Automatic testing method and system for credential equipment

Abstract

The invention discloses an automatic testing method and system for credential equipment, and relates to the technical field of credential equipment test.The method comprises the steps that testing requirements input by a user are analyzed through a natural language interface, key parameters are extracted, historical cases are retrieved, and a structured testing requirement table is generated; based on the structured test demand table, constructing a multi-objective optimization model, and outputting a test strategy matrix after iterative updating by using an NSGA-III algorithm improved by chaotic mapping; performing feature extraction and anomaly detection on the test logs, and identifying anomaly features through a dynamic threshold model; constructing a fault propagation graph model based on the abnormal features, positioning a root cause node by applying an improved causal reasoning algorithm, and generating a strategy iteration scheme; the improved NSGA-III algorithm initializes a population and dynamic target space adaptation mechanism through chaotic mapping, so that the Pareto front coverage rate is improved, and meanwhile, in combination with a resource weight allocation formula of an aging-time decay factor, collaborative optimization of the test cost and the extreme scene coverage rate is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of information and communication technology (ICT) equipment testing, and particularly to an automatic testing method and system for ICT equipment. Background Art

[0002] In recent years, multi-modal intelligent testing systems have gradually become a research hotspot in the fields of industrial inspection and equipment verification. The core challenge lies in how to fuse multi-source heterogeneous data (text, speech, image) to achieve accurate parameter extraction and generate robust testing strategies through dynamic optimization algorithms. Existing technologies usually adopt rule-based multi-modal semantic alignment methods and combine static knowledge bases for parameter disambiguation, but there are limitations in dealing with domain-specific terms and dynamic semantic associations. In addition, test resource allocation is mostly based on linear programming or traditional genetic algorithms, and it is difficult to balance the conflict relationship among parameter approximation, test cost, and scenario coverage rate in a non-uniform objective space. In terms of environmental calibration, existing methods generally adopt a fixed-step adjustment strategy, resulting in low overshoot compensation efficiency and lacking a dynamic verification mechanism for impedance matching and signal reflection coefficient.

[0003] There are mainly two deficiencies in the current technology: First, the existing semantic disambiguation models do not combine the topological relevance of domain knowledge graphs, resulting in the core parameter screening being easily interfered by synonyms, and the probability calculation of candidate terms does not consider the dynamic weight allocation of multi-modal query vectors. Second, traditional multi-objective optimization algorithms (such as NSGA-II) are prone to falling into local optima when dealing with non-uniform Pareto fronts, especially in the coupled scenarios of test resource allocation and dynamic attenuation of aging factors, and it is difficult to maintain the balance between the solution set distribution and the convergence speed. Summary of the Invention

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

[0005] Therefore, the present invention provides an automatic testing method for ICT equipment to solve the problems of inaccurate parameter semantic disambiguation and insufficient distribution of multi-objective optimization solution sets in multi-modal testing scenarios.

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

[0007] In a first aspect, the present invention provides an automatic testing method for ICT equipment, which includes:

[0008] Parsing the test requirements input by the user through a natural language interface, extracting key parameters and retrieving historical cases, and generating a structured test requirement table;

[0009] Constructing a multi-objective optimization model, using the NSGA-III algorithm improved by chaotic mapping, and iteratively updating to output a test strategy matrix;

[0010] Build a physical test environment with a programmable power supply and a temperature control device to generate a real-time environmental status vector;

[0011] Combine the real-time environmental status vector and the test strategy matrix, and dynamically adjust the test parameters through an online learning algorithm to form a test log with environmental compensation;

[0012] Extract features and detect anomalies from the test log, and identify abnormal features through a dynamic threshold model;

[0013] Build a fault propagation graph model, apply an improved causal inference algorithm to locate the root cause node, and generate a policy iteration plan.

[0014] As a preferred solution of the automatic test method for the information technology application innovation device described in the present invention, wherein: the extraction of key parameters includes semantic disambiguation of the test requirements input by the user to generate candidate terms, and screening of the candidate terms.

