System and method for manufacturing, packaging and testing of optical communication memory modules

Abstract

The application provides a manufacturing, packaging and testing integrated system and method of an optical communication storage module, and relates to the technical field of optical communication. The manufacturing, packaging and testing integrated system and method of the optical communication storage module comprises a manufacturing modeling module. Based on structure design data and manufacturing process parameters of the optical communication storage module, a multi-physical field joint modeling method is adopted, and through digital twin mapping, the optical device mounting state, the optical waveguide alignment deviation and the storage chip welding stress are cooperatively modeled to generate a manufacturing coupling model representing the optical-electric-storage coupling relationship in the manufacturing stage. By cooperatively introducing the structure design data and the manufacturing process parameters, and uniformly modeling the optical device mounting state, the optical waveguide alignment deviation and the storage chip welding stress in the manufacturing stage, the optical, electric and storage related factors form a quantifiable correlation in the same calculation framework.

Description

Technical Field

[0001] This invention relates to the field of optical communication technology, specifically to an integrated system and method for manufacturing, packaging and testing optical communication storage modules. Background Technology

[0002] The field of optical communication technology primarily involves using light as an information carrier to achieve high-speed transmission, processing, storage, and exchange of information in optical fibers or integrated optical waveguides. This technology integrates multiple disciplines, including optoelectronic devices, semiconductor processes, precision manufacturing, packaging engineering, signal processing, and system testing. Its core lies in leveraging the advantages of optical signals over electrical signals in terms of bandwidth, speed, anti-interference capability, and energy consumption to meet the demands of data centers, high-performance computing, artificial intelligence, cloud computing, and next-generation communication networks for ultra-high speed, low latency, and high reliability. Specifically, the integrated system and method for manufacturing, packaging, and testing optical communication storage modules refers to a highly integrated approach for optical communication applications, combining optical communication functional units with storage functional units. Through a unified manufacturing process, coordinated packaging control, and integrated performance testing, it achieves a holistic technical solution from manufacturing to packaging and testing of the module. Its purpose is to improve the manufacturing consistency, packaging reliability, and testing accuracy of optical communication storage modules, reduce system integration complexity, and enhance the overall performance and long-term stability of the module in high-speed data transmission and data storage applications.

[0003] In practical applications, existing optical communication technologies often divide processing flows by function or stage. Manufacturing, packaging, and testing typically rely on independent data systems and evaluation standards, lacking a unified way to express the correlation between parameters at each stage. This makes it difficult to directly identify and quantify structural deviations or stress hazards formed during the manufacturing stage in the packaging and testing phases. After manufacturing, the packaging process often focuses on the stability assessment of materials and processes themselves, while paying insufficient attention to the inheritance relationship between the optical and electrical states of the manufacturing stage. As a result, when optical path attenuation or stress concentration occurs after packaging, it can only be corrected through empirical rework or adjustments, increasing time costs and uncertainty. In the testing phase, optical communication performance testing and storage performance testing are usually conducted separately, and the test results exist in the form of independent indicators, making it difficult to reflect the mutual influence between the two under high-density integration conditions. Some latent defects are not easily exposed in individual tests, only gradually appearing during system-level operation, thus leading to performance fluctuations or failure risks. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an integrated system and method for manufacturing, packaging, and testing optical communication storage modules. This solves the problem that existing optical communication technologies often divide processing flows by function or stage in practical applications. Manufacturing, packaging, and testing typically rely on independent data systems and evaluation standards, and there is a lack of a unified way to express the correlation between parameters at each stage. This makes it difficult to directly identify and quantify structural deviations or stress hazards formed during the manufacturing stage in the packaging and testing stages.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: an integrated system and method for manufacturing, packaging and testing optical communication storage modules, comprising the following modules: a manufacturing modeling module, which, based on the structural design data and manufacturing process parameters of the optical communication storage module, adopts a multi-physics joint modeling method and, through digital twin mapping, collaboratively models the mounting state of optical devices, the alignment deviation of optical waveguides and the welding stress of storage chips, generating a manufacturing coupling model that characterizes the optical-electrical-storage coupling relationship in the manufacturing stage;

[0006] The packaging collaboration module, based on the manufacturing coupling model, adopts the optical coupling compensation calculation method and uses the thermo-mechanical-optical path collaborative analysis method to dynamically evaluate and compensate for the material deformation, thermal stress distribution and optical path attenuation generated during the packaging process, and generates a set of packaging compensation parameters for packaging correction.

[0007] The test fusion module, based on the encapsulation compensation parameter set, adopts a multi-dimensional joint test method and, through the synchronous execution mechanism of optical communication performance test and storage access performance test, performs fusion acquisition and feature construction on optical communication bit error rate, optical power stability and storage read / write latency, and generates a joint test feature set that comprehensively characterizes the module's performance status.

[0008] The defect reasoning module, based on the joint test feature set, adopts a defect intelligent reasoning method and performs cross-stage correlation analysis and defect localization on optical communication performance anomalies, packaging stress mismatch and storage interface anomalies through process parameter knowledge correlation analysis, and generates corresponding defect tracing results.

[0009] The optimization feedback module, based on the defect tracing results, adopts a closed-loop parameter optimization method and a manufacturing-packaging-testing parameter reverse adjustment mechanism to comprehensively optimize manufacturing process parameters, packaging compensation strategies, and test thresholds, generating a full-process optimization scheme for system self-iterative improvement.

[0010] Preferably, the manufacturing modeling module includes a process acquisition submodule, a structure mapping submodule, a parameter modeling submodule, and a state simulation submodule;

[0011] The process acquisition submodule uses manufacturing production line sensor information and adopts a real-time process parameter acquisition method to obtain mounting accuracy, welding temperature curve and fiber alignment offset data, and generates a manufacturing process dataset.

[0012] The structure mapping submodule constructs a virtual manufacturing structure and generates a virtual structure mapping model based on the manufacturing process dataset using a digital twin structure mapping method.

[0013] The parameter modeling submodule, based on the virtual structure mapping model, uses a multi-physics joint modeling method to construct the optical, electrical and thermal coupling relationship and generate a multi-physics coupled parameter model.

[0014] The state simulation submodule, based on the multi-physics coupling parameter model, uses finite element simulation and ray tracing calculation methods to simulate the coupling state during the manufacturing stage and generate a manufacturing coupling model.

[0015] Preferably, the encapsulation collaboration module includes a material analysis submodule, an optical path correction submodule, a stress assessment submodule, and a compensation calculation submodule;

[0016] The material analysis submodule generates packaging material characteristic parameters based on the manufacturing coupling model and using a thermo-mechanical property analysis method for packaging materials.

[0017] The optical path correction submodule generates optical path correction parameters based on the packaging material properties using an optical coupling offset compensation method.

[0018] The stress assessment submodule generates packaging stress assessment results based on the optical path correction parameters using a thermo-mechanical stress assessment method.

[0019] The compensation calculation submodule generates a set of packaging compensation parameters based on the packaging stress assessment results using an adaptive Kalman filter compensation method.

[0020] Preferably, the test fusion module includes an optical parameter test submodule, a storage test submodule, a data synchronization submodule, and a feature construction submodule;

[0021] The optical parameter testing submodule generates optical communication test data based on the encapsulation compensation parameter set and using optical communication performance testing methods.

[0022] The storage test submodule generates storage performance test data based on the optical communication test data and using a storage access performance test method.

[0023] The data synchronization submodule generates a synchronization test dataset based on the storage performance test data using a multi-source data time synchronization fusion method.

[0024] The feature construction submodule generates a joint test feature set based on the synchronous test dataset using feature normalization and principal component analysis methods.

[0025] Preferably, the defect reasoning module includes a feature analysis submodule, a knowledge association submodule, a defect classification submodule, and a location determination submodule;

[0026] The feature analysis submodule generates anomaly feature vectors based on the joint test feature set using a multidimensional feature statistical analysis method.

[0027] The knowledge association submodule generates a defect association graph based on the abnormal feature vector using a process parameter knowledge graph association reasoning method.

[0028] The defect classification submodule generates defect type determination results based on the defect association map using support vector machine or deep neural network classification methods.

[0029] The location determination submodule generates defect tracing results based on the defect type determination result using a cross-stage defect location method.

