A model construction method for quantum computing industry quality big data

Abstract

The application discloses a model construction method for quantum computing industry quality big data, and relates to the technical field of industry product quality detection, and comprises the following steps: acquiring the dark count stability standard and historical dark count stability quality data of a single photon detector, and acquiring the equipment production operation data of a production equipment; obtaining standard quality data and standard operation data; constructing a single photon detector intelligent sampling inspection model based on the standard quality data and the standard operation data; constructing a sampling inspection correction strategy based on a product sampling inspection reference scheme, and sampling inspecting the single photon detector; the application is used to solve the problem that the existing industry product quality detection technology cannot locate single photon detector products with high risk of poor quality according to equipment parameters during production when detecting the dark count stability of the single photon detector products, and cannot accurately sample inspect the single photon detector products with a purpose, thereby reducing the detection cost and after-sales cost.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of industrial product quality detection, and particularly relates to a model construction method for quantum computing industrial quality big data. BACKGROUND

[0002] Industrial product quality detection technology refers to a technical system for testing, testing and evaluating the performance, safety, reliability, compliance and other quality characteristics of industrial products through a series of scientific methods, standards and processes. The core goal is to ensure that the product meets the design requirements, industry standards and user needs, and to provide data support for production process optimization, quality control and market access.

[0003] The existing industrial product quality detection technology often detects all the core parameters of the single photon detector product when detecting the single photon detector product, and the non-core parameters are often selected for full detection or sampling inspection according to the application scene; the dark count stability in the non-core parameters refers to the fluctuation amplitude of the dark count rate with time under constant working conditions, and the dark count stability is the premise of reliable operation of the system in the scene requiring continuous operation or long-time data accumulation; for full detection of the dark count stability, the dark count stability of each single photon detector needs to be measured, for the batch production scene, the detection cost increases linearly with the yield, the long-term high-load operation of the equipment in the full detection process is easy to wear, and the fatigue effect of manual detection may cause the misjudgment rate to rise, indirectly increasing the quality risk cost; although random sampling inspection can reduce the detection cost, only the samples are extracted according to the probability, and continuous defective products caused by equipment problems may be missed, and the defective products flow into the market, resulting in increased after-sales cost and defective loss, and the total cost cannot be reduced; therefore, when detecting the dark count stability of the single photon detector product, the existing industrial product quality detection technology cannot locate the single photon detector product with high defective risk according to the equipment parameters during production, and accurately sample for the purpose, so as to reduce the detection cost and the after-sales cost. SUMMARY

[0004] The present application aims to at least solve one of the technical problems in the prior art, by obtaining the dark count stability standard of a single photon detector and historical dark count stability quality data, and obtaining equipment production operation data of a production device; performing first data processing based on the historical dark count stability quality data and the equipment production operation data to obtain standard quality data and standard operation data; constructing a single photon detector intelligent sampling inspection model based on the standard quality data and the standard operation data, and obtaining a product sampling inspection reference scheme; constructing a sampling inspection correction strategy based on the product sampling inspection reference scheme, and sampling inspecting the single photon detector; to solve the problem that the existing industrial product quality detection technology cannot locate single photon detector products with high risk of poor quality according to equipment parameters during production when detecting the dark count stability of single photon detector products, and cannot perform purposeful and accurate sampling inspection, thereby reducing detection cost and after-sales cost.

[0005] To achieve the above-mentioned purpose, the present application provides a model construction method for quantum computing industrial quality big data, comprising the following steps:

[0006] Obtain the dark count stability standard of a single photon detector and historical dark count stability quality data, and obtain equipment production operation data of a production device;

[0007] Perform first data processing based on the historical dark count stability quality data and the equipment production operation data to obtain standard quality data and standard operation data;

[0008] Construct a single photon detector intelligent sampling inspection model based on the standard quality data and the standard operation data, and obtain a product sampling inspection reference scheme;

[0009] Construct a sampling inspection correction strategy based on the product sampling inspection reference scheme, and sample inspect the single photon detector.

[0010] Further, obtaining the dark count stability standard of a single photon detector and historical dark count stability quality data, and obtaining equipment production operation data of a production device comprises the following sub-steps:

[0011] Obtain the normal standard dark count stability of a single photon detector, denoted as the dark count stability standard of the single photon detector;

[0012] Obtain the number of single photon detectors produced in each batch, denoted as L0; for any one production batch, denoted as the first batch, for all single photon detectors produced in the first batch, sort and number them according to the production order of the single photon detectors and the corresponding batch, obtain the dark count stability of each single photon detector and the production time, and obtain the average time T0 for producing a single photon detector, and complete the dark count stability quality data of the first batch;

[0013] The dark count stability quality data of multiple batches is repeatedly acquired, and each batch is sequentially numbered according to the production order, and is respectively recorded as batch 1-n0, and the historical dark count stability quality data is obtained after completion.

[0014] Further, the dark count stability standard of the single photon detector and the historical dark count stability quality data are acquired, and the device production operation data of the production equipment further includes the following sub-steps:

[0015] For any single photon detector in the first batch, it is recorded as the first single photon detector; the first time interval is set as t1; the input voltage of the production equipment when producing the first single photon detector is collected at the first time interval, and is recorded as the production voltage information of the first single photon detector;

[0016] The production voltage information of all single photon detectors in the first batch is repeatedly acquired, and is recorded as the production voltage data of the first batch, and the production voltage data of batches 1-n0 is repeatedly acquired, and the device production operation data is obtained.