[0015] As a preferred solution of the automatic test method for the information technology application innovation device described in the present invention, wherein: the construction of the multi-objective optimization model is to define the weighted deviation between the minimum test setting value and the target value, the minimum total test cost, and the maximum coverage of potential fault scenarios through a structured requirements test table.

[0016] As a preferred solution of the automatic test method for the information technology application innovation device described in the present invention, wherein: the NSGA-III algorithm improved by using a chaotic map, and the output test strategy matrix after iterative update includes the following steps,

[0017] Perform a chaotic map on the objective optimization model to generate an initial population and dynamically adjust it;

[0018] Use simulated binary crossover and polynomial mutation to perform multi-objective iterative optimization on the dynamically adjusted initial population to generate a Pareto optimal solution set;

[0019] Select the scenario with the highest scenario coverage rate in the Pareto optimal solution set as the test sequence;

[0020] Dynamically adjust the resource allocation weight according to the Pareto optimal solution set;

[0021] Combine the test sequence, the parameter combination in the Pareto optimal solution set, and the resource allocation weight to obtain the test strategy matrix.

[0022] As a preferred solution of the automatic test method for the information technology application innovation device described in the present invention, wherein: the dynamic adjustment of the test parameters through the online learning algorithm includes injecting bit error rate data into the test signal, adjusting the frequency offset of the test signal through the online learning algorithm, and dynamically changing the frequency adjustment depth of the test signal.

[0023] As a preferred solution of the automatic testing method for the information technology application innovation device of the present invention, wherein: the identification of abnormal features through the dynamic threshold model includes the following steps,

[0024] Use an LSTM autoencoder to learn the latent distribution of normal data through an encoding-decoding process, and calculate the reconstruction error as the deviation degree;

[0025] Calculate the mean and standard deviation of all deviation degrees;

[0026] Based on the mean and standard deviation of all deviation degrees, dynamically adjust the deviation degree threshold;

[0027] Compare the calculated deviation degree with the adjusted deviation degree threshold to identify abnormal features.

[0028] As a preferred solution of the automatic testing method for the information technology application innovation device of the present invention, wherein: the generation of the strategy iteration scheme includes the following steps,

[0029] Construct a fault propagation graph and calculate the root cause probability of each component;

[0030] Sort the root cause probabilities of each component in descending order, select the key root causes, and generate optimization suggestions;

[0031] According to the temperature sensitivity of the key root causes, adjust the environmental parameters to obtain the updated environmental configuration;

[0032] Merge the optimization suggestions and the updated environmental configuration into the test strategy matrix, define the priority rules, and update the test strategy matrix to generate the strategy iteration scheme.

[0033] In a second aspect, the present invention provides an automatic testing system for information technology application innovation devices, including,

[0034] A requirements analysis module, which parses the test requirements input by the user through a natural language interface, extracts key parameters and retrieves historical cases, and generates a structured test requirements table;

[0035] A strategy generation module, which constructs a multi-objective optimization model based on the structured test requirements table, uses the NSGA-III algorithm improved by chaotic mapping, and outputs the test strategy matrix after iterative update;

[0036] An environment control module, which constructs a physical test environment through a programmable power supply and a temperature control device according to the test strategy matrix, and generates a real-time environment state vector;

[0037] An adaptive testing module, which combines the real-time environment state vector and the test strategy matrix, starts a multi-dimensional test case execution engine, and dynamically adjusts test parameters through an online learning algorithm to form a structured test log;

[0038] Anomaly diagnosis module, which extracts features and detects anomalies from test logs, and identifies anomaly features through a dynamic threshold model;

[0039] Policy iteration module, which constructs a fault propagation graph model based on anomaly features, applies an improved causal reasoning algorithm to locate the root cause node, and generates a policy iteration scheme.

