probability that the semiconductor is in the \(a\)-th aging state based on the feature similarity information vector \(z\), \(G(K)_a\) a represents the feature similarity information related to the \(a\)-th aging state in the feature similarity information extracted from the quantum kernel matrix \(K\) related to the aging characteristics, \(L\) represents the total number of aging states, and \(l\) represents the index variable;

[0035] Based on the probabilities of the semiconductor in different aging states, obtain the aging degree of the semiconductor.

[0036] As a preferred scheme of the semiconductor aging test method described in the present invention, wherein: analyze the semiconductor monitoring data to obtain the state information of the semiconductor, specifically including the following steps,

[0037] Use the short-time Fourier transform to generate the time-frequency diagram of the semiconductor monitoring data in the time-frequency domain;

[0038] Extract the maximum peak frequency and the average power spectral density from the time-frequency diagram, and combine them into a microscopic feature vector;

[0039] Based on the microscopic feature vectors, a convolutional neural network is introduced;

[0040] Using the method of supervised learning and taking cross-entropy as the loss function, the convolutional neural network is trained to obtain the state information of the semiconductor;

[0041] The state information of the semiconductor includes the microscopic structure changes and defect initiation of the semiconductor.

[0042] As a preferred embodiment of the semiconductor aging test method of the present invention, wherein: the aging degree and state information of the semiconductor are combined to form a complete semiconductor test result, which specifically includes the following steps,

[0043] Define a comprehensive evaluation function to combine the aging degree and state information of the semiconductor to form a comprehensive semiconductor test score, and its expression is:

[0044] Z = λ1 × A(P(z) a ) + λ2 × Y;

[0045] Wherein, Z represents the comprehensive semiconductor test score, λ1 represents the weight of the semiconductor aging degree, A(P(z) a ) represents the aging degree of the semiconductor, λ2 represents the weight of the semiconductor state information, and Y represents the state information of the semiconductor;

[0046] Based on the semiconductor historical data, a test threshold is set, and a complete semiconductor test result is obtained according to the interval where the comprehensive semiconductor test score is located with respect to the test threshold.

[0047] As a preferred embodiment of the semiconductor aging test method of the present invention, wherein: according to the semiconductor test result, maintenance suggestions and visualization icons are provided, which specifically include the following steps,

[0048] According to the semiconductor test result, specific maintenance suggestions are provided;

[0049] Using Matplotlib of Python, a visualization chart of the aging degree and state information of the semiconductor is drawn;

[0050] The results of each test are recorded in the database.

[0051] In a second aspect, the present invention provides a semiconductor aging test system, including,

[0052] A preprocessing module that collects the performance parameters of the semiconductor and performs preprocessing;

[0053] A monitoring module that builds a quantum platform and deploys a nano-sensor network to monitor the semiconductor to obtain monitored semiconductor data;

[0054] A prediction module that constructs an aging prediction model and obtains the aging trend of a semiconductor based on the collected performance parameters;

[0055] An analysis module that analyzes semiconductor monitoring data to obtain the status information of the semiconductor;

[0056] A testing module that combines the aging trend and status information of the semiconductor to form a complete semiconductor test result;

[0057] A maintenance module that provides maintenance suggestions and visualization icons based on the semiconductor test results.

[0058] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the semiconductor aging test method described in the first aspect of the present invention is implemented.

[0059] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the semiconductor aging test method described in the first aspect of the present invention is implemented.

[0060] The beneficial effects of the present invention are as follows: An aging prediction model is established. By defining a feature extraction function and using integral transformation to capture the implicit patterns in the collected data, the deep mining of semiconductor aging characteristics is realized. This method reveals the implicit patterns that are difficult to capture by traditional methods, improves the accuracy of aging prediction, and is particularly significant for the identification of early aging signs; in addition, the short-time Fourier transform is used to generate the time-frequency diagram of semiconductor monitoring data in the time-frequency domain. This method captures the trend of microstructural changes, helps to detect potential problems early, reduces the failure risk, and is particularly excellent in detecting the initiation of early defects. Description of the Drawings

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0062] Figure 1 It is a flowchart of the semiconductor aging test method in Embodiment 1.

[0063] Figure 2 It is a schematic diagram of generating a test result in Embodiment 1. Detailed Embodiments

[0064] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following detailed description of the specific embodiments of the present invention will be provided in conjunction with the accompanying drawings of the specification.

