Artificial intelligence assisted trace sample pretreatment data traceability quality control system and method

Abstract

This invention relates to an AI-assisted micro-sample preprocessing data traceability and quality control system and method, belonging to the technical field of data traceability and quality control. It includes: a construction module for collecting transient evolution features and constructing a preprocessing state folding field; a processing module for performing cross-process continuous cross-sectional slicing operations to generate a cross-sectional perturbation density spectrum; an analysis module for generating a latent variable embedding vector set and constructing a non-explicit causal topological skeleton; a calculation module for calculating topological convergence behavior and constructing a traceability reliability collapse index matrix; and a management module for performing storage topology reconstruction to generate a reversible traceability storage structure. This invention solves the technical problems in existing technologies where the micro-sample preprocessing process lacks multi-dimensional dynamic monitoring and deep perturbation traceability capabilities, leading to a black-box processing process, difficulty in tracing the root causes of deviations, and insufficient data integrity and reliability. It achieves the technical effect of improving the reliability, repeatability, and data integrity of micro-sample analysis results.

Description

Technical Field

[0001] This invention relates to the field of data traceability and quality control technology, specifically to an artificial intelligence-assisted micro-sample preprocessing data traceability and quality control system and method. Background Technology

[0002] Micro-sample pretreatment is a crucial step in life sciences, clinical diagnostics, precision medicine, and other fields. The process typically involves delicate operations such as lysis, enrichment, extraction, and amplification. Due to the small sample volume, extremely low target concentration, and complex interference background, the stability, repeatability, and reliability of the entire pretreatment process face severe challenges. Traditional pretreatment quality control mainly relies on the concentration and purity analysis of the final product and the monitoring of process parameters at discrete time points. This often only reflects the static results at a specific moment or endpoint, and cannot fully capture the complex disturbance behavior of dynamic, continuous, and multi-physics coupled at the microscale during pretreatment. Among these, transient disturbances such as local flow field shear anomalies, interfacial mass transfer fluctuations, and transient changes in the reaction microenvironment are the root causes of batch-to-batch differences, result deviations, and even experimental failures. Furthermore, how to extract effective information that can characterize the true state of the pretreatment process from high-dimensional, strongly coupled time-series data, and establish its correlation with the quality of the final processing results and data traceability, is a crucial part of current data traceability quality control. Existing data analysis lacks in-depth exploration of multi-field coupling effects and their cross-process transmission patterns, and has failed to establish a quantitative traceability chain from microscopic disturbances to macroscopic data credibility.

[0003] Therefore, current technologies suffer from a lack of multi-dimensional dynamic monitoring and deep disturbance tracing capabilities in the preprocessing of trace samples, resulting in a black box process, difficulty in tracing the root causes of deviations, and insufficient data integrity and reliability. Summary of the Invention

[0004] This application provides an AI-assisted traceability and quality control system and method for traceable micro-sample preprocessing data, which solves the technical problems in the prior art where the micro-sample preprocessing process lacks multi-dimensional dynamic monitoring and deep disturbance tracing capabilities, resulting in a black box process, difficulty in tracing the root cause of deviations, and insufficient data integrity and reliability. It achieves the technical effect of improving the reliability, repeatability, and data integrity of micro-analysis results.

[0005] This application provides an AI-assisted traceability and quality control system for micro-sample preprocessing data. The system includes: a construction module, used to collect transient evolution characteristics of coupled perturbations of microscale physical-chemical-rheological fields in the micro-sample preprocessing process involving lysis, enrichment, extraction, and amplification; and to construct a preprocessing state folding field with a time-recursive folding structure based on these transient evolution characteristics; and a processing module, used to perform cross-process continuous cross-sectional slicing operations on the preprocessing state folding field, introducing a self-constraint mechanism during the cross-sectional slicing process to generate a cross-sectional perturbation density spectrum that satisfies local topological conservation constraints. The analysis module is used to input the cross-sectional disturbance density spectrum into the latent source latent mapping model to generate a set of latent variable embedding vectors for the disturbance latent sources, and to construct a non-explicit causal topological skeleton for disturbance propagation based on the set of latent variable embedding vectors; the calculation module is used to calculate the topological convergence behavior of the disturbance propagation path in the high-dimensional collapse domain based on the non-explicit causal topological skeleton, and to construct a source tracing credible collapse index matrix; the management module is used to inject the source tracing credible collapse index matrix into the source tracing index control layer, drive the data lake storage engine to perform storage topology reconstruction based on topological convergence behavior, and generate a reversible source tracing storage structure.