[0040] 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 Xinchuang device automatic test method described in the first aspect of the present invention is implemented.

[0041] 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 Xinchuang device automatic test method described in the first aspect of the present invention is implemented.

[0042] The beneficial effects of the present invention are as follows: calculating the candidate term probability using the cosine similarity between the pre-trained vector and the query vector, and screening out the accurate terms higher than the threshold in combination with the semantic topological relationship of the knowledge graph, the term misjudgment rate is reduced compared with the traditional disambiguation method based on word frequency; the improved NSGA-III algorithm initializes the population through chaotic mapping and the dynamic objective space adaptation mechanism, improving the Pareto front coverage rate. At the same time, combined with the resource weight allocation formula of the aging-time decay factor, the collaborative optimization of the test cost and the extreme scenario coverage rate is realized; the progressive approximation algorithm improves the convergence speed of the environmental parameters through the coarse adjustment-fine adjustment two-stage control and the discretized first-order inertia correction; through the closed-loop control of the impedance matching verification model and the voltage standing wave ratio, the reflection coefficient of the link calibration is reduced; the accuracy of the anomaly type determination is improved through the dynamic deviation threshold, and combined with the root cause location model of the improved PageRank algorithm, the recognition accuracy of the fault propagation path is improved. Description of the Drawings

[0043] 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.

[0044] Figure 1 It is a flowchart of the Xinchuang device automatic test method in Embodiment 1.

[0045] Figure 2 It is a module diagram of the Xinchuang device automatic test system in Embodiment 1.

[0046] Figure 3It is the logic diagram of the improved NSGA-III algorithm in Embodiment 1.

[0047] Figure 4 It is the flow chart of anomaly diagnosis and root cause location in Embodiment 1. Specific implementation manners

[0048] To make the above objects, features and advantages of the present invention more obvious and understandable, the specific implementation manners of the present invention will be described in detail below with reference to the accompanying drawings of the specification.

[0049] Many specific details are set forth in the following description in order to provide a thorough understanding of the present invention, but 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, so the present invention is not limited by the specific embodiments disclosed below.

[0050] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure or characteristic that may be included in at least one implementation manner 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 other embodiments.

[0051] Embodiment 1, referring to Figures 1 to 4 , which is the first embodiment of the present invention. This embodiment provides an automatic testing method for information technology application innovation equipment, including the following steps:

[0052] S1. Receive the text, language, and image input by the user, map the data of different modalities to a unified semantic space through a pre-trained multi-modal encoder to generate a query vector; construct a disambiguation matrix based on the domain knowledge graph, perform semantic disambiguation on the query vector, and calculate the candidate term probability. The expression is:

[0053]

[0054] where E(t u |Q) represents the probability of the candidate term h u in the original query Q input by the user, h u represents the u-th candidate term, u represents the candidate term index, Q represents the original query input by the user, represents the pre-trained vector of the candidate term h u , v represents the query vector, n represents the total number of terms in the knowledge graph, represents the semantic vector of the term in the knowledge graph, h b represents the b-th term in the knowledge graph, and b represents the query term index in the knowledge graph.

[0055] Set the term threshold to filter out terms whose candidate term probabilities are greater than the term threshold (e.g., filter out "85°C" instead of the ambiguous "high temperature"), and generate a core parameter set. Semantic disambiguation is to avoid confusion between synonyms such as "temperature" and "humidity".

[0056] Retrieve the Top-K similar cases from the knowledge graph, construct a historical case vector set, weight them through the attention mechanism to obtain the comprehensive parameter weights, automatically match the test standard template according to the parameter type, generate dynamic constraint conditions, dynamically calculate the allowable deviation through Monte Carlo simulation, perform demand structured mapping, and generate a structured test requirement table.