[0065] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0066] Secondly, as used herein, "one embodiment" or "an embodiment" refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other.

[0067] Example 1, referring to Figure 1 and Figure 2 , is the first embodiment of the present invention. This embodiment provides a semiconductor aging test method, including the following steps:

[0068] S1. Collect the performance parameters of the semiconductor and perform preprocessing.

[0069] Specifically, it includes the following steps.

[0070] Install intelligent sensors with IIoT functions on the production line. These sensors monitor parameters such as the working voltage, current, temperature, and humidity of semiconductor devices as performance parameters.

[0071] Use ADAS for data collection. ADAS is a unified automated data collection platform that collects the performance parameters of sensors in real time through the IIoT platform. IIoT is the Industrial Internet of Things.

[0072] Start ADAS, set a fixed time interval (for example, once per minute), and use the Network Time Protocol (NTP) server to synchronize the time of all data collection nodes.

[0073] S1.1. Clean and convert the format of the collected performance parameters. Remove outliers (such as measurement results outside the reasonable range) and fill in missing values (such as using interpolation). Then, apply the Z-score normalization formula to the cleaned data to unify the format for subsequent analysis.

[0074] Further explanation is that it lays a solid foundation for constructing an efficient aging prediction model in the future.

[0075] S2. Build a quantum platform, deploy a nano-sensor network to monitor the semiconductor, and obtain semiconductor monitoring data.

[0076] Specifically, it includes the following steps:

[0077] Select IBM Quantum as the quantum cloud service platform. Utilize its advanced quantum computing resources and technical support to connect to IBM Quantum through the API interface, ensuring that the classical computer can seamlessly access the quantum processor.

[0078] Configure the communication interface between the classical computer and the quantum processor to ensure efficient data transmission between the two.

[0079] Write a Python script to initialize the quantum circuit. This script will set the initial state vector (representing a uniform superposition state) and define a series of common quantum gate operations. For example, the Hadamard gate: used to create a superposition state; the Pauli-X gate: equivalent to the classical NOT gate, which flips the state of the quantum bit.

[0080] Arrange a graphene-based nanosensor array at key positions in the semiconductor (such as the gate, source, and drain). Connect the individual nanosensors in a wired or wireless manner to form a densely distributed sensing network, enabling centralized management and real-time transmission of data.

[0081] To maintain the long-term stable operation of the nanosensor network, select an efficient switching-mode DC-DC converter and pair it with a large-capacity supercapacitor to ensure the stable operation of the nanosensor network.

[0082] S3. Build an aging prediction model to obtain the aging degree of the semiconductor based on the collected performance parameters.

[0083] Specifically, it includes the following steps:

[0084] Define a feature extraction function to capture the hidden patterns in the collected performance parameters through integral transformation. Hidden patterns refer to the potential structures or regularities that are not directly obvious in the original data but can be revealed through calculations. For example, changes in the aging rate and periodic behavior. Its expression is:

[0085]

[0086] where F(x) represents the result of the captured hidden pattern, x represents the performance parameter vector, α represents the regularization coefficient, t represents time, n represents the number of performance parameters, and x' h represents the h-th standardized performance parameter, and dt represents the differential symbol;

[0087] Furthermore, by using the Gaussian kernel for integral transformation, hidden patterns in the time domain and frequency domain can be effectively captured, revealing potential aging characteristics.

[0088] Map the result of the implicit mode to the quantum state φ(F(x))> using a mapping function. Mapping the result of the implicit mode to a quantum state can utilize the powerful parallel processing ability and complex pattern recognition ability of quantum computing to significantly improve the data analysis efficiency;

[0089] According to the result of the implicit mode mapped to the quantum state, introduce the results of the implicit modes of multiple quantum states, calculate the inner product similarity of the results of the implicit modes of any two quantum states in the quantum state space using the quantum kernel function, and form a quantum kernel matrix from the inner product similarities between the results of all implicit modes of the quantum state. Its expression is:

[0090] K ij =|<φ(F(x i ))|φ(F(x j ))>| 2 ;

[0091] where K ij represents the inner product similarity between the i-th and j-th implicit mode results in the quantum state space in the quantum kernel matrix, F(x i ) represents the i-th captured implicit mode result, F(x j ) represents the j-th captured implicit mode result, and φ() represents the mapping function.