[0017] Further, the first data processing is performed based on the historical dark count stability quality data and the device production operation data, and the standard quality data and the standard operation data are obtained, including the following sub-steps:

[0018] All input voltages in the device production operation data are normalized by the max-min normalization method, and the sizes of all input voltages are mapped to [0, 1], and the normalized device operation data is obtained after completion;

[0019] The production voltage information of the first single photon detector in the first batch in the normalized device operation data is recorded as the first voltage information, and the first voltage information is sorted from far to near according to the time sequence of collection, and is recorded as the first production voltage sequence;

[0020] For each voltage value in the first production voltage sequence, the first probability amplitude corresponding to each voltage value is calculated by the first formula, and is arranged according to the corresponding order, and the first probability amplitude sequence is obtained; the first formula is as follows: Wherein, represents the first probability amplitude of the i-th voltage value, and represents the i-th voltage;

[0021] The first entanglement matrix CE is constructed based on the first probability amplitude sequence and the second formula, and the second formula is as follows: Wherein, represents the element of the i-th row and the j-th column of the first entanglement matrix, and represents the first probability amplitude of the j-th voltage value;

[0022] Absolute values of all eigenvalues of the first entanglement matrix are obtained, normalized to obtain a normalized eigenvalue sequence, and the first entanglement entropy of the first single-photon detector is calculated according to a third formula as follows: where λm represents the mth eigenvalue in the normalized eigenvalue sequence, QCE represents the first entanglement entropy, and M represents the total number of eigenvalues in the normalized eigenvalue sequence.

[0023] Further, the first data processing is performed based on the historical dark count stability quality data and the equipment production operation data to obtain standard quality data and standard operation data, including the following sub-steps:

[0024] The mean and the standard deviation of the first production voltage sequence are calculated and are sequentially denoted as AP and AB, respectively, and the kurtosis corresponding to the first single-photon detector is calculated according to a fourth formula as follows: where QD is the kurtosis, and n1 is the total number of voltage values in the first production voltage sequence.

[0025] According to the equipment production operation data, the kurtosis corresponding to all single-photon detectors is repeatedly obtained, and normalized according to a max-min normalization method to map the sizes of all kurtosis to [0, 1];

[0026] According to the normalized kurtosis corresponding to the first single-photon detector, a quantum feature vector LR corresponding to the first single-photon detector is generated, where LR = [cos(GD*π / 2), sin(GD*π / 2)], and GD is the normalized QD.

[0027] The quantum feature vectors corresponding to all single-photon detectors are repeatedly obtained and denoted as feature vector data, the average vector of the quantum feature vectors corresponding to the single-photon detectors in the first batch in the feature vector data is obtained and denoted as a first average vector DR1, and the average vectors of the remaining quantum feature vectors are obtained and denoted as a second average vector DR2; and a weighted average vector DRO of DR1 and DR2 is obtained and denoted as the average feature vector of the first batch, where DRO = q1*DR1 + q2*DR2, and q1 and q2 are set weight coefficients.

[0028] The second voltage kurtosis corresponding to the first single-photon detector is calculated according to a fifth formula as follows: where VFR is the second voltage kurtosis.

[0029] According to the production voltage data of the first batch, the first entanglement entropy and the second voltage kurtosis of all single-photon detectors in the first batch are repeatedly obtained, and the first operation data of the first batch is obtained after the operation.

[0030] The first operation data of all batches is repeatedly obtained to obtain the standard operation data.

[0031] Further, the first data processing is performed based on the historical dark count stability quality data and the device production operation data to obtain standard quality data and standard operation data, including the following sub-steps:

[0032] Based on the dark count stability standard of the single photon detector, it is judged whether the dark count stability of each single photon detector in the historical dark count stability quality data meets the dark count stability standard. If it meets, it is marked as normal dark count stability, and if it does not meet, it is marked as abnormal dark count stability.

[0033] For any one sequence number position in the batch, denoted as the first sequence number position, the dark count stability of the first sequence number position in all batches of the historical dark count stability quality data is obtained, and the probability of occurrence of abnormal dark count stability is calculated, denoted as the historical abnormal rate BYi of the first sequence number position. The historical abnormal rates of all sequence number positions in the batch are repeatedly obtained, and after completion, the first quality data is obtained.

[0034] The detection cost of each single photon detector is obtained, denoted as CB0; and the average loss cost caused by the missed detection of abnormal dark count stability is obtained, denoted as CB1;

[0035] The total number of abnormal dark count stability of the first batch is obtained, denoted as YW, and YW*CB0 is denoted as the minimum detection loss cost of the first batch. The minimum detection loss cost of each batch is repeatedly obtained, and is stored together with the first quality data, denoted as the standard quality data.

[0036] Further, the single photon detector intelligent sampling model is constructed based on the standard quality data and the standard operation data, and the product sampling reference scheme is obtained, including the following sub-steps:

[0037] CU is denoted as the total detection loss cost, wherein CU=CB0*V0+CB1*V1, V0 is the number of single photon detectors sampled in a batch, and V1 is the number of missed detection of abnormal dark count stability in a batch;

[0038] Based on the quantum annealing algorithm, an original sampling model is constructed, and the minimization of CU is set as the core target of the model. The quantum bit is defined as σi∈{0,1}, wherein σi=0 represents not detecting the i-th single photon detector in the batch, and σi=1 represents detecting the i-th single photon detector in the batch. A Hamiltonian is constructed, and the Hamiltonian is as follows: wherein hi=q3*QCEi+q4*VFRi+q5*BYi, q3, q4 and q5 are weight coefficients; Jij is a coupling parameter, Jij=q6*exp(|ti-tj| / T0), q6 is a coupling strength coefficient, and ti and tj are the production times of the i-th single photon detector and the j-th single photon detector, respectively;

[0039] Set the minimum sampling number as Fmin, calibrate and optimize the parameters of the model based on the standard quality data and the standard operation data; obtain the detection accuracy of the model for each batch, and record the minimum detection accuracy FM and the average detection accuracy FD, to ensure that the average detection accuracy of the sampling scheme output by the model is greater than k1, and obtain the single photon detector intelligent sampling model after completion, wherein k1 is a set missed detection rate threshold.