[0057] S2. The structured test requirement table includes environmental parameter categories, target values, and allowable deviations. Based on the structured test requirement table, a multi-objective optimization model for minimizing the weighted deviation between the test setting value and the target value, minimizing the total test cost, and maximizing the coverage of potential failure scenarios is defined as follows:

[0058] Parameter approximation degree objective function for minimizing the weighted deviation between the test setting value and the target value:

[0059]

[0060] Among them, f1(x) represents the parameter approximation degree objective function, i represents the parameter index, n represents the total number of parameters, x i represents the test setting value of the i-th parameter, w i represents the weight of the i-th parameter, y i represents the target value of the i-th parameter, and x represents the test setting value.

[0061] Minimize the test cost:

[0062]

[0063] Among them, f2(x) represents the test cost objective function, j represents the test resource index, m represents the total number of test resources, c j represents the cost of the j-th test resource.

[0064] Maximize the coverage of potential failure scenarios:

[0065]

[0066] Among them, f3(x) represents the scenario coverage rate objective function, k represents the extreme scenario index, P(k|x) represents the probability of the k-th extreme scenario under the test setting value x, and p represents the total number of extreme scenarios.

[0067] An improved NSGA-III algorithm is adopted. By performing chaotic mapping on the target optimization model to generate an initial population, the population diversity is enhanced, premature convergence is avoided, and the initial population is dynamically adjusted according to the target space distribution to adapt to the non-uniform target space and improve the Pareto front coverage rate. Simulated binary crossover and polynomial mutation are used to perform multi-objective iterative optimization on the dynamically adjusted initial population, and the solution set after iterative optimization is divided into multiple front levels through non-dominated sorting. The high-level solutions are preferentially retained. During this period, the niche retention method is used to calculate the crowding distance to maintain the distribution of the solution set in the target space, and finally a Pareto optimal solution set is generated. Sort the scenario coverage rates according to each Pareto solution in the Pareto optimal solution set, and select the scenario with the highest probability as its test sequence. Adjust the aging-time dynamic decay factor in the test values according to the generated Pareto optimal solution set, and the expression is:

[0068] Ψ i =1-exp(-λl i );

[0069] where, Ψ i represents the aging-time dynamic decay factor of the i-th parameter, λ represents the device aging factor, and l i represents the expected execution time of the i-th test resource during the test process.

[0070] Adjust the resource allocation weight according to the aging-time dynamic decay factor, and the expression is:

[0071]

[0072] where, R j represents the weight of the j-th test resource, I ij represents the indicator function, that is, it is 1 when the test parameter is test resource j of i, otherwise it is 0, and I ik represents 1 when the test resource is i and the extreme scenario is k, otherwise it is 0.

[0073] Combine the test sequence, the parameter combination in the Pareto optimal solution set, and the resource allocation weight to obtain the test strategy matrix.

[0074] S3. According to the test strategy matrix, analyze the defined initial environmental parameters, including the core voltage, initial temperature, initial humidity, and clock frequency reference, and send instructions to the programmable power supply through the GPIB / USB bus to set the voltage output and control the temperature control box to calibrate the initial environmental parameters.

[0075] Extract target environment parameters from the test strategy matrix, such as high-temperature test scenarios and high-humidity test scenarios; based on the target environment, use a progressive environment approximation algorithm to adjust the initial environment parameters in stages, including a rough adjustment stage and a fine adjustment stage; in the rough adjustment stage, when target -T current || > 10%·T target When, set a large step size and set a high rate factor according to the device safety margin for accelerated adjustment. When problems are found, use a discretized first-order inertia link for correction and limit the maximum single-step change in the large step size, and continue to iterate until the fine adjustment stage is reached, that is, when target -T current || ≤ 10%·T target When, start fine adjustment, set a small step size and a low rate factor to suppress overshoot, define the overshoot amount. When the fine adjustment result is greater than the overshoot amount, trigger the overshoot compensation mechanism for continuous iteration until the environment is stable, and obtain the calibrated environment parameters. Among them, T target represents the target environment parameter, and T current represents the initial environment parameter.