[0092] Furthermore, by calculating the inner product similarity, the similarity degree between different samples can be quantified, providing a basis for subsequent information filtering. And the quantum kernel matrix provides a global perspective, reflecting the mutual relationship between all samples, which helps to discover potential patterns and associations.

[0093] Define an information filtering function to extract the feature similarity information related to the aging characteristics from the quantum kernel matrix, and emphasize the high-similarity feature pairs (those feature pairs that show high similarity between different samples will be given higher weights because they are likely to capture the key patterns in the data). Its expression is:

[0094]

[0095] where G(K) represents extracting the feature similarity information related to the aging characteristics from the quantum kernel matrix K, K represents the quantum kernel matrix, K ij represents the inner product similarity between the i-th and j-th quantum state implicit mode results in the quantum state space, β represents the temperature parameter, K ii represents the similarity of the i-th quantum state implicit mode result with itself, and K jj represents the similarity of the j-th quantum state implicit mode result with itself;

[0096] Furthermore, the exponential decay factor can highlight high - similarity feature pairs and reduce the influence of low - similarity features, making the aging prediction model pay more attention to important features.

[0097] Define three aging states: normal, mild aging, and severe aging.

[0098] Extract the feature similarity information related to aging features from the quantum kernel matrix and combine it into a feature similarity information vector z, and input it into the fully - connected layer. Apply the softmax function to obtain the probabilities of the semiconductor in different aging states. Its expression is:

[0099]

[0100] where P(z) a represents the probability that the semiconductor is in the a - th aging state based on the feature similarity information vector z, G(K) a represents the feature similarity information related to the a - th aging state among the feature similarity information related to aging features extracted from the quantum kernel matrix K. L represents the total number of aging states, and l represents the index variable;

[0101] According to the obtained probabilities of the semiconductor in different aging states, it can be judged based on the highest probability. For example, if the output probabilities are [0.1, 0.6, 0.3], then it is judged that the semiconductor is currently in the mild - aging state.

[0102] Similarly, weighted average can be used and combined with the probabilities of all states for comprehensive scoring. Its expression is:

[0103]

[0104] where w a represents the weight corresponding to the a - th aging state.

[0105] First, set a weight for the probability P(z) a of each aging state. The setting of the weight can reflect the importance and severity of different aging states. For example, for the normal state, the weight w1 is equal to 0, which means that if the device is in the normal state, it has no contribution to the aging score; for the mild - aging state, the weight w2 is equal to 0.5, which means that if the device is in the mild - aging state, the aging score will increase by half; for the severe - aging state, the weight w3 is equal to 1, which means that if the device is in the severe - aging state, the aging score will double.

[0106] These weights can be set according to actual needs to reflect the importance of different aging states to the overall evaluation.

[0107] Suppose the output probabilities are [0.7, 0.2, 0.1], then:

[0108] Degree of aging = 0.7×0 + 0.2×0.5 + 0.1×1;

[0109] Degree of aging = 0 + 0.1 + 0.1;

[0110] Degree of aging = 0.2.

[0111] Set an aging threshold in the range of [0, 1], and obtain the aging degree of the semiconductor according to the interval where the value of the aging degree is located with respect to the aging threshold.

[0112] For example, if the value of the aging degree is less than 0.3, it is determined that the semiconductor is in a normal state; if the value of the aging degree is greater than or equal to 0.3 and less than 0.7, it is determined that the semiconductor is in a mild aging state; if the value of the aging degree is greater than or equal to 0.7 and less than or equal to 1, it is determined that the semiconductor is in a severe aging state.

[0113] The specific setting of the aging threshold can be customized according to the usage scenario and actual situation.

[0114] S4. Analyze the semiconductor monitoring data to obtain the status information of the semiconductor.

[0115] Specifically, it includes the following steps

[0116] S4.1. Use the short-time Fourier transform to generate the time-frequency diagram of the semiconductor monitoring data in the time-frequency domain. Since the short-time Fourier transform can provide both time information and frequency information, it is particularly suitable for analyzing short-term phenomena in non-stationary signals, such as microstructural changes in semiconductor devices. Its expression is:

[0117]

[0118] Among them, S(t, f) represents the time-frequency diagram, representing the signal intensity at time point t and frequency f, v(τ) represents the semiconductor monitoring data, τ represents the time variable, r() represents the window, usually a Hanning window, Gaussian window, etc., which is used to limit the range of each Fourier transform, and q represents the imaginary unit.