[0040] Further, based on the standard quality data and the standard operation data, the single photon detector intelligent sampling model is constructed, and the product sampling reference scheme is obtained, which further includes the following sub-steps:

[0041] For any one production batch of single photon detectors, denoted as the second batch, the equipment production operation data corresponding to the second batch and the production time of each single photon detector are collected, and the equipment production operation data is subjected to first data processing to obtain the standard operation data corresponding to the second batch. Then, the corresponding data is input into the single photon detector intelligent sampling model to obtain {σ1, σ2, …, σG}, denoted as the product sampling reference scheme.

[0042] Further, based on the product sampling reference scheme and the historical quality data, a sampling correction strategy is constructed, and the single photon detectors are subjected to sampling, including the following sub-steps:

[0043] Set the defective rate qualified threshold as BK0; based on the product sampling reference scheme, the single photon detectors of the second batch are subjected to first sampling, to obtain the total number of single photon detectors subjected to first sampling, denoted as YU0, and the number of single photon detectors with abnormal dark count stability subjected to first sampling, denoted as YU1; and the single photon detectors with abnormal dark count stability subjected to first sampling are marked as first defective dark count stability.

[0044] Calculate YU3 = YU1 / YU0, if YU3 is less than 2FM-FD, it is determined that the sampling fails; the first defective dark count stability of the second batch is removed, and the single photon detectors of the second batch are subjected to random sampling, and the defective rate BK1 of the single photon detectors subjected to random sampling is calculated; if BK1 is not greater than BK0, it is determined that the second batch is qualified; if BK1 is greater than BK0, it is determined that the second batch is unqualified, and the remaining single photon detectors of the second batch are subjected to full detection.

[0045] Further, based on the product sampling reference scheme and the historical quality data, a sampling correction strategy is constructed, and the single photon detectors are subjected to sampling, including the following sub-steps:

[0046] If YU3 is not less than 2FM-FD, it is determined that the sampling inspection is successful; BK2 is calculated, BK2=YU1 / G; BK3 is obtained by correcting BK2 based on FD, BK3=BK2 / FD, if BK3 is not greater than BK0, it is determined that the second batch is qualified; if BK1 is greater than BK0, the first poor dark count stability of the second batch is removed, then the single photon detector of the second batch is randomly sampled, and the poor rate BK1 is calculated; if BK1 is not greater than BK0, it is determined that the second batch is qualified; if BK1 is greater than BK0, it is determined that the second batch is unqualified, and the single photon detector of the remaining second batch is fully inspected.

[0047] The present application has the following advantages: the dark count stability standard of the single photon detector and the historical dark count stability quality data are obtained, and the equipment production operation data of the production equipment are obtained; the first data processing is performed based on the historical dark count stability quality data and the equipment production operation data, to obtain standard quality data and standard operation data; the single photon detector intelligent sampling inspection model is constructed based on the standard quality data and the standard operation data, and the product sampling inspection reference scheme is obtained; the sampling inspection correction strategy is constructed based on the product sampling inspection reference scheme, and the single photon detector is sampled; when the dark count stability of the single photon detector product is detected, the single photon detector product with high risk can be located according to the equipment parameters during production, and accurate sampling inspection with a purpose is performed, so that the detection cost and the after-sales cost are reduced.

[0048] The present application captures the chaotic correlation and extreme fluctuation characteristics of voltage fluctuation through quantum characteristic parameters such as probability amplitude sequence, entanglement matrix and entanglement entropy, and can reflect the microscopic quality risk in dark count stability production more than traditional statistical characteristics, improve the accuracy of sampling inspection of abnormal single photon detectors; the coupling strength is calculated through the production time difference, the time dependence and spatial correlation in the production process of the single photon detector are excavated, and the pertinence of the sampling inspection strategy is improved; the success or failure of sampling inspection is determined through the sampling inspection correction strategy, the sampling inspection scheme is dynamically adjusted, the real-time sampling inspection result is fed back, the decision logic is optimized, and the detection accuracy is improved. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 The step flowchart of the method of the present application is shown in the figure;

[0050] Figure 2 The second voltage kurtosis calculation flowchart of the present application is shown in the figure;

[0051] Figure 3 The sampling inspection correction strategy flowchart of the present application is shown in the figure;

[0052] Figure 4 The structure schematic diagram of the electronic equipment of the present application is shown in the figure. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0054] Embodiment 1, please refer to Figure 1 As shown in the figure, the present application provides a model construction method for quantum computing industry quality big data, comprising the following steps:

[0055] Step S1, acquiring the dark count stability standard of the single photon detector and the historical dark count stability quality data, and acquiring the equipment production operation data of the production equipment; step S1 comprises the following sub-steps:

[0056] Step S101, acquiring the normal standard dark count stability of the single photon detector, denoted as the dark count stability standard of the single photon detector; providing a unified standard for the eligibility of the dark count stability of the single photon detector, which is the core basis for subsequent marking of "normal dark count stability" and "abnormal dark count stability";

[0057] Step S102, acquiring the number of single photon detectors in each batch, denoted as L0; for any one production batch, denoted as the first batch, for all single photon detectors produced in the first batch, sorting and numbering according to the production order of the single photon detector and the corresponding batch, acquiring the dark count stability of each single photon detector and the production time, and acquiring the average time T0 for producing a single photon detector, after which the dark count stability quality data of the first batch is obtained; the average production time T0 is used to measure the production line stability, and at the same time serves as a normalization benchmark for time distance in the subsequent process, for quantifying the time correlation of adjacent produced single photon detectors;

[0058] Step S103, repeatedly acquiring the dark count stability quality data of multiple batches, and sorting and numbering each batch according to the production order, respectively denoted as batches 1-n0, after which the historical dark count stability quality data is obtained;