[0076] Based on the calibrated environment parameters, extract the test sequence requirements from the test strategy matrix. If the test item includes "PCIe endpoint to root complex latency test", then port direct connection is required. If the test item includes "multi-device concurrent bandwidth test", then it needs to be extended through a PCIe Switch; for the physical ports of the device under test, generate a connection matrix, and the expression is:

[0077]

[0078] Among them, C ad represents the connection strength coefficient between port a and d. When C ad = 1.0 represents direct connection, and strict impedance matching is required (such as direct connection between CPU and GPU). C ad = 0.5 represents connection through a switch, allowing a certain signal attenuation (such as extended peripherals). P direct represents the set of all port pairs that need to be directly connected, and P switch represents the set of all port pairs that need to be connected through a switch.

[0079] Traverse all C ad , and filter out the port pairs (a, d) where C ad = 1.0. For each direct connection link, calculate the matching resistance value, and the expression is:

[0080]

[0081] Among them, Denote the terminal resistance value, which is used to eliminate signal reflection. The calculation formula derivation is as follows:

[0082] Measure the signal reflection coefficient using a vector network analyzer When Z L = R term it is required that |Γ| < 0.1. Solve the simultaneous equations to obtain R term and verify the voltage standing wave ratio Z0 = 85Ω, which is the characteristic impedance of the transmission line defined by the PCIe 4.0 standard and is determined by the transmission line geometry and dielectric material. Γ represents the reflection coefficient, and Z L represents the equivalent load of the terminal resistance and the transmission line impedance; based on the impedance matching verification result, construct the dynamic test environment and generate the real-time environment state vector.

[0083] S4. Read the initial value of the fundamental frequency amplitude from the signal parameters in the test strategy matrix, and combine it with the terminal resistance value in the real-time environment state vector to calculate the actual fundamental frequency signal amplitude. The expression is:

[0084]

[0085] where A represents the actual fundamental frequency signal amplitude, and A base represents the initial value of the fundamental frequency signal amplitude.

[0086] Configure the compensated fundamental frequency signal amplitude to the amplitude register of the arbitrary waveform generator through the GPIB bus to complete the initialization of the fundamental frequency amplitude; read the initial pulse amplitude from the stress test in the test strategy matrix and increase it according to the preset gradient rule. For example, after every 10 pulse cycles, the initial pulse amplitude increases by 10% (upper limit 0.5×A). The gradient parameter is monitored in real time by the FPGA counter and dynamically updated to the pulse amplitude register of the waveform generator; the initial pulse interval is set by the test strategy and is dynamically corrected according to the bit error rate feedback later. The adjusted pulse interval is written to the time series controller of the waveform generator through a high-precision timer; in the arbitrary waveform generator, the fundamental frequency sine wave is generated by the DDS (direct digital synthesis) module, and the pulse sequence is generated by the high-speed arbitrary sequence generator (ASG). Inject the test signal through the arbitrary waveform generator to generate bit error rate data and generate the composite reference signal. The expression is:

[0087]

[0088] where s(t) represents the composite reference signal at time t, f0 represents the fundamental frequency, which is related to the test protocol, and B qrepresents the amplitude of the q-th pulse, which is used to simulate the burst load pressure. q represents the pulse index, N represents the total number of pulses, t represents time, Δ represents the pulse time interval, which is used to control the density of the pressure test, and δ represents the Dirac function, which is used to simulate the burst load pressure.

[0089] Through the physical layer injection interface of the protocol analyzer, the composite reference signal is superimposed on the differential signal lines of the device under test, and at the same time, the reflection coefficient Γ is monitored to ensure impedance matching. The bit error rate data ∈ generated by injecting the test signal is monitored in real time BER (such as PCIe 4.0 requires ∈ BER <10 -12 ) and the frequency offset Δf of the test signal is adjusted through an online learning algorithm. When ∈ BER <0.1×10 -12 , Δf is increased to strengthen the pressure test. When ∈ BER ≥0.1×10 -12 , Δf is decreased to avoid device overload; the adjusted Δf is input into the clock generator to dynamically change the frequency modulation depth of the test signal.