[0030] Preferably, the optimization feedback module includes a parameter backtracking submodule, a strategy generation submodule, a process adjustment submodule, and a scheme output submodule;

[0031] Based on the defect tracing results, the parameter backtracking submodule generates a set of key parameter deviations using a manufacturing-packaging parameter backtracking method.

[0032] The strategy generation submodule generates an optimized strategy parameter set based on the key parameter deviation set using a model predictive control optimization method.

[0033] The process adjustment submodule generates an optimized process configuration based on the optimization strategy parameter set using a closed-loop process parameter adjustment method.

[0034] The output submodule generates a full-process optimization scheme based on the optimized process configuration.

[0035] Preferably, the integrated method for manufacturing, packaging, and testing optical communication storage modules includes the following steps:

[0036] S1: Based on the manufacturing production line data of optical communication storage modules, manufacturing process parameters such as mounting accuracy, welding temperature curve, and fiber alignment offset are acquired in real time using a process parameter acquisition method. The process parameters are then cleaned and normalized. On this basis, a virtual mapping model of the manufacturing structure is constructed using a digital twin structure mapping method. Furthermore, a multi-physics joint modeling method is used to establish the coupling relationship between optics, electricity, and heat, thereby forming a parameter model that can characterize the state of the manufacturing stage and generating a multi-physics coupled parameter model.

[0037] S2: Based on the multi-physics coupling parameter model generated in S1, the deformation characteristics of the packaging material under thermal and mechanical loads are analyzed using the thermo-mechanical property analysis method to obtain the characteristic parameters of the packaging material. Based on the characteristic parameters of the packaging material, the optical path offset generated during the packaging process is corrected using the optical coupling offset compensation method. The stress distribution during the packaging stage is calculated using the thermo-mechanical stress assessment method, and the packaging parameters are dynamically adjusted using the Kalman filter compensation method to generate a packaging compensation parameter set.

[0038] S3: Based on the encapsulation compensation parameter set generated in S2, optical power, bit error rate, and eye diagram quality are tested using optical communication performance testing methods to obtain optical communication performance data. On this basis, storage performance testing methods are used to obtain read / write latency, bandwidth, and error count data of the storage module. A multi-source data time synchronization fusion method is used to synchronize the optical communication performance data and storage performance data, forming a unified dataset. For the synchronized dataset, feature normalization and principal component analysis methods are used to construct joint test features. Furthermore, support vector machines or deep neural network classification methods are used to analyze the joint test features to determine defect types. Combined with process parameter correlation reasoning methods, the source of defects is determined, generating defect tracing results.

[0039] Preferably, step S1 includes the following steps:

[0040] S101: Based on the real-time operation status of the optical communication storage module manufacturing production line, a real-time process parameter acquisition method is adopted. Through a multi-sensor synchronous acquisition mechanism, manufacturing process parameters such as mounting accuracy, welding temperature curve, and fiber alignment offset are acquired and time-calibrated to generate the original manufacturing process parameter set.

[0041] S102: Based on the original manufacturing process parameter set, data cleaning and normalization methods are used to standardize the manufacturing process parameters by outlier removal, noise filtering and scale unification, and a standardized manufacturing process parameter set is generated.

[0042] S103: Based on the standardized manufacturing process parameter set, a digital twin structure mapping method is adopted to construct a virtual mapping model of the manufacturing structure through a one-to-one mapping mechanism between physical structure parameters and virtual model parameters, and to generate a virtual structure mapping model.

[0043] S104: Based on the virtual structure mapping model, a multi-physics joint modeling method is adopted. Through the coupled modeling of optical, electrical and thermal parameters, the multi-physics coupling relationship in the manufacturing stage is calculated and modeled to generate a multi-physics coupling parameter model.

[0044] Preferably, step S2 includes the following steps:

[0045] S201: Based on the multi-physics coupled parameter model, the thermo-mechanical property analysis method of the packaging material is adopted. The deformation characteristics of the packaging material under thermal and mechanical loads are analyzed by finite element calculation, and the property parameters of the packaging material are generated.

[0046] S202: Based on the characteristic parameters of the packaging material, the optical coupling offset compensation method is used to correct the optical path offset generated during the packaging process by inverting the optical path offset and generating optical path correction parameters.

[0047] S203: Based on the optical path correction parameters, a thermo-mechanical stress assessment method is adopted to comprehensively assess the thermal and mechanical stresses during the packaging stage through stress distribution calculation, and generate packaging stress assessment results;

[0048] S204: Based on the packaging stress assessment results, a Kalman filter compensation method is adopted to adaptively compensate key parameters in the packaging process through a dynamic parameter correction mechanism, thereby generating a packaging compensation parameter set.

[0049] Preferably, step S3 includes the following steps:

[0050] S301: Based on the encapsulation compensation parameter set, an optical communication performance testing method is adopted. By testing optical power, bit error rate and eye diagram quality, optical communication performance data is obtained and optical communication performance test data is generated.

[0051] S302: Based on the optical communication performance test data, a storage performance test method is adopted to obtain storage performance data by testing read / write latency, bandwidth and error count, and then generating storage performance test data.

[0052] S303: Based on the optical communication performance test data and storage performance test data, a multi-source data time synchronization fusion method is adopted to unify the multi-source test data through time axis alignment processing, and generate a synchronous test dataset;

[0053] S304: Based on the aforementioned synchronous test dataset, feature normalization and principal component analysis methods are used to extract joint test features through feature dimensionality reduction and feature reconstruction, generating a joint test feature set;

[0054] S305: Based on the joint test feature set, a support vector machine or deep neural network classification method is used, combined with a process parameter correlation reasoning method, to determine the defect type and its source, and generate defect tracing results.

[0055] This invention provides an integrated system and method for manufacturing, packaging, and testing optical communication storage modules. It offers the following advantages:

[0056] This invention integrates structural design data with manufacturing process parameters, and performs unified modeling of optical device mounting status, waveguide alignment deviations, and memory chip soldering stress during the manufacturing stage. This establishes a quantifiable correlation between optical, electrical, and storage-related factors within the same computational framework, preventing the accumulation of deviations caused by the fragmented processing of parameters from different physical dimensions. In the packaging stage, the coupling relationship formed during manufacturing is used as input to simultaneously evaluate and compensate for material deformation, thermal stress distribution, and optical path attenuation. This ensures that structural changes and optical performance changes during packaging remain dynamically consistent, reducing the risk of post-packaging optical power fluctuations and long-term reliability degradation. In the testing stage, optical communication performance indicators and memory access performance indicators are incorporated into a single process. By integrating data collection and feature construction under inter-benchmark conditions, the module performance status can reflect the coupling effects across functional units in a unified feature form, avoiding the situation where a single test indicator is insufficient to reveal potential hidden problems. On this basis, by reasoning and analyzing the correlation between test features and historical process parameters, cross-stage correspondence between performance anomalies and manufacturing and packaging processes can be achieved. This allows defect localization to move beyond judgment of a single process and point to more targeted source factors. Furthermore, closed-loop optimization is formed through reverse parameter adjustment, enabling the manufacturing process, packaging strategy, and test thresholds to gradually converge in multiple rounds of operation. This improves the overall stability, consistency, and predictability of module performance, while reducing resource consumption caused by repeated trial and error and reliance on human experience. Attached Figure Description

[0057] Figure 1 This is a system block diagram of the present invention;

[0058] Figure 2 This is a schematic diagram of the main steps of the present invention;

[0059] Figure 3 This is a detailed schematic diagram of S1 of the present invention;

[0060] Figure 4 This is a detailed schematic diagram of S2 of the present invention;

[0061] Figure 5 This is a detailed schematic diagram of S3 of the present invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] Example:

[0064] like Figure 1-5 As shown, this embodiment of the invention provides an integrated system and method for manufacturing, packaging and testing optical communication storage modules, including the following modules: a manufacturing modeling module, which, based on the structural design data and manufacturing process parameters of the optical communication storage module, adopts a multi-physics joint modeling method and uses a digital twin mapping method to collaboratively model the mounting state of optical devices, the alignment deviation of optical waveguides and the welding stress of storage chips, and generates a manufacturing coupling model that characterizes the optical-electrical-storage coupling relationship in the manufacturing stage;

[0065] The packaging collaboration module, based on the manufacturing coupling model, adopts the optical coupling compensation calculation method and uses the thermo-mechanical-optical path collaborative analysis method to dynamically evaluate and compensate for the material deformation, thermal stress distribution and optical path attenuation generated during the packaging process, and generates a set of packaging compensation parameters for packaging correction.