[0059] Step S104, for any single-photon detector in the first batch, denoted as the first single-photon detector, setting the first time interval as t1, collecting the input voltage of the production equipment when producing the first single-photon detector with the first time interval, denoted as the production voltage information of the first single-photon detector; that is, each single-photon detector has a corresponding voltage sequence during production; the production equipment when producing the first single-photon detector is various, and the dark count stability of the single-photon detector depends on the uniformity of pn junction doping, and the key equipment of the doping process is an ion implanter; the stability in the input voltage has the greatest influence on the dark count stability of the single-photon detector, so when collecting the input voltage of the production equipment, the input voltage of the ion implanter is mainly collected;

[0060] Step S105, repeatedly collecting the production voltage information of all single-photon detectors in the first batch, denoted as the production voltage data of the first batch, repeatedly collecting the production voltage data of batches 1-n0, and obtaining the equipment production operation data;

[0061] In the specific implementation process, the core of the single-photon detector is the pn junction avalanche multiplication structure under reverse bias, and the dark count is mainly from the "tunneling current" and "hot excited carriers" in the depletion region of the pn junction. The most critical step of the dark count stability is the pn junction doping. When the impurity concentration is not uniform during ion implantation or diffusion doping, the electric field intensity in the depletion region will fluctuate, random tunneling current will be easily generated in some areas, and the fluctuation with temperature change will be greater, thereby causing the dark count stability to be too large; and if the voltage fluctuation of the ion implanter suddenly increases when producing a single-photon detector, it may indicate that the equipment is malfunctioning, and the detection weight of the single-photon detector needs to be increased in the sampling inspection; the production voltage information is also the core input of the subsequent quantum characteristics.

[0062] Step S2, performing first data processing based on the historical dark count stability quality data and the equipment production operation data to obtain standard quality data and standard operation data; step S2 includes the following substeps:

[0063] Step S201, normalizing all input voltages in the equipment production operation data by the max-min normalization method, mapping the size of all input voltages to [0, 1], and obtaining the normalized equipment operation data after completion; eliminating the dimension influence of the original value of the voltage, so that the voltage data of different batches is comparable; at the same time, the probability amplitude, phase angle and other parameters in quantum calculation are usually defined in the interval [0, 1] or [-1, 1], and normalization ensures that the voltage data meets the mathematical requirements of quantum state mapping;

[0064] Step S202, record the production voltage information of the first batch of single photon detectors in the normalized device running data as first voltage information, sort the first voltage information according to the time sequence from far to near, and record as a first production voltage sequence;

[0065] Step S203, for each voltage value in the first production voltage sequence, calculate the first probability amplitude corresponding to each voltage value by a first formula, and arrange according to the corresponding order to obtain a first probability amplitude sequence; the first formula is as follows: Wherein, represents the first probability amplitude of the i-th voltage value, and represents the i-th voltage; the classical voltage data is converted into the probability amplitude representation in quantum mechanics, so that the voltage fluctuation has the superposition characteristics of the quantum state; the fluctuation amplitude of the probability amplitude directly reflects the voltage stability - if the voltage fluctuation is large, the change of the sequence is more violent, and the subsequent entanglement entropy calculation will capture higher chaotic characteristics, prompting the production quality risk;

[0066] Step S204, based on the first probability amplitude sequence and a second formula, a first entanglement matrix CE is constructed, and the second formula is as follows: Wherein, represents the element of the i-th row and the j-th column of the first entanglement matrix, and represents the first probability amplitude of the j-th voltage value;

[0067] Step S205, obtain the absolute value of all eigenvalues of the first entanglement matrix, and perform normalization processing to obtain a normalized eigenvalue sequence, and calculate the first entanglement entropy of the first single photon detector according to a third formula, and the third formula is as follows: Wherein, represents the m-th eigenvalue of the normalized eigenvalue sequence, QCE represents the first entanglement entropy, and M represents the total number of eigenvalues in the normalized eigenvalue sequence; the entanglement matrix element CEij reflects the correlation of voltage probability amplitudes at different times, and the entanglement entropy QCE quantifies the chaotic degree of the entire voltage sequence through the eigenvalue distribution; the higher QCE is, the stronger the disorder of voltage fluctuation is, the worse the uniformity in the production process of dark count stability is, and the higher the risk of quality problems is;

[0068] Step S206, please refer to Figure 2 As shown in the figure, the mean and the standard deviation of the first production voltage sequence are calculated, and are recorded as AP and AB in order respectively, and the kurtosis corresponding to the first single photon detector is calculated according to a fourth formula, and the fourth formula is as follows: Wherein, QD is the kurtosis, and n1 is the total number of voltage values in the first production voltage sequence; the kurtosis QD reflects the "spike thick tail" characteristics of the voltage distribution; QD>0 indicates that there is voltage peak or valley, such as voltage impact when the device starts and stops, which may cause dark count stability defects; QD<0 indicates that the fluctuation is gentle, and the production state is stable;

[0069] Step S207, according to the equipment production operation data, repeatedly obtain the kurtosis corresponding to all single photon detectors, and perform normalization processing according to the max-min normalization method, and map the size of all kurtosis to [0, 1]; the normalized kurtosis is used for subsequent generation of quantum feature vectors LR;

[0070] Step S208, generate the quantum feature vector LR corresponding to the first single photon detector according to the normalized kurtosis corresponding to the first single photon detector, wherein LR=[cos(GD*π / 2), sin(GD*π / 2)], GD is the normalized QD; LR converts the kurtosis risk into a two-dimensional quantum state vector, and the vector direction reflects the risk degree, for example, GD=1 corresponds to a phase angle of 90°, LR=[0, 1], indicating the highest risk, which is convenient for subsequent quantum model to process multi-dimensional features through vector inner product and entanglement operation;

[0071] Step S209, repeatedly obtain the quantum feature vectors corresponding to all single photon detectors, denoted as feature vector data, obtain the average vector of the quantum feature vectors corresponding to the first batch of single photon detectors in the feature vector data, denoted as the first average vector DR1, and obtain the average vector of the remaining quantum feature vectors, denoted as the second average vector DR2; then, the weighted average vector DR0 of DR1 and DR2 is calculated, denoted as the average feature vector of the first batch, wherein DRO=q1*DR1+q2*DR2, q1 and q2 are set weight coefficients; in this embodiment, q1=q2=0.5; the quality features within and across batches are integrated through weighted average, and the consistency of the production process is captured;