[0119] Obtain the time-frequency diagram slices from the time-frequency diagram, which is equivalent to intercepting a series of vertical slices along the time axis on the time-frequency diagram;

[0120] Calculate the amplitude at each frequency f, and find the maximum value of the amplitude and its corresponding frequency;

[0121] The frequency corresponding to the maximum value is the maximum peak frequency;

[0122] Furthermore, the maximum peak frequency reflects the strongest frequency component at each time point, which helps to identify transient events or change trends.

[0123] Calculate the power spectral density based on the time-frequency diagram. The power spectral density is the energy distribution of the signal at different frequencies;

[0124] For each frequency f, calculate the average value of the power spectral density at all time points t, which is the average power spectral density;

[0125] Furthermore, the average power spectral density provides the power distribution over the entire time range, and can reveal the main frequency components of the signal and their relative intensities.

[0126] Use the splicing method to form a microscopic feature vector from the maximum peak frequency and the average power spectral density.

[0127] S4.2. Select the convolutional neural network (CNN) as the main model because it is good at processing two-dimensional data (such as time-frequency diagrams) and can automatically learn complex patterns;

[0128] Use the cross-entropy loss function to measure the difference between the predicted value and the true label, and use the backpropagation algorithm to update the CNN parameters to minimize the loss function and obtain the state information of the semiconductor;

[0129] The state information can detect potential aging signs at an early stage through continuous monitoring of microscopic features, achieve timely warning, and prevent the occurrence of failures.

[0130] S5. Combine the aging degree and state information of the semiconductor to form a complete semiconductor test result.

[0131] Specifically, it includes the following steps:

[0132] Define the comprehensive evaluation function Z, which combines the aging degree and state information of the semiconductor by weighted summation to form a comprehensive semiconductor test score. Its expression is:

[0133] Z = λ1 × A(P(z) a ) + λ2 × Y;

[0134] Among them, Z represents the comprehensive semiconductor test score, λ1 represents the weight of the semiconductor aging degree, reflecting the importance of the aging degree in the final score, A(P(z) a ) represents the aging degree of the semiconductor, λ2 represents the weight of the semiconductor state information, and Y represents the state information of the semiconductor;

[0135] The settings of the semiconductor aging degree weight and the semiconductor state information weight can be customized according to actual needs.

[0136] Use the Sigmoid function to limit the comprehensive semiconductor test score between [0, 1].

[0137] Set a test threshold based on the semiconductor historical data, and set three intervals between [0, 1] to represent different health levels; for example, good, need attention, and emergency maintenance.

[0138] After obtaining the comprehensive semiconductor test score, it can be compared with the test threshold interval to obtain the complete semiconductor test result.

[0139] Furthermore, combining the aging degree and status information makes the evaluation more comprehensive and reduces the risk of misjudgment that may be caused by relying on a single indicator only.

[0140] S6. Provide maintenance suggestions and visualization icons according to the semiconductor test results.

[0141] Specifically, it includes the following steps.

[0142] If the semiconductor test result indicates a good state, continue to monitor and no immediate action is required.

[0143] If the semiconductor test result indicates that attention is needed, increase the monitoring frequency, prepare a preventive maintenance plan, and consider replacing key components or optimizing the working conditions at the same time.

[0144] If the semiconductor test result indicates emergency maintenance, stop using the device, conduct a detailed diagnosis, and formulate a detailed maintenance plan to ensure that it resumes operation as soon as possible.

[0145] Use Matplotlib in Python to draw a visualization chart of the aging degree and status information of the semiconductor. Matplotlib supports various chart types, such as line charts, bar charts, scatter plots, pie charts, radar charts, heat maps, etc., and can meet the data display needs of different types.

[0146] Select a relational database (such as MySQL, PostgreSQL) or a NoSQL database (such as MongoDB), select the appropriate database type according to the requirements, and record each test result in the database for subsequent reference.