[0066] The test fusion module, based on the encapsulation compensation parameter set, adopts a multi-dimensional joint test method and uses a synchronous execution mechanism of optical communication performance test and storage access performance test to perform fusion acquisition and feature construction of optical communication bit error rate, optical power stability and storage read and write latency, and generate a joint test feature set that comprehensively characterizes the module performance status.

[0067] The defect reasoning module, based on the joint test feature set, adopts the intelligent defect reasoning method and performs cross-stage correlation analysis and defect localization on optical communication performance anomalies, packaging stress mismatch and storage interface anomalies through process parameter knowledge correlation analysis, and generates corresponding defect tracing results.

[0068] The feedback module is optimized. Based on the defect tracing results, a closed-loop parameter optimization method is adopted. Through the reverse adjustment mechanism of manufacturing-packaging-testing parameters, the manufacturing process parameters, packaging compensation strategies and test thresholds are comprehensively optimized to generate a full-process optimization scheme for system self-iterative improvement.

[0069] The manufacturing modeling module includes a process acquisition submodule, a structure mapping submodule, a parametric modeling submodule, and a state simulation submodule.

[0070] The process acquisition submodule uses manufacturing production line sensor information and adopts a real-time process parameter acquisition method to obtain mounting accuracy, welding temperature profile and fiber alignment offset data, and generates a manufacturing process dataset.

[0071] The multi-point sensing units on the production line are deployed according to workstation numbers. Each workstation is equipped with a displacement sensor, a thermocouple, and a vision alignment camera. The sampling period is set to a millisecond-level time step. A sequence set is formed within a continuous time window T, and the mounting position coordinates are marked as follows. , and design coordinates Perform deviation calculation, deviation amount ,when The temperature is considered normal if it falls within a preset range of 0 to several micrometers; otherwise, it is marked as abnormal. The welding temperature is obtained instantaneously via a thermocouple. Using the moving average formula Suppressing fluctuations, for the slope of temperature rise Record and compare with the baseline slope In comparison, the fiber-aligned image is calculated using the gray-scale centroid method to determine the center coordinates (cx, cy), and the offset is... All parameters are concatenated according to timestamps to form a matrix. For example, several batches of components were selected during the placement machine operation, and hundreds of samples were collected and written into the database. After uniform formatting, a manufacturing process dataset that can be directly accessed was obtained.

[0072] The structure mapping submodule uses a digital twin structure mapping method to construct a virtual manufacturing structure based on the manufacturing process dataset, and generates a virtual structure mapping model.

[0073] The aforementioned data matrix and the 3D structure file undergo coordinate unification processing. A homogeneous transformation matrix H is used to represent the mapping from actual coordinates to model coordinates. H consists of a rotation matrix R and a translation vector t. For each measured point implement The model's spatial position, mounting accuracy, temperature field, and alignment deviation are obtained as attribute vectors. Attached to nodes, forming a weighted set G of nodes, with weights Assign based on parameter sensitivity and satisfy In the example, similar proportions are used to ensure that all three types of factors participate in the calculation. When constructing the grid cells, several voxels are divided according to the step size h, and the cell attributes are calculated using a weighted average. Subsequently, structural dimensions such as millimeter-scale packages and micrometer-scale optical waveguides were imported into the simulation software and calibrated. The errors were compared using several reference points. ,when Matching is determined within a small-scale interval, resulting in a virtual structure mapping model that is consistent with the real production line.

[0074] The parametric modeling submodule is based on the virtual structure mapping model and uses a multi-physics joint modeling method to construct the optical, electrical and thermal coupling relationship and generate a multi-physics coupled parametric model.

[0075] On the established virtual geometric mesh, three field variables are defined: refractive index n, electrical conductivity σ, and thermal conductivity λ. The variation of each variable with spatial coordinate r is denoted as... The baseline value was obtained through experimental samples and used as a reference coefficient. The optical propagation loss was calculated using... It means that, among them Here, ρ is the loss at room temperature, and β is the temperature coefficient. β is obtained by averaging multiple measurements. The resistivity ρ satisfies the following relationship with temperature. γ is selected from the median value of the range given in the material handbook, and thermal diffusion satisfies The temperature of each node is obtained through discretization. The three types of field quantities form a vector at the same node. Calculate the gradient of adjacent nodes using differential calculation and then calculate the coupling degree. The coefficients a and b are allocated according to the proportion of influence, and a + b = 1. For example, consider calculating the values ​​of several nodes in the neighborhood of a memory chip. The data is stored in a table, and continuous regions are filled in using interpolation to form a multi-physics coupled parameter model that can be directly used in subsequent calculations.

[0076] The state simulation submodule is based on a multi-physics coupling parameter model and uses finite element simulation and ray tracing calculation methods to simulate the coupling state of the manufacturing stage and generate a manufacturing coupling model.

[0077] The calculation process revolves around the state evolution. A finite element mesh is established based on the aforementioned coupling parameters, and a material matrix D is assigned to each element. The displacement field u satisfies K·u=F, where K is the global stiffness matrix and F is the external load vector. The nodal stresses are obtained using a step-by-step iterative method. At the same time, ray aggregation is used for the optical path. Tracing the ray, each ray follows Snell's law. Calculate the angle of refraction and accumulate the path length. Power attenuation according to The thermal field temperature result is obtained as input to update the α value and the calculation is repeated for several steps to form a time series. The iteration terminates when the difference between two adjacent results is less than the set convergence threshold η. The threshold η is a small-scale interval value based on the measurement accuracy. In the example, the cyclic solution is performed on several packaging batches to obtain the stress, temperature and optical power distribution arrays at each moment. The summary output contains a manufacturing coupling model containing multiple field coupling states.

[0078] The packaging collaboration module includes a material analysis submodule, an optical path correction submodule, a stress assessment submodule, and a compensation calculation submodule;

[0079] The materials analysis submodule is based on the manufacturing coupling model and uses the thermo-mechanical property analysis method of packaging materials to generate the property parameters of packaging materials;

[0080] The materials, optical, and mechanical elements involved in the collaborative module are uniformly decomposed, and the packaging structure is regarded as a combination of multiple media and connection interfaces. First, the material categories corresponding to different packaging layers are identified. For example, the chip carrier layer, transition buffer layer, and external protective layer correspond to different material groups. The temperature field distribution, stress distribution, and optical power distribution in the coupling model obtained during the manufacturing stage are used as input sources. During the material analysis process, the temperature change range and stress change range of relevant regions are read layer by layer. A set of parameters such as thermal expansion coefficient, elastic modulus, and Poisson's ratio are established for each type of material. The thermal expansion coefficient is calculated by comparing the structural dimension change at the upper and lower temperature limits. The dimension change ΔL is compared with the original length. Between α is taken from the average interval value of multiple interval fittings, and the elastic modulus is obtained through the proportional relationship between stress σ and strain ε. The strain is estimated by displacement field difference. By summarizing the results of multiple grid nodes and removing outliers, a stable set of material parameters is formed. A case study is conducted in conjunction with actual packaging styles, such as small-sized optoelectronic device packaging structures. The parameters of multilayer materials are summarized and output in tabular form as packaging material characteristic parameters that can be repeatedly called by subsequent modules.

[0081] The optical path correction submodule generates optical path correction parameters based on the packaging material properties and using an optical coupling offset compensation method.

[0082] Based on the existing set of material parameters, the refractive index ranges, thermotropic coefficients, and interface tilt angles corresponding to different encapsulation layers are organized. During the optical path correction process, a representative light propagation path in the manufacturing coupling model is first selected as the reference path. The incident and exit angles of this path under conditions without encapsulation disturbance are used as the reference angle range. After incorporating material properties, the refractive change of each path segment is calculated segment by segment. The refractive angle change is obtained through... The relationship is estimated, where the refractive index n changes linearly with temperature. k represents the range of photothermal coefficients of the material. By comparing the offsets Δx and Δy of the center position of the light spot before and after correction, the offsets are mapped to optical path compensation vectors. The size of the compensation vectors is divided into three levels: small, medium, and large, according to a preset range and recorded separately. In the example, multiple offset results of a certain optical channel under different temperature distributions are selected and averaged. Finally, the compensation vectors of each channel are associated with the corresponding material segments to form a set of optical path correction parameters that can be repeatedly applied.