[0072] Step S210, calculate the second voltage kurtosis corresponding to the first single photon detector according to the fifth formula, and the fifth formula is as follows: , wherein VFR is the second voltage kurtosis; the essence of VFR is a cooperative evaluation index of individual voltage fluctuation risk and batch global quality feature; the larger the VFR is, the more significant the voltage exists extreme fluctuations such as sharp peaks and sharp drops during the production of the single photon detector, and the greater the difference between the quantum feature vector LR and the batch average vector DR0, which means that the production state of the dark count stability deviates from the overall level of the batch, and the probability of the dark count stability becoming a defective product significantly increases, which needs to be preferentially included in the range of sampling inspection;

[0073] Step S211, according to the production voltage data of the first batch, repeatedly obtain the first entanglement entropy and the second voltage kurtosis of all single photon detectors in the first batch, and complete the first running data of the first batch;

[0074] Step S212, repeatedly obtain the first running data of all batches to obtain the standard running data;

[0075] Step S213, based on the dark count stability standard of the single photon detector, judge whether the dark count stability of each single photon detector in the historical dark count stability quality data meets the dark count stability standard, if it meets, mark it as normal dark count stability, if it does not meet, mark it as abnormal dark count stability;

[0076] Step S214, for any one sequence number position in the batch, mark it as the first sequence number position, obtain the dark count stability of the first sequence number position in all batches of the historical dark count stability quality data, and calculate the probability of occurrence of abnormal dark count stability, mark it as the historical abnormal rate BYi of the first sequence number position, repeat the acquisition of the historical abnormal rate of all sequence number positions in the batch, and complete the first quality data;

[0077] Step S215, obtain the detection cost of each single photon detector, mark it as CB0; and obtain the average loss cost caused by the missed detection of abnormal dark count stability, mark it as CB1; the average loss cost caused by the missed detection of abnormal dark count stability will be much larger than the detection cost of the single photon detector;

[0078] Step S216, obtain the total number of abnormal dark count stability of the first batch, mark it as YW, and YW*CB0 is marked as the minimum detection loss cost of the first batch, repeat the acquisition of the minimum detection loss cost of each batch, and store it with the first quality data, mark it as standard quality data; the minimum detection loss cost refers to the total cost of the detection cost and the loss cost caused by the missed detection of abnormal dark count stability; for example, the detection cost CB0 of each single photon detector is 0.5 yuan, the average loss cost CB1 caused by the missed detection of abnormal dark count stability is 20 yuan, the total number of abnormal dark count stability of the first batch YW is 10, and the minimum detection loss cost is 0.5*10=5 yuan, that is, the cost when no abnormal dark count stability is missed;

[0079] In the specific implementation process, the entanglement entropy, as the core parameter of quantum chaos theory, can capture the nonlinear correlation ignored by classical statistical methods, for example, the chaotic oscillation of voltage caused by device aging, and provides a microscopic quality risk indicator for the quantum model; and the VFR combines the kurtosis risk of individual dark count stability and the overall production state, calculates the feature similarity through vector inner product, for example, the smaller the angle between LR and DR0, the more the risk features of the dark count stability conform to the overall risk of the current batch, avoiding single feature misjudgment.

[0080] Step S3, based on the standard quality data and the standard operation data, construct a single photon detector intelligent sampling inspection model, and obtain a product sampling inspection reference scheme; step S3 includes the following substeps:

[0081] Step S301, record the CU as the total cost of detection loss, wherein CU=CB0*V0+CB1*V1, V0 is the number of single-photon detectors in a batch, and V1 is the number of missed detection of abnormal dark count stability in a batch; the detection cost and the missed detection loss are converted into a unified cost function, so that the optimization direction of the quantum model is directly aligned with the minimization of the total cost;

[0082] Step S302, based on the quantum annealing algorithm, an original sampling model is constructed, and the minimization of CU is set as the core target of the model; the quantum bit is defined as σi∈{0, 1}, wherein σi=0 represents that the i-th single-photon detector in the batch is not detected, and σi=1 represents that the i-th single-photon detector in the batch is detected; and a Hamiltonian is constructed, and the Hamiltonian is as follows: wherein hi=q3*QCEi+q4*VFRi+q5*BYi, q3, q4 and q5 are weight coefficients; Jij is a coupling parameter, Jij=q6*exp(|ti-tj| / T0), q6 is a coupling strength coefficient, ti and tj are production times of the i-th single-photon detector and the j-th single-photon detector respectively; the quantum tunneling effect of the quantum annealing algorithm is used to quickly search for the global optimal solution of the minimization of CU, and the convergence speed is faster than that of the classical optimization algorithm when processing large-scale sampling decisions;

[0083] The hi parameter fuses the quantum characteristics (QCEi, VFRi) and the historical abnormal rate (BYi), and quantifies the detection priority of a single dark count stability-QCEi high represents a risk of voltage chaos, VFRi high represents a voltage extreme fluctuation, and BYi high represents a high probability of a defect at the same position, after being weighted by the weight coefficient, hi corresponding to a high-risk dark count stability is larger, and the quantum bit σi is more inclined to take 1, that is, sampling is required;

[0084] The Jij parameter quantifies the time correlation of the production of the dark count stability through the production time difference |ti-tj|, for example, the dark count stability produced within 10 minutes is likely to be affected by the fluctuation of the same device, Jij>0 makes the detection decisions of adjacent dark count stabilities have a positive correlation, for example, simultaneous detection or non-detection, and improves the rationality of the sampling scheme;