[0090] Based on the temperature measurement value in the real-time environmental state vector, drift detection is performed, and a temperature threshold is set. When the absolute value of the difference between the temperature measurement value and the target temperature value is greater than the temperature threshold, voltage compensation is triggered, and the voltage is adjusted through a programmable power supply to offset the voltage change caused by temperature drift and maintain the stability of the device operating point; the performance indicators, abnormal events, and compensated environmental parameters during the dynamic adjustment according to the timestamp are packed into a structured test log.

[0091] S5. Extract the timing data from the structured test log to construct a single-timepoint feature vector; with the time window as the span, construct a timing matrix; use the LSTM autoencoder to learn the latent distribution of normal data through the encoding-decoding process, and calculate the Frobenius norm of the reconstruction error as the calculation method of the deviation degree to measure the reconstruction error between the current data and the learned normal distribution, reflecting the degree of abnormality. The expression is:

[0092]

[0093] where D represents the deviation degree, represents the reconstruction output matrix of the LSTM autoencoder for the timing matrix X, X represents the timing matrix, and ||·|| F represents the matrix Frobenius norm, which calculates the square root of the sum of the squares of all elements.

[0094] Calculate the mean and standard deviation of all deviation degrees within the time window [t - W, t]. The expression is:

[0095]

[0096] Among them, μ D represents the mean of all deviation degrees within the sliding window, W represents the sliding window size, r represents the index variable within the sliding window, and D r represents the deviation degree value calculated at time point r, and σ D represents the standard deviation of all deviation degrees within the sliding window, reflecting the data discreteness.

[0097] Based on the mean and standard deviation of all deviation degrees, the deviation degree threshold is dynamically adjusted, and the expression is:

[0098] θ = μ D + 3σ D ·(1 - e -Ω / γ );

[0099] Among them, θ represents the adjusted deviation degree threshold, Ω represents the total time from the start of the test to the current operation, γ represents the attenuation coefficient, controlling the relaxation speed of the deviation degree threshold over time; when D > θ, the abnormal type determination and time correlation analysis are performed according to the deviation contribution degrees of each dimension in the time series matrix to locate the starting time of the anomaly; combining the abnormal type determination result, the starting time of the anomaly, and the anomaly severity, an abnormal feature vector is generated.

[0100] S6. Based on historical fault data and domain knowledge, define the vertex set as hardware components (such as CPU, PCIe controller, power module, etc.), define the edge set as the fault propagation path (such as "unstable power voltage → CPU clock jitter → rising PCIe error rate"), and the edge weight matrix represents the conditional probability, that is, the probability of one fault causing another fault, which can be calculated by the existing Bayesian network or causal discovery algorithm; according to the abnormal type determination result and the starting time of the anomaly in the abnormal feature vector, activate the relevant subgraph to complete the construction of the fault propagation graph; apply the improved PageRank algorithm to iteratively calculate the root cause probability of each component, and the expression is:

[0101]

[0102] Among them, G(g α ) represents the root cause probability of the α-th hardware component node g α , g α represents the α-th hardware component node, represents the damping coefficient (usually set to 0.85), controlling the balance between random jump and propagation, g β represents the β-th hardware component node, M(g α ) represents the set of all neighbor nodes pointing to g α , L(g β ) represents the total number of out-edges of node g β (that is, the number of branches of the fault propagation path), G(gβ ) represents the root cause probability of the β-th hardware component node g β .