[0083] The stress assessment submodule generates packaging stress assessment results based on optical path correction parameters and using a thermo-mechanical stress assessment method.

[0084] After correcting the optical path parameters, the structural deformation information corresponding to the corrected parameters is introduced as an additional condition into the mechanical analysis process. During stress assessment, the free expansion of the encapsulated structure in different temperature ranges is first estimated based on the elastic modulus and thermal expansion coefficient ranges in the material parameters. Then, the free expansion is superimposed with the structural constraints to obtain the actual displacement field. The strain tensor is calculated using the displacement gradient, and the strain in each direction is obtained using a finite difference method. The stress is then calculated through... The relationship calculation is performed, where D is the interval representation of the material stiffness matrix. The calculation results are classified by region, and the stress values ​​are divided into low, medium and high intervals. The interval boundaries are set according to the statistical distribution of historical encapsulation samples. The stress results of key interfaces and areas adjacent to the light channel are recorded in detail. Examples are given to illustrate the trend interval of stress change under multiple thermal cycles under a certain type of encapsulation. Finally, the encapsulation stress evaluation results corresponding to each region of the structure are summarized.

[0085] The compensation calculation submodule generates a set of packaging compensation parameters based on the packaging stress assessment results using an adaptive Kalman filter compensation method.

[0086] After obtaining the stress assessment results, the stress intervals of each region and the corresponding optical path offset intervals are used together as the state observation inputs for the compensation calculation process. In the compensation calculation, the state vector is first defined to include two components: optical path offset and structural displacement. The observation vector consists of the intervalized results obtained from the current assessment. The prediction process is calculated based on the state vector and state transition relationship at the previous moment. The prediction error covariance is initialized through an empirical interval. In the update stage, the observation results are introduced and the Kalman gain is calculated. The gain is adjusted according to the relative proportion between the prediction uncertainty interval and the observation uncertainty interval. The state update is completed by weighted superposition. Through multiple iterations, the state change tends to a stable interval. The example uses several continuous encapsulation conditions as a time series for demonstration. The compensation vector after each iteration is recorded and smoothed to finally form an encapsulation compensation parameter set containing information from multiple time steps and multiple regions.

[0087] The test fusion module includes an optical parameter testing submodule, a storage testing submodule, a data synchronization submodule, and a feature construction submodule;

[0088] The optical parameter testing submodule generates optical communication test data based on the encapsulation compensation parameter set and using optical communication performance testing methods.

[0089] The optical link measurement, storage access measurement, time alignment processing, and feature construction elements involved in the test fusion module are decomposed. The pre-encapsulation compensation parameter set is regarded as the input constraint set of the test conditions. First, the compensation vectors related to optical path offset and the compensation amounts related to structural displacement in the compensation parameter set are classified and labeled. Then, in the test preparation stage, the connectivity of the optical module port, switching interface, and storage control channel is verified. During the verification, the bit error count of port loopback and the link lock duration are recorded. The bit error count is divided into low, medium, and high intervals and written into the test form. At the same time, read and write command receipt checks are performed on the storage channel, and the receipt delay is statistically analyzed and grouped by interval. Subsequently, the optical link test sampling period and the storage access sampling period are aligned and set. During the setting, the rising edge of the synchronous trigger signal is used as the reference time. The time deviation Δt of each sampling source is calculated and corrected. Δt is obtained through... Obtain, among which Let i be the sampling timestamp of the i-th source. Using the reference timestamp, the example selects a packaged optoelectronic device for continuous sampling, and maps the optical power sampling sequence and the storage throughput sampling sequence to the same time axis to form a set of original test records that can be used for subsequent construction, serving as the input structured list for the test fusion module.

[0090] The storage test submodule generates storage performance test data based on optical communication test data and using storage access performance test methods.

[0091] The test conditions are decomposed based on the packaged compensation parameter set. The offset components in the compensation parameters are denoted as Δx and Δy, and the compensation amount related to temperature drift is denoted as... These quantities are written into the test script configuration items as test conditions. During optical parameter testing, the transmitting end code pattern and rate level are first set and link training is performed. During the training phase, the sampling point position and equalizer tap coefficient changes at the receiving end are recorded. Then, key indicators, including received optical power, are collected. Transmitted optical power Error counting With total number of bits Bit error rate according to Calculation, where To count the number of error bits within the window, To count the number of bits transmitted within the window, the example samples three temperature ranges and substitutes the compensation amount into the correction. The correction method is to map the measured spot center offset δ to the compensated offset. ,in To compensate for the weighting coefficient, The value of is determined by referring to the offset suppression ratio range of similar historical packaging samples and by minimizing the offset residual in pre-experiments. When r falls within a small interval, the link is considered to be in a stable state; when r falls within a large interval, it is recorded as an abnormal state. Simultaneously, the eye diagram opening height is collected. With shaking And labeled by interval, finally The corresponding compensation condition labels are summarized and written into a record file to generate optical communication test data.

[0092] The data synchronization submodule generates a synchronization test dataset based on storage performance test data using a multi-source data time synchronization fusion method.

[0093] Based on the time window division and link status labels in the optical communication test data, the sampling window for storage testing is consistent with the optical link sampling window. During storage access testing, a read / write mixing ratio and queue depth levels are defined, and the number of access requests corresponding to each level is recorded as follows: After the request is issued, the number of completed requests will be counted. With cumulative time Throughput by The calculation is performed, where Q represents the number of requests completed per unit time. To calculate the cumulative time consumed within the statistics window, the average access latency is calculated as follows: Calculation, where To average latency, the example measures and records Q and Q under four types of load: sequential read, random read, sequential write, and random write. The latency is then divided into low-latency, medium-latency, and high-latency intervals, with interval boundaries set based on historical batch statistical percentiles for the same interface specification. Simultaneously, bandwidth utilization is also considered. Perform calculations. ,in To measure the effective bandwidth, For nominal bandwidth, if If it falls into the lower range, it is recorded as a bandwidth-limited segment. If it falls into the high range, it is recorded as saturation. Then, the data within each time window is recorded. The data is associated with the corresponding optical link status tag and written to form storage performance test data.

[0094] The feature construction submodule generates a joint test feature set based on the synchronous test dataset, using feature normalization and principal component analysis methods.

[0095] Based on storage performance test data, the optical link index sequence and storage index sequence are used as multi-source inputs. First, the timestamps of each source are unified. During unification, the period count of the reference clock source is used as the index to map the sampling points of each source to the same index k. The mapping error is expressed as... It means that when If it falls into a small interval, it will be directly aligned. If the value falls within a large interval, linear interpolation is used to fill the gap. The interpolation is performed according to... Calculate, where x is the index to be filled, and k1 and k2 are the indices of adjacent valid sampling points. After alignment, points under the same k will be... After concatenating the data to form a synchronized record row, a synchronized test dataset is obtained. Then, normalization is performed on each metric. Normalization uses... Where x is the original value, μ is the mean of the indicator in the current batch, and σ is the standard deviation. The mean and standard deviation are calculated by summing the batch samples. Outliers are marked according to the rule that |z| falls into the high interval, and then principal component construction is performed, first calculating the covariance matrix. Where X is the normalized sample matrix, then the eigenvectors are calculated and sorted by eigenvalue. The top m principal components whose cumulative contribution rates fall within the target interval are selected, and the contribution rates are ranked according to... Calculation, where Let be the j-th feature value, and p be the original feature dimension. In the example, the scores of the first few principal components are used as joint features and time window labels are attached to the output to form a joint test feature set.

[0096] The defect reasoning module includes a feature analysis submodule, a knowledge association submodule, a defect classification submodule, and a location determination submodule;

[0097] The feature analysis submodule generates abnormal feature vectors based on the joint test feature set and using multi-dimensional feature statistical analysis methods.

[0098] The feature analysis submodule processes the joint test feature set using multidimensional feature statistical analysis methods. First, it preprocesses the data for each feature vector, removing missing and outlier values ​​to ensure data integrity and accuracy. Then, based on the distribution of each feature, it normalizes the features using standardization methods. Where x is the feature value, μ is the mean, and σ is the standard deviation, next, anomaly detection is performed. The distribution range of a certain feature is selected, and the standard deviation of each feature is calculated to determine whether there is a significant anomaly. If the z value of a certain feature is greater than a certain threshold (such as 3), it is considered an anomaly. Then, the abnormal features are marked to form an anomaly feature vector. In the example, if the z value of feature P_rx is greater than 3, the optical power data is determined to be abnormal. Finally, a vector set containing abnormal features is generated for use by the subsequent knowledge association reasoning module.