[0085] Step S303, set the minimum sampling number as Fmin, calibrate and optimize the parameters of the model based on the standard quality data and the standard operation data; and obtain the detection accuracy of the model for each batch, and record the minimum detection accuracy FM and the average detection accuracy FD, so as to ensure that the average detection accuracy of the sampling scheme output by the model is greater than k1, and the intelligent sampling model of the single-photon detector is obtained, wherein k1 is a set missed detection rate threshold;

[0086] The Fmin avoidance model avoids excessive reduction of the sampling quantity for cost saving, prevents the false negative rate from rising sharply due to insufficient sample size, and is the bottom line guarantee of the cost-risk balance; the optimal weight coefficient is determined through historical data training, so that the minimum value of the Hamiltonian energy corresponds to the lowest point of the actual cost; the detection accuracy is the ratio of the number of abnormal dark count stability of the single photon detector in the sampling to the total sampling number; in this embodiment, k1=90%; the detection accuracy is verified to ensure the reliability of the model in actual application;

[0087] In step S304, for any single photon detector of a production batch, denoted as a second batch, the device production operation data corresponding to the second batch and the production time of each single photon detector are collected, the device production operation data is subjected to first data processing to obtain standard operation data corresponding to the second batch, and then the corresponding data is input into the single photon detector intelligent sampling model to obtain {σ1, σ2, …, σG}, denoted as a product sampling reference scheme;

[0088] In the specific implementation process, the value of σ in {σ1, σ2, …, σG} represents whether the corresponding single photon detector needs to be sampled, for example, σ2=1 represents that the second single photon detector in the batch needs to be sampled, and σ5=0 represents that the fifth single photon detector in the batch does not need to be sampled.

[0089] In step S4, a sampling correction strategy is constructed based on the product sampling reference scheme, and the single photon detector is sampled; step S4 includes the following sub-steps:

[0090] In step S401, the unqualified threshold of the defective rate is set as BK0; the single photon detector of the second batch is subjected to first sampling based on the product sampling reference scheme, the total number of the single photon detectors subjected to the first sampling is obtained and denoted as YU0, and the number of the single photon detectors with abnormal dark count stability subjected to the first sampling is obtained and denoted as YU1; the single photon detector with abnormal dark count stability subjected to the first sampling is marked as the first defective dark count stability; BK0 can be set according to the actual application scene; in this embodiment, BK0=2%;

[0091] In step S402, please refer to Figure 3As shown, YU3=YU1 / YU0 is calculated, if YU3 is less than 2FM-FD, it is determined that the sampling inspection fails; and the first bad dark count stability of the second batch is removed, and then the single photon detector of the second batch is randomly sampled, and the bad rate BK1 of the randomly sampled single photon detector is calculated; if BK1 is not greater than BK0, it is determined that the second batch is qualified; if BK1 is greater than BK0, it is determined that the second batch is unqualified, and the remaining single photon detector of the second batch is fully inspected; a dynamic threshold 2FM-FD is set according to the historical performance index of the model, instead of a fixed ratio, to adapt to the performance fluctuation of the model under different batches; for example, if the average accuracy of the model FD=90%, the minimum accuracy FM=85%, the threshold is 2*85%-90%=80%, when the actual sampling bad rate YU3<80%, it means that the model has a problem, and the batch risk is underestimated, so the product sampling reference scheme output by the model is not completely reliable, and the bad rate of the batch is determined by the random sampling scheme to avoid missing detection caused by model parameter drift, and to cause non-standard batches to flow into the market;

[0092] In step S403, if YU3 is not less than 2FM-FD, it is determined that the sampling inspection is successful; that is, the sampling scheme of the model has no problem, BK2 is calculated, BK2=YU1 / G; and BK3 is obtained by correcting BK2 based on FD, BK3=BK2 / FD, if BK3 is not greater than BK0, it is determined that the second batch is qualified; if BK1 is greater than BK0, the first bad dark count stability of the second batch is removed, and then the single photon detector of the second batch is randomly sampled, and the bad rate BK1 is calculated; if BK1 is not greater than BK0, it is determined that the second batch is qualified; if BK1 is greater than BK0, it is determined that the second batch is unqualified, and the remaining single photon detector of the second batch is fully inspected; when part of the single photon detector is sampled and part of the abnormal dark count stability is removed, the bad rate of the batch still cannot meet the standard, then the risk is forced to control through full inspection to avoid unqualified products flowing into the market, causing too large loss cost of missing detection;

[0093] In the specific implementation process, the measured bad rate BK2 is corrected by using the historical average accuracy FD of the model, if the model FD=90%, BK3=BK2 / 0.9 is equivalent to amplifying the measured result, which makes up for the tendency of "optimistic estimation" of the model, for example, the model considers that the sampling scheme can sample all abnormal dark count stabilities, but there may be sampling bias in reality.

[0094] Embodiment 3, please refer to Figure 4 As shown, Figure 4An example is shown in a structural diagram of an electronic device, which can include a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete communication with each other through the communication bus. The memory stores computer readable instructions, and the processor can call the instructions in the memory. When the computer readable instructions are executed by the processor, the steps in a model construction method for quantum computing industry quality big data are run to implement the following functions: obtaining dark count stability standards of a single photon detector and historical dark count stability quality data, and obtaining equipment production operation data of a production device; performing first data processing based on the historical dark count stability quality data and the equipment production operation data to obtain standard quality data and standard operation data; constructing an intelligent sampling inspection model for the single photon detector based on the standard quality data and the standard operation data, and obtaining a product sampling inspection reference scheme; constructing a sampling inspection correction strategy based on the product sampling inspection reference scheme, and sampling inspecting the single photon detector.