[0103] Sort the calculated root cause probabilities of each component in descending order, set a root cause threshold based on historical failure data, and select the vertices with root cause probabilities greater than the root cause threshold as the key root causes; for the key root cause components, match the best parameter adjustment strategy from the historical optimization solution library to generate optimization suggestions; adjust the environmental parameters according to the temperature sensitivity of the root cause components to obtain the updated environmental configuration; merge the optimization suggestions and the updated environmental configuration into the test strategy matrix, and define the priority rules to update the test strategy matrix.

[0104] This embodiment also provides an automatic testing system for information and communication technology (ICT) equipment, including: a requirement analysis module that parses the test requirements input by the user through a natural language interface, extracts key parameters, and retrieves historical cases to generate a structured test requirement table; a strategy generation module that constructs a multi-objective optimization model based on the structured test requirement table, uses the NSGA-III algorithm improved by chaotic mapping, and outputs the test strategy matrix after iterative update; an environmental control module that constructs a physical test environment through a programmable power supply and a temperature control device according to the test strategy matrix to generate a real-time environmental state vector; an adaptive testing module that combines the real-time environmental state vector and the test strategy matrix, starts a multi-dimensional test case execution engine, and dynamically adjusts test parameters through an online learning algorithm to form a structured test log; an anomaly diagnosis module that extracts features and detects anomalies from the test log, and identifies anomaly features through a dynamic threshold model; a strategy iteration module that constructs a fault propagation graph model based on the anomaly features, applies an improved causal reasoning algorithm to locate the root cause nodes, and generates a strategy iteration plan.

[0105] This embodiment also provides a computer device applicable to the automatic testing method for ICT equipment, 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 automatic testing method for ICT equipment as proposed in the above embodiment.

[0106] The computer device can be a terminal, which includes a processor, a memory, a communication interface, a display screen, and an input device connected via 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 computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be implemented through WIFI, 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 buttons, trackballs, or touchpads provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0107] 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 method for automatically testing a Xinchuang device as 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 (abbreviated as SRAM), electrically erasable programmable read-only memory (abbreviated as EEPROM), erasable programmable read-only memory (abbreviated as EPROM), programmable read-only memory (abbreviated as PROM), read-only memory (abbreviated as ROM), magnetic memory, flash memory, a magnetic disk, or an optical disc.

[0108] In summary, the present invention: calculates the candidate term probability by using the cosine similarity between the pre-trained vector and the query vector, and combines the semantic topological relationship of the knowledge graph to screen out the accurate terms higher than the threshold, reducing the term misjudgment rate compared with the traditional disambiguation method based on word frequency; the improved NSGA-III algorithm initializes the population through chaotic mapping and the dynamic objective space adaptation mechanism, improving the coverage rate of the Pareto front. At the same time, combined with the resource weight allocation formula of the aging-time decay factor, it realizes the co-optimization of the test cost and the extreme scenario coverage rate; the progressive approximation algorithm improves the convergence speed of the environmental parameters through the coarse-tuning and fine-tuning dual-stage control and the discretized first-order inertia correction; through the closed-loop control of the impedance matching verification model and the voltage standing wave ratio, it reduces the reflection coefficient of the link calibration; through the dynamic deviation threshold, it improves the accuracy of the abnormal type determination, and combines the root cause location model of the improved PageRank algorithm to improve the recognition accuracy of the fault propagation path.

[0109] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the 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 method for automatic testing of ICT equipment, characterized in that: include, Parse the test requirements input by users through the natural language interface, extract key parameters and retrieve historical cases to generate a structured test requirements table; Construct a multi-objective optimization model, use the NSGA-III algorithm improved by chaotic mapping, and output the test strategy matrix after iterative update; Build a physical test environment through programmable power supply and temperature control equipment to generate real-time environmental state vector; Combining the real-time environment state vector and the test strategy matrix, the test parameters are dynamically adjusted through the online learning algorithm to form a test log with environmental compensation; Perform feature extraction and anomaly detection on test logs, and identify abnormal features through dynamic threshold models; Construct a fault propagation graph model, apply an improved causal reasoning algorithm to locate the root cause node, and generate a policy iteration plan.