[0099] The knowledge association submodule generates a defect association graph based on anomaly feature vectors and using a process parameter knowledge graph association reasoning method.

[0100] The knowledge association submodule processes abnormal feature vectors using a process parameter knowledge graph association reasoning method. First, based on the abnormal information in the feature vector, relevant process parameters are extracted to construct a knowledge graph, which includes the relationship between each process parameter and the defect type. Through graph reasoning technology, nodes in the graph that match the abnormal features are found. Then, the correlation between nodes is analyzed, and the similarity between nodes is calculated. Based on these correlations, process parameter configurations that may lead to defects are inferred. If a node (such as a temperature parameter node) has a high correlation with an abnormal feature (such as optical power with an offset greater than 3) and the node is within a significant threshold range, then the process parameter configuration is determined to be a defect source, and a defect association graph is generated. In the example, if the temperature value has a high correlation with the abnormal value of P_rx, the final generated graph will mark temperature as a key defect factor.

[0101] The defect classification submodule generates defect type determination results based on the defect association map and using support vector machine or deep neural network classification methods.

[0102] The defect classification submodule classifies defect types using either Support Vector Machines (SVM) or Deep Neural Networks (DNN). First, the generated defect association map is converted into feature input and fed into the classification algorithm. The SVM uses a kernel function to map the feature data, selecting an appropriate kernel function (such as a Gaussian kernel or a linear kernel). By dividing the training and test sets, the support of each feature is calculated, and the classification margin is used to judge the classification effect. Finally, based on the support, the defect type result is output. The Deep Neural Network is trained using a multilayer perceptron structure, passing the input features through hidden layers, calculating the activation value of each layer, and finally outputting the predicted defect type value. In the example, if a feature value is classified as "optical power imbalance" by the DNN, then the defect type is determined to be optical power imbalance.

[0103] The location determination submodule generates defect tracing results based on the defect type determination result and using a cross-stage defect location method.

[0104] The location determination submodule generates defect tracing results based on the defect type determination result and through a cross-stage defect location method. First, according to the defect type information, the corresponding location strategy is selected. If the defect type is "optical power misalignment", then a cross-stage location strategy for the optical link is selected. Combining historical data, the key nodes in each stage that may lead to this type of defect are analyzed, such as temperature changes in the optical module or angle offset of the fiber access. Then, a multi-stage feedback analysis method is used to further narrow down the defect range by comparing the process parameters of each stage, and finally outputs the specific location of the defect, such as "optical module port 2" or "fiber connector". In the example, if the optical power misalignment occurs at a certain optical module port, the defect location is finally determined to be that port through the location method.

[0105] The optimization feedback module includes a parameter backtracking submodule, a strategy generation submodule, a process adjustment submodule, and a solution output submodule;

[0106] Based on the defect tracing results, the parameter backtracking submodule generates a set of key parameter deviations using the manufacturing-packaging parameter backtracking method.

[0107] The parameter backtracking submodule generates a set of key parameter deviations using a manufacturing-packaging parameter backtracking method. First, based on the defect tracing results, it extracts various process parameters, such as temperature, humidity, and pressure, from relevant historical process data and performs backtracking analysis according to the process flow, gradually deriving the parameter deviations in each process stage. For each stage, it analyzes the difference between the parameter value and the ideal value and calculates the deviation amount. For example, assuming the deviation of the temperature parameter from the set target temperature in a certain stage is ΔT, if this deviation value is greater than a set threshold (such as ±5°C) in the actual process, the temperature deviation in that stage is considered significant, and a corresponding set of key parameter deviations is generated. Finally, a result set containing all key process deviations is formed. In the example, if the temperature deviation in the packaging stage is ΔT=7°C and exceeds the threshold range, this deviation will be included in the deviation set for analysis. The final generated deviation set is used in the subsequent optimization strategy generation submodule.

[0108] The strategy generation submodule generates an optimized strategy parameter set based on the key parameter deviation set and using a model predictive control optimization method.

[0109] The strategy generation submodule generates an optimized strategy parameter set using model predictive control optimization (MDI). First, based on the backtracked set of key parameter deviations, it selects variables related to these deviations for modeling. Then, it employs common control algorithms, such as PID controllers, to optimize the deviation (e.g., ΔT) and calculate the ideal control target. For example, assuming the target temperature is set to... The current actual temperature is Calculate the deviation Based on the PID control principle, the control quantity Δu is calculated using the formula. Where Kp is the proportional coefficient, Ki is the integral coefficient, Kd is the derivative coefficient, and e(t) is the temperature deviation, the optimized strategy parameter set is obtained. If, after calculation, Δu = 1.2 is obtained, the corresponding adjustment parameters will be output through the feedback system and applied to process adjustment. Finally, the optimized strategy parameter set is formed, providing a basis for subsequent process adjustment.

[0110] The process adjustment submodule generates an optimized process configuration based on the optimized strategy parameter set and using a closed-loop process parameter adjustment method.

[0111] The process adjustment submodule generates an optimized process configuration based on an optimized strategy parameter set using a closed-loop process parameter adjustment method. First, by analyzing the adjustment requirements in the optimized strategy parameter set, a suitable adjustment method is selected to correct key process parameters. For example, if the optimized strategy suggests adjusting the temperature parameter to 26°C, the system will automatically adjust the equipment temperature control system using a closed-loop adjustment method. This involves real-time monitoring of temperature values ​​and feedback on deviations during the adjustment process to ensure the temperature fluctuates within the set range. The adjustment method in this process is based on actual measurements and feedback information. The control system fine-tunes various process parameters such as temperature and humidity to ensure they conform to the set optimization strategy. During the adjustment process, if the temperature exceeds the set target, the system will continue to adjust and correct through a feedback mechanism, ultimately achieving the ideal process configuration. In the example, if the temperature is set to 26°C after adjustment, and the system feedback indicates the temperature has reached the target temperature range, then the optimized process configuration is confirmed to be complete.

[0112] The solution output submodule generates a full-process optimization solution based on the optimized process configuration.

[0113] The solution output submodule generates a full-process optimization solution based on the optimized process configuration. First, based on the adjusted optimized process configuration, it integrates the improvement measures of each process link to formulate a full-process optimization solution. For example, assuming that the adjusted parameters such as temperature, pressure, and humidity have been optimized to the best state, the system will combine the parameters of each link to output the optimized process operation guide and generate the final full-process solution, including operation procedures, control requirements, equipment adjustments, personnel operation specifications, etc., ultimately forming a complete optimization solution and outputting it for easy implementation in actual production. In the example, if the generated solution includes temperature control requirements, pressure setpoints, etc., the system will automatically generate an operation manual for process engineers to refer to, ensuring the smooth implementation of the process.

[0114] An integrated method for manufacturing, packaging, and testing optical communication storage modules includes the following steps:

[0115] S1: Based on the manufacturing production line data of optical communication storage modules, a real-time process parameter acquisition method is used to obtain manufacturing process parameters such as mounting accuracy, welding temperature curve, and fiber alignment offset. The process parameters are then cleaned and normalized. On this basis, a virtual mapping model of the manufacturing structure is constructed using a digital twin structure mapping method. Furthermore, a multi-physics joint modeling method is used to establish the coupling relationship between optics, electricity, and heat, thereby forming a parameter model that can characterize the state of the manufacturing stage and generating a multi-physics coupled parameter model.

[0116] S2: Based on the multi-physics coupling parameter model generated in S1, the deformation characteristics of the packaging material under thermal and mechanical loads are analyzed using the thermo-mechanical property analysis method to obtain the characteristic parameters of the packaging material. Based on the characteristic parameters of the packaging material, the optical path offset generated during the packaging process is corrected using the optical coupling offset compensation method. The stress distribution during the packaging stage is calculated using the thermo-mechanical stress assessment method, and the packaging parameters are dynamically adjusted using the Kalman filter compensation method to generate a packaging compensation parameter set.