[0095] In addition, the logical instructions in the memory described above can be implemented in the form of a software functional unit and sold or used as an independent product, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0096] Embodiment 4, the present application also provides a computer readable storage medium, and the present application provides a storage medium, which stores a computer program. When the computer program is executed by a processor, the steps in the above model construction method for quantum computing industry quality big data are run to implement the following functions: obtaining dark count stability standards of a single photon detector and historical dark count stability quality data, and obtaining equipment production operation data of a production device; performing first data processing based on the historical dark count stability quality data and the equipment production operation data to obtain standard quality data and standard operation data; constructing an intelligent sampling inspection model for the single photon detector based on the standard quality data and the standard operation data, and obtaining a product sampling inspection reference scheme; constructing a sampling inspection correction strategy based on the product sampling inspection reference scheme, and sampling inspecting the single photon detector.

[0097] Through the description of the above embodiments, the embodiments of the present application can be provided as a method, a system or a computer program product. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in various embodiments or some parts of the embodiments.

[0098] In the embodiments provided by the present application, it should be understood that the disclosed system or method can be implemented in other manners. The embodiments described above are merely schematic, and should not be construed as limiting. For example, the division of the modules or the units is merely logical function division, and there can be other division manners in actual implementation. For example, a plurality of modules or units can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different modules can be indirect couplings or communication connections through some interfaces, and there can be electric, mechanical or other forms.

[0099] Finally, it should be noted that the above-mentioned embodiments are merely used to illustrate the technical solutions of the present application, rather than limit them; even if the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features; and these modifications or replacements do not cause the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for constructing a model using big data on industrial quality through quantum computing, characterized in that, Includes the following steps: Obtain the dark count stability standard and historical dark count stability quality data of the single-photon detector, and obtain the equipment production and operation data of the production equipment; The first data processing is performed based on historical dark count stability quality data and equipment production operation data to obtain standard quality data and standard operation data. A smart sampling inspection model for single-photon detectors is constructed based on standard quality data and standard operating data, and a product sampling inspection reference scheme is obtained. A sampling inspection correction strategy was constructed based on the product sampling inspection reference scheme, and sampling inspections were carried out on single-photon detectors. Obtaining the dark count stability standard and historical dark count stability quality data of the single-photon detector, and acquiring the equipment production operation data of the production equipment, also includes the following sub-steps: For any single-photon detector in the first batch, it is denoted as the first single-photon detector; the first time interval is set to t1; the input voltage of the production equipment when producing the first single-photon detector is collected at the first time interval, and it is denoted as the production voltage information of the first single-photon detector; Repeatedly acquire the production voltage information of all single-photon detectors in the first batch, and record it as the production voltage data of the first batch. Repeatedly acquire the production voltage data of batches 1-n0 to obtain the equipment production operation data. The first data processing, based on historical dark count stability quality data and equipment production operation data, yields standard quality data and standard operation data, including the following sub-steps: All input voltages in the equipment production and operation data are normalized using the max-min normalization method, and the magnitudes of all input voltages are mapped to [0, 1]. After completion, the normalized equipment operation data is obtained. The production voltage information of the first single-photon detector in the first batch of the unified equipment operation data is recorded as the first voltage information. The first voltage information is sorted from the farthest to the nearest according to the acquisition time order and recorded as the first production voltage sequence. For each voltage value in the first production voltage sequence, the first probability amplitude corresponding to each voltage value is calculated using the first formula, and then arranged in the corresponding order to obtain the first probability amplitude sequence; the first formula is as follows: , where Ψi represents the first probability amplitude of the i-th voltage value, and xi represents the i-th voltage; The first entanglement matrix CE is constructed based on the first probability amplitude sequence and the second formula, which is as follows: , where CEij represents the element in the i-th row and j-th column of the first entanglement matrix, and Ψj represents the first probability amplitude of the j-th voltage value; Obtain the absolute values ​​of all eigenvalues ​​of the first entanglement matrix and normalize them to obtain a normalized eigenvalue sequence. Then, calculate the first entanglement entropy of the first single-photon detector according to the third formula, which is as follows: , where λm represents the m-th eigenvalue of the normalized eigenvalue sequence, QCE represents the first entanglement entropy, and M represents the total number of eigenvalues ​​in the normalized eigenvalue sequence; Calculate the mean and standard deviation of the first production voltage sequence, denoted as AP and AB respectively, and calculate the kurtosis corresponding to the first single-photon detector according to the fourth formula, which is as follows: Where QD is the kurtosis and n1 is the total number of voltage values ​​in the first production voltage sequence; Based on the equipment production and operation data, the kurtosis corresponding to all single-photon detectors is repeatedly obtained and normalized according to the max-min normalization method, so that the magnitude of all kurtosis is mapped to [0, 1]. Based on the normalized kurtosis corresponding to the first single-photon detector, the quantum feature vector LR corresponding to the first single-photon detector is generated, where LR=[cos(GD*π / 2),sin(GD*π / 2)], and GD is the normalized QD; Repeatedly acquire the quantum eigenvectors corresponding to all single-photon detectors, denoted as eigenvector data. Obtain the average vector of the quantum eigenvectors corresponding to the first batch of single-photon detectors in the eigenvector data, denoted as the first average vector DR1. Calculate the average vector of the remaining quantum eigenvectors, denoted as the second average vector DR2. Then calculate the weighted average vector DR0 of DR1 and DR2, denoted as the average eigenvector of the first batch, where DRO = q1*DR1 + q2*DR2, and q1 and q2 are the set weight coefficients. The second voltage kurtosis corresponding to the first single-photon detector is calculated according to the fifth formula, which is as follows: Where VFR is the second voltage kurtosis; Based on the production voltage data of the first batch, the first entanglement entropy and the second voltage kurtosis of all single-photon detectors in the first batch are repeatedly obtained, and the first operating data of the first batch is obtained after completion. Repeatedly obtain the first run data of all batches to obtain standard run data; Based on the dark count stability standard of single-photon detectors, it is determined whether the dark count stability of each single-photon detector in the historical dark count stability quality data meets the dark count stability standard. If it meets the standard, it is marked as normal dark count stability; otherwise, it is marked as abnormal dark count stability. For any position in a batch, we denote it as the first position. We obtain the dark count stability of the first position in all batches of historical dark count stability quality data, and calculate the probability of abnormal dark count stability, which is denoted as the historical abnormality rate BYi of the first position. We repeat the process of obtaining the historical abnormality rate of the positions in all batches. After completion, we obtain the first quality data. Obtain the detection cost for each single-photon detector, denoted as CB0; and obtain the average loss cost caused by the stability of the dark count of missed detection anomalies, denoted as CB1; The total number of abnormal dark counts in the first batch is denoted as YW, and YW*CB0 is denoted as the minimum detection loss cost of the first batch. The minimum detection loss cost of each batch is repeatedly obtained and merged with the first quality data for storage, which is denoted as the standard quality data. A smart sampling inspection model for single-photon detectors is constructed based on standard quality data and standard operating data, and a product sampling inspection reference scheme is obtained, including the following sub-steps: Let CU be the total cost of detection loss, where CU = CB0 * V0 + CB1 * V1, V0 is the number of single-photon detectors sampled in a batch, and V1 is the number of dark count stability of missed detection anomalies in a batch. The original sampling model is constructed based on the quantum annealing algorithm, with minimizing the CU as the core objective. A qubit is defined as σi∈{0,1}, where σi=0 represents the i-th single-photon detector in the batch that is not detected, and σi=1 represents the i-th single-photon detector in the batch that is detected. The Hamiltonian is constructed as follows: , where hi=q3*QCEi+q4*VFRi+q5*BYi, q3, q4 and q5 are weighting coefficients; Jij is the coupling parameter, Jij=q6*exp(-|ti-tj| / T0), q6 is the coupling strength coefficient, ti and tj are the production times of the i-th single-photon detector and the j-th single-photon detector, respectively; The minimum number of samples to be sampled is set to Fmin. The parameters of the model are calibrated and optimized based on standard quality data and standard operating data. The detection accuracy of the model for each batch is obtained, and the minimum detection accuracy FM and the average detection accuracy FD are recorded to ensure that the average detection accuracy of the sampling scheme output by the model is greater than k1. After completion, the intelligent sampling model of single photon detector is obtained, where k1 is the set false negative rate threshold.