[0147] This embodiment also provides a semiconductor aging test system, including:

[0148] A preprocessing module that collects the performance parameters of the semiconductor and performs preprocessing;

[0149] A monitoring module that builds a quantum platform and deploys a nano-sensor network to monitor the semiconductor to obtain monitored semiconductor data;

[0150] A prediction module that constructs an aging prediction model and obtains the aging trend of the semiconductor according to the collected performance parameters;

[0151] A parsing module that parses semiconductor monitoring data to obtain the status information of the semiconductor;

[0152] A testing module that combines the aging trend and status information of the semiconductor to form a complete semiconductor test result;

[0153] A maintenance module that provides maintenance suggestions and visual icons based on the semiconductor test results.

[0154] This embodiment also provides a computer device applicable to the case of the semiconductor aging test method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the semiconductor aging test method proposed in the above embodiment.

[0155] The computer device can be a terminal. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, a touchpad, or a mouse, etc.

[0156] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the semiconductor aging test method proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM for short), Electrically Erasable Programmable Read-Only Memory (EEPROM for short), Erasable Programmable Read Only Memory (EPROM for short), Programmable Red-Only Memory (PROM for short), Read-Only Memory (ROM for short), magnetic memory, flash memory, a magnetic disk, or an optical disc.

[0157] In summary, the present invention achieves in-depth mining of semiconductor aging characteristics by: establishing an aging prediction model, defining a feature extraction function, and using integral transform to capture implicit patterns in the collected data. This method reveals implicit patterns that are difficult to capture with traditional methods, improves the accuracy of aging prediction, and is particularly important for identifying early signs of aging. In addition, short-time Fourier transform is used to generate time-frequency diagrams of semiconductor monitoring data in the time-frequency domain. This method captures the trend of microstructural changes, helps to discover potential problems early, and reduces the risk of failure, especially in detecting early defect initiation.

[0158] Example 2, referring to Table 1, is the second example of the present invention. To further verify the technical solution of the present invention, experimental simulation data of a semiconductor aging test method is provided.

[0159] In order to verify the effectiveness of the semiconductor aging test method proposed in the present invention, a set of experiments was designed to compare the comparison method in the prior art. The experiment uses four different semiconductor devices as test objects, which are respectively recorded as comparison method test device 1, comparison method test device 2, invention method test device 1 and invention method test device 2. The experimental environment is set under standard laboratory conditions, with the temperature controlled at 25℃±1℃ and the humidity maintained at 40%±5%.

[0160] First, for each test object, collect its operating voltage, current, temperature, humidity and other performance parameters and perform preprocessing. The specific steps include:

[0161] Real-time data collection: Use high-precision data collection equipment to record the above performance parameters every 1 second to ensure the accuracy and completeness of the data.

[0162] Data cleaning: Clean the collected data to remove outliers and fill in missing values. For example, data points that are beyond the normal range (such as voltage exceeding the rated value) are marked as abnormal and removed; missing values ​​are filled using linear interpolation.

[0163] Standardization: The cleaned data is standardized so that the values ​​of all performance parameters are distributed in the interval [0,1] to facilitate subsequent analysis.

[0164] Next, we built a quantum platform and deployed a nanosensor network to monitor semiconductors. We selected IBM Quantum as the quantum cloud service platform, configured the hardware interface between the classical computer and the quantum processor, and wrote an initialization script to set the initial state vector. We arranged nanosensor arrays at key locations of the semiconductor to form a densely distributed sensor network to monitor the state of the semiconductor in real time and obtain detailed monitoring data.

[0165] Constructing an aging prediction model is one of the core parts of the present invention. Define a feature extraction function to capture the implicit patterns in the collected data through integral transformation.

[0166] Furthermore, map the results of the implicit patterns to quantum states, and based on the results of the implicit patterns, use the quantum kernel function to calculate the inner product similarity of the results of any two implicit patterns in the quantum state space to form a quantum kernel matrix. Then, define an information filtering function to extract the feature similarity information related to the aging characteristics from the quantum kernel matrix and emphasize the high-similarity feature pairs.

[0167] Finally, combine the feature similarity information related to the aging characteristics extracted from the quantum kernel matrix into a feature similarity information vector and input it into the fully connected layer, and apply the softmax function to obtain the probabilities of the semiconductor being in different aging states, so as to determine the aging degree of the semiconductor.

[0168] When analyzing semiconductor monitoring data, use the short-time Fourier transform to generate a time-frequency diagram, extract the maximum peak frequency and the average power spectral density from it, and combine them into a microscopic feature vector. Based on this microscopic feature vector, introduce a convolutional neural network (CNN), and use the supervised learning method with cross-entropy as the loss function to train the CNN to obtain the state information of the semiconductor, including microscopic structure changes and defect initiation.