[0117] S3: Based on the encapsulation compensation parameter set generated in S2, optical power, bit error rate, and eye diagram quality are tested using optical communication performance testing methods to obtain optical communication performance data. On this basis, storage performance testing methods are used to obtain read / write latency, bandwidth, and error count data for the storage module. A multi-source data time synchronization fusion method is used to synchronize the optical communication performance data and storage performance data, forming a unified dataset. For the synchronized dataset, feature normalization and principal component analysis methods are used to construct joint test features. Furthermore, support vector machines or deep neural network classification methods are used to analyze the joint test features to determine defect types. Combined with process parameter correlation reasoning methods, the source of defects is determined, generating defect tracing results.

[0118] S1 includes the following steps:

[0119] S101: Based on the real-time operation status of the optical communication storage module manufacturing production line, a real-time process parameter acquisition method is adopted. Through a multi-sensor synchronous acquisition mechanism, manufacturing process parameters such as mounting accuracy, welding temperature curve, and fiber alignment offset are acquired and time-calibrated to generate the original manufacturing process parameter set.

[0120] First, multiple types of sensors are deployed at key workstations on the production line to synchronously collect data on placement position offset, temperature change trends in the soldering area, and relative position changes of optical fibers. The data collection process is recorded according to a unified time rhythm. By performing time alignment processing on signals from different sources, errors caused by inconsistent collection starting points are eliminated. In specific implementation, for example, placement accuracy is continuously output by the vision sensing unit to measure position offset, the soldering process is generated by the temperature acquisition unit to form a continuous change curve, and the optical fiber alignment status is output by the displacement acquisition unit to measure the relative offset interval value. Before entering the buffer, all types of data are first compared with timestamps, and data within the same manufacturing cycle are grouped into the same set of records. For data segments with short-term missing data, the change trends of adjacent acquisition cycles are used to fill in the missing data. Finally, the multi-dimensional parameters obtained within the same manufacturing cycle are summarized in a unified format to form a set of manufacturing process parameters that can be directly called, serving as the basis for the original parameters required for subsequent processing.

[0121] S102: Based on the original manufacturing process parameter set, data cleaning and normalization methods are used to standardize the manufacturing process parameters through outlier removal, noise filtering and scale unification to generate a standardized manufacturing process parameter set.

[0122] First, a completeness check is performed on all parameters, and records that significantly exceed the normal physical range are marked as abnormal intervals. For example, when the placement offset is much higher than the normal fluctuation range, it is marked. Then, high-frequency jitter in continuous parameters is smoothed by comparing the change amplitude of adjacent sampling points and eliminating data segments with abrupt changes. Subsequently, the scale of parameters with different dimensions is unified, and position, temperature, and offset parameters are transformed into interval expressions that can be compared side by side. In the implementation example, the welding temperature curve is transformed into interval descriptions of heating segment, stabilization segment, and fallback segment, the placement accuracy is described as an offset level interval, and the fiber alignment status is described as a relative position interval. After the above processing, all parameters are expressed with unified rules, forming a data set with a consistent structure. This set is the standardized parameter result called in the subsequent modeling stage.

[0123] S103: Based on a standardized set of manufacturing process parameters, a digital twin structure mapping method is adopted. Through a one-to-one mapping mechanism between physical structure parameters and virtual model parameters, a virtual mapping model of the manufacturing structure is constructed, and a virtual structure mapping model is generated.

[0124] First, extract the physical unit information directly related to the parameters in the manufacturing structure, such as the positional relationship of the components, the connection method, and the spatial size range. Then, establish a corresponding structural description entry for each physical unit in the virtual environment. Through a one-to-one correspondence, the real-world structural state is mapped to the parameter state in the virtual model. During implementation, for example, the actual installation position range of a certain optical device is mapped to the position variable range in the virtual model, and the heat-affected zone of the welding area is mapped to the heat-affected area description in the virtual model. Each mapping relationship needs to be consistent when it is established to ensure that there are no missing or duplicate mappings in the virtual structure. Then, the standardized parameters are written into the corresponding virtual structural entries so that the virtual model can reflect the structural state changes within the same manufacturing cycle. Finally, a complete structural mapping description result is formed, which serves as the input object for subsequent multi-physics modeling.

[0125] S104: Based on the virtual structure mapping model, a multi-physics joint modeling method is adopted. Through the coupled modeling of optical, electrical and thermal parameters, the multi-physics coupling relationship in the manufacturing stage is calculated and modeled to generate a multi-physics coupling parameter model.

[0126] First, optical, electrical, and thermal parameters are introduced into the virtual structure. These parameters are then linked to their corresponding structural units according to their objects of action. The relationships between different physical attributes are then systematically analyzed, such as the relationship between changes in optical path and structural position ranges, and the relationship between electrical operating states and temperature range changes. These relationships are then jointly calculated and described in the virtual environment. By synchronously analyzing the changing trends of multiple parameters within the same structural unit, a description of multi-attribute interactions is formed. In the implementation example, when a device changes its position range, its thermal state range also adjusts accordingly. This relationship is recorded as a set of coupling relationship descriptions. After all structural units have undergone the above processing, a set of multi-attribute related parameters covering the manufacturing stage is summarized, resulting in a comprehensive model that characterizes multiple physical relationships.

[0127] S2 includes the following steps:

[0128] S201: Based on the multi-physics coupled parameter model, the thermo-mechanical property analysis method of the packaging material is adopted. The deformation characteristics of the packaging material under thermal and mechanical loads are analyzed by finite element calculation, and the property parameters of the packaging material are generated.

[0129] First, the temperature range and applied load range corresponding to different packaging areas in the model are read, and the packaging material is divided into several finite element regions. A basic attribute parameter library such as the elastic modulus range, thermal expansion coefficient range, and Poisson's ratio range of the material is established in the engineering simulation software environment. Then, thermal load range and mechanical load range from the coupled model are applied to each element. In a specific implementation example, the packaging shell of a certain optical communication storage module is selected as the object and divided into multiple continuous elements. Elements in the high temperature range are marked as the first-level temperature segment, and elements in the medium temperature range are marked as the second-level temperature segment. Then, the deformation of each element in the corresponding temperature segment is calculated and extrapolated according to the thermal expansion coefficient range of the material. At the same time, the deformation trend of the structure is superimposed and analyzed by combining the external clamping force range. During the analysis, the deformation results of different elements are compared. Elements whose deformation is above the preset deformation threshold range are marked as key areas. The deformation threshold is set according to the empirical range of allowable deformation of the material. By summarizing the results of multiple elements, a material property description data set containing the thermal deformation level and mechanical deformation level of each region is formed, and the packaging material property parameters used for subsequent steps are obtained.

[0130] S202: Based on the characteristic parameters of the packaging material, the optical coupling offset compensation method is adopted. The optical path offset generated during the packaging process is corrected by optical path offset inversion calculation, and optical path correction parameters are generated.

[0131] First, the deformation range of each region is correlated and matched with the optical path structure parameters. The theoretical optical axis position range between the light emitter and receiver in the packaging structure is used as the reference range. The structural offset range caused by material deformation is read and used as the initial input data for the optical path offset. In the implementation example, a certain packaging area has a displacement level of medium offset due to thermal expansion. Combined with the optical structure geometry, this displacement is projected onto the optical axis direction to form the optical axis offset range. Then, the theoretical optical path position is corrected by reverse calculation. During the correction process, the actual measured optical power attenuation range is compared with the theoretical coupling efficiency range. When the attenuation level is above the preset compensation threshold range, it is determined to be a state that needs compensation. This compensation threshold is set according to the statistical range of historical packaging batches. The offset range is mapped to the adjustment displacement range according to the proportional relationship to form the corresponding optical path position correction data set. Finally, the optical path correction parameters used for structural adjustment are output.

[0132] S203: Based on optical path correction parameters, a thermo-mechanical stress assessment method is adopted. The thermal and mechanical stresses in the packaging stage are comprehensively assessed through stress distribution calculation to generate packaging stress assessment results.

[0133] First, the thermal load range and mechanical constraint range of each structural unit in the packaging stage are extracted and input into the stress assessment module. The thermal stress range and mechanical stress range inside each unit are calculated and assessed separately. In specific implementation, the packaging shell and internal optical components are divided into several regions. The thermal stress level of the region in the high-temperature range is calculated, and the mechanical stress level of the pressure-clamped region is calculated. Then, the two types of stress ranges are superimposed and compared to set a comprehensive stress grading threshold range. When the comprehensive stress is in a higher level range, it is marked. The higher level is defined based on the empirical range of material yield strength. Then, all regions are screened, and regions that exceed the upper limit of the safe stress range are included in the key monitoring list. Finally, the stress assessment results of the packaging stage, which include the stress level distribution of each region, are summarized.