2. The method for constructing a model using big data on industrial quality through quantum computing as described in claim 1, characterized in that, Obtaining the dark count stability standard and historical dark count stability quality data of the single-photon detector, and acquiring the equipment operation data of the production equipment, includes the following sub-steps: Obtain the normal standard dark count stability of the single-photon detector, denoted as the dark count stability standard of the single-photon detector; The number of single-photon detectors produced in each batch is obtained and denoted as L0. Any production batch is denoted as the first batch. For all single-photon detectors produced in the first batch, they are sorted and numbered according to the production order of the single-photon detectors and the corresponding batch. The dark count stability and production time of each single-photon detector are obtained, and the average time T0 for producing one single-photon detector is obtained. After completion, the dark count stability quality data of the first batch is obtained. Repeatedly acquire dark count stability quality data for multiple batches, and sort and number each batch according to the production sequence, denoted as batch 1-n0 respectively. After completion, historical dark count stability quality data is obtained.

3. The method for constructing a model using big data on industrial quality through quantum computing according to claim 2, characterized in that, The intelligent sampling inspection model for single-photon detectors, constructed based on standard quality data and standard operating data, and the resulting product sampling inspection reference scheme also includes the following sub-steps: For any single-photon detector in a production batch, referred to as the second batch, the equipment production and operation data corresponding to the second batch and the production time of each single-photon detector are collected. The equipment production and operation data is then processed to obtain the standard operation data corresponding to the second batch. The corresponding data is then input into the intelligent sampling inspection model of the single-photon detector to obtain {σ1, σ2, ..., σG}, which is denoted as the product sampling inspection reference scheme.

4. The method for constructing a model using big data on industrial quality through quantum computing according to claim 3, characterized in that, Based on the product sampling reference scheme and historical quality data, a sampling correction strategy is constructed, and the sampling inspection of single-photon detectors includes the following sub-steps: Set the defect rate pass threshold to BK0; conduct the first sampling inspection on the second batch of single-photon detectors based on the product sampling inspection reference scheme, obtain the total number of single-photon detectors in the first sampling inspection, denoted as YU0, and obtain the number of single-photon detectors with abnormal dark count stability in the first sampling inspection, denoted as YU1; and mark the single-photon detectors with abnormal dark count stability in the first sampling inspection as the first defective dark count stability. Calculate YU3 = YU1 / YU0. If YU3 is less than 2FM-FD, the sampling inspection is deemed a failure. Then, remove the first defective dark count stability of the second batch and randomly sample the single-photon detectors of the second batch, and calculate the defect rate BK1 of the randomly sampled single-photon detectors. If BK1 is not greater than BK0, the second batch is deemed qualified. If BK1 is greater than BK0, the second batch is deemed unqualified, and the remaining single-photon detectors of the second batch are subject to full inspection.

5. The method for constructing a model using big data on industrial quality through quantum computing according to claim 4, characterized in that, Based on the product sampling reference scheme and historical quality data, a sampling correction strategy is constructed, and the sampling inspection of single-photon detectors includes the following sub-steps: If YU3 is not less than 2FM-FD, the sampling inspection is considered successful; and BK2 is calculated, BK2=YU1 / G; and BK2 is corrected based on FD to obtain BK3, BK3=BK2 / FD. If BK3 is not greater than BK0, the second batch is considered qualified. If BK1 is greater than BK0, the first defective dark count stability of the second batch is removed, and then the single-photon detectors of the second batch are randomly sampled and the defect rate BK1 is calculated. If BK1 is not greater than BK0, the second batch is deemed qualified; if BK1 is greater than BK0, the second batch is deemed unqualified, and the remaining single-photon detectors of the second batch are subject to full inspection.

Images