[0169] When analyzing semiconductor monitoring data, use the short-time Fourier transform to generate a time-frequency diagram, extract the maximum peak frequency and the average power spectral density from it, and combine them into a microscopic feature vector. Based on this microscopic feature vector, introduce a convolutional neural network (CNN), and use the supervised learning method with cross-entropy as the loss function to train the CNN to obtain the state information of the semiconductor, including microscopic structure changes and defect initiation.

[0170] Finally, combine the aging degree and the state information of the semiconductor, define a comprehensive evaluation function to obtain the complete semiconductor test results, and provide specific maintenance suggestions and visualization charts. Each test result is recorded in the database for long-term tracking and analysis.

[0171] As shown in Table 1 below:

[0172] Table 1 Test Comparison Table

[0173]

[0174] Through the analysis of the above table content, it can be clearly seen that the method of the present invention has significant advantages and innovative effects compared with the comparative method:

[0175] Aging degree score: The aging degree scores of the devices tested by the comparative method are 0.73 and 0.81, while the aging degree scores of the devices tested by the method of the present invention are 0.59 and 0.63. This indicates that the method of the present invention can identify the aging signs of semiconductors earlier, so as to take preventive measures in advance and extend the device life.

[0176] Status information score: The status information scores of the devices tested by the comparative method are 0.68 and 0.72, while the status information scores of the devices tested by the method of the present invention are 0.86 and 0.89. This shows that the method of the present invention can capture the status information of semiconductors more comprehensively and accurately, especially in aspects such as microstructure changes, providing a more detailed description of the health status.

[0177] Comprehensive score: Although the comprehensive scores are similar (0.71, 0.77 and 0.72, 0.76 respectively), the method of the present invention shows higher accuracy in both aging degree and status information scores, making the comprehensive score more reliable. In addition, since the method of the present invention can detect potential problems earlier, the comprehensive score can better reflect the actual health status.

[0178] Maintenance suggestions: The maintenance suggestions provided by the method of the present invention are more accurate. It can not only identify the situations that need attention, but also give more targeted operation suggestions. For example, the suggestions of "continue to monitor" and "increase the monitoring frequency" are both based on detailed status information, which helps to formulate a reasonable maintenance plan and avoid unnecessary downtime.

[0179] Visualization chart score: The method of the present invention is significantly better than the comparative method in terms of the visualization chart score, which are 0.94 and 0.91 respectively. This is mainly because the method of the present invention can generate more intuitive and detailed charts, helping technicians quickly understand the meaning behind the data and improving the decision-making efficiency.

[0180] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not restrictive. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A semiconductor aging test method, characterized in that: include, Collect semiconductor performance parameters and perform preprocessing; Build a quantum platform and deploy a nanosensor network to monitor semiconductors and obtain semiconductor monitoring data; Construct an aging prediction model to obtain the aging degree of the semiconductor based on the collected performance parameters; Analyze semiconductor monitoring data to obtain semiconductor status information; Combine the semiconductor aging degree and status information to form a complete semiconductor test result; Provides maintenance recommendations and visualization icons based on semiconductor test results.

2. The semiconductor aging test method according to claim 1, wherein: Collect the performance parameters of the semiconductor and perform preprocessing, which specifically includes the following steps: Real-time collection of semiconductor operating voltage, current, temperature, and humidity as performance parameters; Clean the collected performance parameters, convert their formats, remove outliers, and fill in missing values; The cleaned data are standardized.

3. The semiconductor aging test method according to claim 2, wherein: Build a quantum platform, deploy a nanosensor network to monitor semiconductors, and obtain semiconductor monitoring data. The specific steps include the following: Select IBM Quantum as the quantum cloud service platform; Configure the hardware interface between the classical computer and the quantum processor, and write an initialization script to set the initial state vector; Arrange nanosensor arrays at key locations on semiconductors to form a densely distributed sensor network; The nanosensor network monitors the status of the semiconductor in real time and obtains monitoring data.