[0134] S204: Based on the packaging stress assessment results, the Kalman filter compensation method is adopted, and the key parameters in the packaging process are adaptively compensated through a dynamic parameter correction mechanism to generate a packaging compensation parameter set.

[0135] First, a mapping relationship is established between the stress level of each region and the corresponding structural parameters. Key parameters such as the encapsulation pressure setting range, heating power range, and positioning compensation displacement range are selected as adjustable variables. Historical encapsulation data from multiple batches are used as a reference sequence. In the dynamic processing module, the parameters of the current batch are compared with the historical average range. When a key parameter deviates from the upper or lower limit of the reference range, it is determined to be in a state that needs correction. Then, the current measured value and the predicted range are weighted and fused according to the Kalman filter approach. The weight range is set according to the fluctuation range of historical data. Parameters with large fluctuations are assigned a lower weight range, and stable parameters are assigned a higher weight range. The fused result is rewritten into the key parameter set, and the cyclic correction process continues for the next sampling period, finally forming an encapsulation compensation parameter set containing real-time correction values.

[0136] S3 includes the following steps:

[0137] S301: Based on the encapsulation compensation parameter set, an optical communication performance testing method is adopted. By testing optical power, bit error rate and eye diagram quality, optical communication performance data is obtained and optical communication performance test data is generated.

[0138] First, the displacement correction range, drive parameter correction range, and temperature control parameter correction range related to the optical channel structure are read from the compensation parameter set and used as the state input conditions before testing. In a specific implementation example, an optoelectronic module that has been packaged and loaded with compensation parameters is used as the test object. The module is connected to a standard optical communication test station, and the optical output end and receiver end are connected and fixed. During the test, the optical power change range, signal error statistics range, and eye diagram opening characteristic range are collected under multiple time slices. The optical power data is obtained by averaging the continuous sampling points over the range. The error data is obtained by converting the ratio of the number of error bits per unit time to the total number of bits. The eye diagram quality is determined by classifying the density of the sampled waveform distribution on the time axis and amplitude axis. When performing comparison, the current acquisition range is compared with the historical stable operation range as a benchmark. When a certain indicator falls into the deviation range, it is marked. By recording and summarizing multiple test indicators in parallel, a data set containing multi-dimensional optical communication test indicator range descriptions is formed, which is used as the output of optical communication performance test data.

[0139] S302: Based on optical communication performance test data, storage performance test methods are adopted. By testing read / write latency, bandwidth, and error count, storage performance data is obtained and storage performance test data is generated.

[0140] First, keeping the compensation parameters and operating status unchanged during optical communication testing, the module is switched to the storage performance testing station. Multiple rounds of read and write operations are initiated on the storage unit through the control interface. In the implementation example, for continuous write operations, the time span from the issuance of the command to the completion of data confirmation is recorded. This time span is used as the source of write latency data. Read latency intervals are obtained in the same way. For bandwidth data, the bandwidth interval description is obtained by statistically analyzing the total amount of data transmission completed within a fixed time window and proportionally converting it to the window duration. Error count data is obtained by accumulating the number of data inconsistencies that occur during multiple read and write verification processes and normalizing it according to the number of operations. When performing judgment, the latency interval is divided into three levels: low, medium, and high. The bandwidth interval is divided into three levels: insufficient, normal, and high. The error count interval is divided into two segments: acceptable and abnormal. The above interval divisions are set based on the historical statistical distribution of similar devices. Through the interval comparison and filtering of multiple rounds of test results, a structured storage performance test data set is formed.

[0141] S303: Based on optical communication performance test data and storage performance test data, a multi-source data time synchronization fusion method is adopted. The time axis alignment process is used to unify the multi-source test data and generate a synchronous test dataset.

[0142] First, the time stamp information of the two types of test data is extracted. The data generated by different test stations are reordered according to a unified time base. In the implementation example, by mapping the optical power sampling time period with the storage read and write operation time period, the data within the same operating cycle are divided into the same time window. On this basis, data segments with time offsets are interpolated or truncated to align the various types of data on the time axis. After alignment, the optical communication index interval and storage performance index interval within each time window are combined and encapsulated to generate a joint record entry containing multi-dimensional features. During the filtering process, time window data with missing key indicators or incomplete sampling are removed. The remaining data are summarized and organized in chronological order to finally form a time-consistent, multi-source unified data set, which is output as a synchronous test dataset.

[0143] S304: Based on the synchronous test dataset, feature normalization and principal component analysis are used to extract joint test features through feature dimensionality reduction and feature reconstruction, generating a joint test feature set;

[0144] First, the various feature indicators contained in the synchronous dataset are scaled uniformly to map feature values ​​of different dimensions to a unified range. In the implementation example, the original features are converted into dimensionless range descriptions by setting the maximum and minimum range values ​​of each feature. After normalization, correlation analysis is performed on the feature set, and features with highly similar trends are marked as the same feature group. For feature groups with high redundancy, only their representative features are retained. When performing dimensionality reduction, the feature subset with the contribution of each feature to the fluctuation in historical samples is selected by sorting the features. Then, the subset is reconstructed based on the selected features to form a joint feature description set that can simultaneously characterize the optical communication state and the storage state, which is output as the joint test feature set.

[0145] S305: Based on the joint test feature set, support vector machine or deep neural network classification method is used, combined with process parameter correlation reasoning method, to determine the defect type and its source, and generate defect tracing results.

[0146] First, the joint feature set is imported into the classification and determination module. Each feature vector is matched with the established feature template. In the implementation example, the similarity interval between the current feature vector and different defect category templates is calculated. The category with the high similarity interval is taken as the candidate result. At the same time, process parameter association information is introduced. The packaging temperature range, pressure range, and compensation parameter adjustment records of the corresponding batch are used as auxiliary judgment basis. When performing the judgment, when the feature matching result and the process parameter association result point to the same defect category, the category is directly recorded. When there is a discrepancy, a weighted comparison is performed based on the association weight interval between each process parameter and defect type in historical statistics. The category with the comprehensive weight in the advantageous interval is selected as the final judgment result. Then, the defect category is associated with the corresponding process stage and parameter interval to form the defect tracing result output.

[0147] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An integrated system for manufacturing, packaging, and testing optical communication storage modules, characterized in that, It includes the following modules: Manufacturing modeling module, which, based on the structural design data and manufacturing process parameters of optical communication storage modules, adopts a multi-physics joint modeling method and uses digital twin mapping to collaboratively model the mounting status of optical devices, optical waveguide alignment deviation and memory chip welding stress, and generates a manufacturing coupling model that characterizes the optical-electrical-storage coupling relationship in the manufacturing stage; The packaging collaboration module, based on the manufacturing coupling model, adopts the optical coupling compensation calculation method and uses the thermo-mechanical-optical path collaborative analysis method to dynamically evaluate and compensate for the material deformation, thermal stress distribution and optical path attenuation generated during the packaging process, and generates a set of packaging compensation parameters for packaging correction. The test fusion module, based on the encapsulation compensation parameter set, adopts a multi-dimensional joint test method and, through the synchronous execution mechanism of optical communication performance test and storage access performance test, performs fusion acquisition and feature construction on optical communication bit error rate, optical power stability and storage read / write latency, and generates a joint test feature set that comprehensively characterizes the module's performance status. The defect reasoning module, based on the joint test feature set, adopts a defect intelligent reasoning method and performs cross-stage correlation analysis and defect localization on optical communication performance anomalies, packaging stress mismatch and storage interface anomalies through process parameter knowledge correlation analysis, and generates corresponding defect tracing results. The optimization feedback module, based on the defect tracing results, adopts a closed-loop parameter optimization method and a manufacturing-packaging-testing parameter reverse adjustment mechanism to comprehensively optimize manufacturing process parameters, packaging compensation strategies, and test thresholds, generating a full-process optimization scheme for system self-iterative improvement.