2. The automatic testing method for the ICT equipment according to claim 1, characterized in that: The extracting of key parameters includes semantically disambiguating the test requirements input by the user, generating candidate terms, and screening the candidate terms.

3. The automatic testing method for the ICT equipment according to claim 1, characterized in that: The multi-objective optimization model is constructed by defining, through a structured requirement test table, minimizing the weighted deviation between the test setting value and the target value, minimizing the total test cost, and maximizing the coverage of potential failure scenarios.

4. The automatic testing method for the ICT equipment according to claim 1, characterized in that: The NSGA-III algorithm improved by chaotic mapping and outputting the test strategy matrix after iterative update includes the following steps: Perform chaotic mapping on the target optimization model to generate an initial population and dynamically adjust it; Simulated binary crossover and polynomial mutation are used to perform multi-objective iterative optimization on the dynamically adjusted initialization population to generate a Pareto optimal solution set. Select the scenario with the highest scenario coverage in the Pareto optimal solution set as the test sequence; Dynamically adjust resource allocation weights based on the Pareto optimal solution set; The test strategy matrix is ​​obtained by combining the test sequence, the parameter combination in the Pareto optimal solution set, and the resource allocation weight.

5. The automatic testing method for the ICT equipment according to claim 1, characterized in that: The dynamically adjusting the test parameters through the online learning algorithm includes injecting bit error rate data into the test signal, adjusting the frequency offset of the test signal through the online learning algorithm, and dynamically changing the frequency adjustment depth of the test signal.

6. The automatic testing method for the ICT equipment according to claim 1, characterized in that: The method of identifying abnormal features by using a dynamic threshold model comprises the following steps: Use LSTM autoencoder to learn the latent distribution of normal data through encoding-decoding process and calculate the reconstruction error as deviation; Calculate the mean and standard deviation of all deviations; Dynamically adjust the deviation threshold based on the mean and standard deviation of all deviations; The calculated deviation is compared with the adjusted deviation threshold to identify abnormal features.

7. The automatic testing method for the ICT equipment according to claim 1, characterized in that: The generation strategy iteration scheme comprises the following steps: Construct fault propagation diagrams and calculate the root cause probability of each component; Sort the root cause probabilities of each component in descending order, select the key root causes, and generate optimization suggestions; According to the temperature sensitivity of the key root causes, the environmental parameters are adjusted to obtain the updated environmental configuration; Merge the optimization suggestions and updated environment configuration into the test strategy matrix, define priority rules, update the test strategy matrix, and generate a strategy iteration plan.

8. An automatic testing system for ICT equipment, based on the automatic testing method for ICT equipment according to any one of claims 1 to 7, characterized in that: include, The requirement analysis module analyzes the test requirements input by the user through the natural language interface, extracts key parameters and retrieves historical cases to generate a structured test requirement table; The strategy generation module builds a multi-objective optimization model based on the structured test requirement table, uses the NSGA-III algorithm improved by chaos mapping, and outputs the test strategy matrix after iterative updates; The environmental control module builds a physical test environment through programmable power supplies and temperature control devices according to the test strategy matrix and generates a real-time environmental state vector; The adaptive test module combines the real-time environment state vector and the test strategy matrix to start the multi-dimensional test case execution engine, dynamically adjusts the test parameters through the online learning algorithm, and forms a structured test log; The anomaly diagnosis module performs feature extraction and anomaly detection on the test logs and identifies abnormal features through a dynamic threshold model; The policy iteration module builds a fault propagation graph model based on abnormal features, applies an improved causal reasoning algorithm to locate the root cause node, and generates a policy iteration plan.

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, it implements the steps of the automatic testing method for information technology equipment described in any one of claims 1 to 7.

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 automatic testing method for information technology equipment described in any one of claims 1 to 7 are implemented.

Images