4. The semiconductor aging test method according to claim 3, wherein: Construct an aging prediction model to obtain the aging degree of the semiconductor based on the collected performance parameters. The specific steps include: Define the feature extraction function to capture the implicit pattern in the collected performance parameters through integral transformation, and its expression is: Among them, F(x) represents the result of capturing the implicit pattern, x represents the performance parameter vector, α represents the regularization coefficient, t represents the time, n represents the number of performance parameters, and x' i represents the i-th standardized performance parameter, dt represents the differential sign; Mapping the results of the hidden pattern into quantum states; According to the implicit pattern results mapped to quantum states, multiple implicit pattern results of quantum states are introduced, and the inner product similarity of the results of the implicit patterns of any two quantum states in the quantum state space is calculated using the quantum kernel function, and the inner product similarities between the results of the implicit patterns of all quantum states form a quantum kernel matrix; An information filtering function is defined to extract feature similarity information related to aging features from the quantum kernel matrix and emphasize high similarity feature pairs. Its expression is: Where G(K) represents the feature similarity information related to aging features extracted from the quantum kernel matrix K, K represents the quantum kernel matrix, K ij represents the inner product similarity of the results of the hidden patterns of the i-th and j-th quantum states in the quantum state space, β represents the temperature parameter, K ii represents the similarity between the result of the hidden pattern of the i-th quantum state and itself, K jj Indicates the similarity between the result of the hidden pattern of the jth quantum state and itself; The feature similarity information related to the aging characteristics extracted from the quantum kernel matrix is ​​combined into a feature similarity information vector and input into the fully connected layer. The softmax function is applied to obtain the probability of the semiconductor being in different aging states, which is expressed as: Among them, P(z) a represents the probability that the semiconductor is in the ath aging state based on the feature similarity information vector z, G(K) a It represents the feature similarity information related to the a-th aging state extracted from the feature similarity information related to the aging feature from the quantum kernel matrix K, L represents the total number of aging states, and l represents the index variable; Based on the probability of the semiconductor being in different aging states, the aging degree of the semiconductor is obtained.

5. The semiconductor aging test method according to claim 4, characterized in that: Parsing semiconductor monitoring data to obtain semiconductor status information includes the following steps: Generate time-frequency diagrams of semiconductor monitoring data in the time-frequency domain using short-time Fourier transform; The maximum peak frequency and average power spectrum density are extracted from the time-frequency diagram and combined into a micro-feature vector; Based on microscopic feature vectors, convolutional neural networks are introduced; By using supervised learning and cross entropy as the loss function, the convolutional neural network is trained to obtain the state information of the semiconductor. The state information of the semiconductor includes microstructure changes and defect initiation of the semiconductor.

6. The semiconductor aging test method according to claim 5, characterized in that: Combining the semiconductor aging degree and status information to form a complete semiconductor test result includes the following steps: A comprehensive evaluation function is defined to combine the semiconductor aging degree and status information to form a comprehensive semiconductor test score, which is expressed as: Z=λ1×A(P(z) a )+λ2×Y; Where Z represents the comprehensive semiconductor test score, λ1 represents the weight of semiconductor aging degree, A(P(z) a ) represents the aging degree of the semiconductor, λ2 represents the weight of the semiconductor state information, and Y represents the semiconductor state information; A test threshold is established based on semiconductor historical data, and a complete semiconductor test result is obtained according to the interval of the test threshold according to the comprehensive semiconductor test score.

7. The semiconductor aging test method according to claim 6, wherein: Provide maintenance suggestions and visualization icons based on semiconductor test results, including the following steps: Provide specific maintenance recommendations based on semiconductor test results; Use Python's Matplotlib to draw a visual chart of the semiconductor's aging and status information; The results of each test are recorded in the database.

8. A semiconductor aging test system, based on the semiconductor aging test method according to any one of claims 1 to 7, characterized in that: include, A preprocessing module collects semiconductor performance parameters and performs preprocessing; Monitoring module, build a quantum platform, deploy a nanosensor network to monitor semiconductors, and obtain monitoring semiconductor data; The prediction module builds an aging prediction model and obtains the aging trend of the semiconductor based on the collected performance parameters; An analysis module analyzes semiconductor monitoring data to obtain semiconductor status information; The test module combines the semiconductor aging trend and status information to form a complete semiconductor test result; The maintenance module provides maintenance recommendations and visualization icons based on semiconductor test results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the semiconductor aging test method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the semiconductor aging test method according to any one of claims 1 to 7 are implemented.

Images