2. The integrated system for manufacturing, packaging, and testing optical communication storage modules according to claim 1, characterized in that: The manufacturing modeling module includes a process acquisition submodule, a structure mapping submodule, a parameter modeling submodule, and a state simulation submodule. The process acquisition submodule uses manufacturing production line sensor information and adopts a real-time process parameter acquisition method to obtain mounting accuracy, welding temperature curve and fiber alignment offset data, and generates a manufacturing process dataset. The structure mapping submodule constructs a virtual manufacturing structure and generates a virtual structure mapping model based on the manufacturing process dataset using a digital twin structure mapping method. The parameter modeling submodule, based on the virtual structure mapping model, uses a multi-physics joint modeling method to construct the optical, electrical and thermal coupling relationship and generate a multi-physics coupled parameter model. The state simulation submodule, based on the multi-physics coupling parameter model, uses finite element simulation and ray tracing calculation methods to simulate the coupling state during the manufacturing stage and generate a manufacturing coupling model.

3. The integrated system for manufacturing, packaging, and testing optical communication storage modules according to claim 1, characterized in that: The encapsulation collaboration module includes a material analysis submodule, an optical path correction submodule, a stress assessment submodule, and a compensation calculation submodule. The material analysis submodule generates packaging material characteristic parameters based on the manufacturing coupling model and using a thermo-mechanical property analysis method for packaging materials. The optical path correction submodule generates optical path correction parameters based on the packaging material properties using an optical coupling offset compensation method. The stress assessment submodule generates packaging stress assessment results based on the optical path correction parameters using a thermo-mechanical stress assessment method. The compensation calculation submodule generates a set of packaging compensation parameters based on the packaging stress assessment results using an adaptive Kalman filter compensation method.

4. The integrated system for manufacturing, packaging, and testing optical communication storage modules according to claim 1, characterized in that: The test fusion module includes an optical parameter test submodule, a storage test submodule, a data synchronization submodule, and a feature construction submodule; The optical parameter testing submodule generates optical communication test data based on the encapsulation compensation parameter set and using optical communication performance testing methods. The storage test submodule generates storage performance test data based on the optical communication test data and using a storage access performance test method. The data synchronization submodule generates a synchronization test dataset based on the storage performance test data using a multi-source data time synchronization fusion method. The feature construction submodule generates a joint test feature set based on the synchronous test dataset using feature normalization and principal component analysis methods.

5. The integrated system for manufacturing, packaging, and testing optical communication storage modules according to claim 1, characterized in that: The defect reasoning module includes a feature analysis submodule, a knowledge association submodule, a defect classification submodule, and a location determination submodule. The feature analysis submodule generates anomaly feature vectors based on the joint test feature set using a multidimensional feature statistical analysis method. The knowledge association submodule generates a defect association graph based on the abnormal feature vector using a process parameter knowledge graph association reasoning method. The defect classification submodule generates defect type determination results based on the defect association map using support vector machine or deep neural network classification methods. The location determination submodule generates defect tracing results based on the defect type determination result using a cross-stage defect location method.

6. The integrated system for manufacturing, packaging, and testing optical communication storage modules according to claim 1, characterized in that: The optimization feedback module includes a parameter backtracking submodule, a strategy generation submodule, a process adjustment submodule, and a scheme output submodule. Based on the defect tracing results, the parameter backtracking submodule generates a set of key parameter deviations using a manufacturing-packaging parameter backtracking method. The strategy generation submodule generates an optimized strategy parameter set based on the key parameter deviation set using a model predictive control optimization method. The process adjustment submodule generates an optimized process configuration based on the optimization strategy parameter set using a closed-loop process parameter adjustment method. The output submodule generates a full-process optimization scheme based on the optimized process configuration.

7. An integrated method for manufacturing, packaging, and testing optical communication storage modules, characterized in that, Includes the following steps: S1: Based on the manufacturing production line data of optical communication storage modules, manufacturing process parameters such as mounting accuracy, welding temperature curve, and fiber alignment offset are acquired in real time using a process parameter acquisition method. The process parameters are then cleaned and normalized. On this basis, a virtual mapping model of the manufacturing structure is constructed using a digital twin structure mapping method. Furthermore, a multi-physics joint modeling method is used to establish the coupling relationship between optics, electricity, and heat, thereby forming a parameter model that can characterize the state of the manufacturing stage and generating a multi-physics coupled parameter model. S2: Based on the multi-physics coupling parameter model generated in S1, the deformation characteristics of the packaging material under thermal and mechanical loads are analyzed using the thermo-mechanical property analysis method to obtain the characteristic parameters of the packaging material. Based on the characteristic parameters of the packaging material, the optical path offset generated during the packaging process is corrected using the optical coupling offset compensation method. The stress distribution during the packaging stage is calculated using the thermo-mechanical stress assessment method, and the packaging parameters are dynamically adjusted using the Kalman filter compensation method to generate a packaging compensation parameter set. S3: Based on the encapsulation compensation parameter set generated by S2, optical power, bit error rate and eye diagram quality are tested using optical communication performance testing methods to obtain optical communication performance data; and on this basis, storage performance testing methods are used to obtain read / write latency, bandwidth and error count data of the storage module. A multi-source data time synchronization fusion method is used to synchronize optical communication performance data and storage performance data to form a unified dataset. For the synchronized dataset, feature normalization and principal component analysis are used to construct joint test features. Furthermore, support vector machine or deep neural network classification methods are used to analyze the joint test features to determine the defect type. Combined with process parameter correlation reasoning method, the defect source is determined to generate defect tracing results.

8. The integrated method for manufacturing, packaging, and testing an optical communication storage module according to claim 7, characterized in that: S1 includes the following steps: S101: Based on the real-time operation status of the optical communication storage module manufacturing production line, a real-time process parameter acquisition method is adopted. Through a multi-sensor synchronous acquisition mechanism, manufacturing process parameters such as mounting accuracy, welding temperature curve, and fiber alignment offset are acquired and time-calibrated to generate the original manufacturing process parameter set. S102: Based on the original manufacturing process parameter set, data cleaning and normalization methods are used to standardize the manufacturing process parameters by outlier removal, noise filtering and scale unification, and a standardized manufacturing process parameter set is generated. S103: Based on the standardized manufacturing process parameter set, a digital twin structure mapping method is adopted to construct a virtual mapping model of the manufacturing structure through a one-to-one mapping mechanism between physical structure parameters and virtual model parameters, and to generate a virtual structure mapping model. S104: Based on the virtual structure mapping model, a multi-physics joint modeling method is adopted. Through the coupled modeling of optical, electrical and thermal parameters, the multi-physics coupling relationship in the manufacturing stage is calculated and modeled to generate a multi-physics coupling parameter model.

9. The integrated method for manufacturing, packaging, and testing an optical communication storage module according to claim 7, characterized in that: S2 includes the following steps: S201: Based on the multi-physics coupled parameter model, the thermo-mechanical property analysis method of the packaging material is adopted. The deformation characteristics of the packaging material under thermal and mechanical loads are analyzed by finite element calculation, and the property parameters of the packaging material are generated. S202: Based on the characteristic parameters of the packaging material, the optical coupling offset compensation method is used to correct the optical path offset generated during the packaging process by inverting the optical path offset and generating optical path correction parameters. S203: Based on the optical path correction parameters, a thermo-mechanical stress assessment method is adopted to comprehensively assess the thermal and mechanical stresses during the packaging stage through stress distribution calculation, and generate packaging stress assessment results; S204: Based on the packaging stress assessment results, a Kalman filter compensation method is adopted to adaptively compensate key parameters in the packaging process through a dynamic parameter correction mechanism, thereby generating a packaging compensation parameter set.

10. The integrated method for manufacturing, packaging, and testing an optical communication storage module according to claim 7, characterized in that: S3 includes the following steps: S301: Based on the encapsulation compensation parameter set, an optical communication performance testing method is adopted. By testing optical power, bit error rate and eye diagram quality, optical communication performance data is obtained and optical communication performance test data is generated. S302: Based on the optical communication performance test data, a storage performance test method is adopted to obtain storage performance data by testing read / write latency, bandwidth and error count, and then generating storage performance test data. S303: Based on the optical communication performance test data and storage performance test data, a multi-source data time synchronization fusion method is adopted to unify the multi-source test data through time axis alignment processing, and generate a synchronous test dataset; S304: Based on the aforementioned synchronous test dataset, feature normalization and principal component analysis methods are used to extract joint test features through feature dimensionality reduction and feature reconstruction, generating a joint test feature set; S305: Based on the joint test feature set, a support vector machine or deep neural network classification method is used, combined with a process parameter correlation reasoning method, to determine the defect type and its source, and generate defect tracing results.

Images