[0006] In a possible implementation, the AI-assisted micro-sample preprocessing data traceability quality control system, in its analysis module, after inputting the cross-sectional perturbation density spectrum into the latent source mapping model, includes: activating a dual-channel latent space mapping structure, wherein the dual-channel latent space mapping structure includes a structure-preserving mapping channel based on perturbation morphology consistency and an energy-constrained mapping channel based on perturbation energy conservation constraints; using the structure-preserving mapping channel to perform nonlinear preservation projection on the local peak gradient distribution, frequency domain energy bandwidth distribution, and phase continuity characteristics of the cross-sectional perturbation density spectrum, generating a structure-constrained latent space feature matrix; using the energy-constrained mapping channel to perform multi-scale tensor quantization embedding on the time-frequency joint energy density distribution, amplitude decay curve, and time-series fluctuation entropy characteristics of the cross-sectional perturbation density spectrum, generating an energy-constrained latent space feature matrix; and performing cross-consistency regularization constraints on the structure-constrained latent space feature matrix and the energy-constrained latent space feature matrix to generate a latent variable embedding vector set.

[0007] In a possible implementation, the AI-assisted micro-sample preprocessing data traceability quality control system, in its analysis module, constructs a non-explicit causal topological skeleton for perturbation propagation based on the latent variable embedding vector set, including: performing calculations of the delayed mutual information tensor and the conditional transition entropy tensor between the latent variable embedding vectors in the latent variable vector set; constructing a latent causal influence matrix based on the delayed mutual information tensor and the conditional transition entropy tensor; and applying topological sparsity constraints and loop structure suppression constraints to the latent causal influence matrix to generate the non-explicit causal topological skeleton for perturbation propagation.

[0008] In a possible implementation, the AI-assisted micro-sample preprocessing data traceability quality control system, in its computing module, calculates the topological convergence behavior of the perturbation propagation path in a high-dimensional collapse domain based on the non-explicit causal topological skeleton, and constructs a traceability credible collapse index matrix. This includes: constructing a high-dimensional embedded state space for the perturbation propagation path based on the non-explicit causal topological skeleton, using the node connectivity tensor of the high-dimensional embedded state space as the initial constraint boundary of the high-dimensional collapse domain; performing multi-scale path folding mapping on the perturbation propagation path, jointly mapping the path length distribution, propagation delay distribution, and path weight decay curve to a trajectory compression representation in the high-dimensional collapse domain; jointly constraining the trajectory compression representation and the node connectivity tensor to generate a collapse potential field for the perturbation propagation path; calculating the topological convergence gradient distribution between perturbation propagation paths based on the collapse potential field, constructing a structural convergence matrix for the perturbation propagation path based on the topological convergence gradient distribution; and generating a traceability credible collapse index matrix using the topological consistency constraint relationship between the structural convergence matrix and the non-explicit causal topological skeleton.

[0009] In a possible implementation, the AI-assisted micro-sample preprocessing data traceability quality control system, in its computing module, utilizes the topological consistency constraint relationship between the structural convergence matrix and the non-explicit causal topological skeleton to generate a traceability credible collapse index matrix. This includes: constructing a topologically stable domain for the perturbation propagation path based on the collapse potential field; performing topological stability domain constraint pruning on the topological convergence gradient distribution to construct a constrained convergence gradient tensor; calibrating the path coupling weights of the convergence gradient tensor using the node connectivity tensor to generate a path credible convergence weight matrix; performing a joint tensor fusion operation on the path credible convergence weight matrix and the structural convergence matrix to generate a collapse credible factor tensor characterizing the structural stability of the perturbation propagation; and outputting the traceability credible collapse index matrix using the high-order topological consistency mapping relationship between the collapse credible factor tensor and the non-explicit causal topological skeleton.

[0010] In a possible implementation, the processing module of the AI-assisted micro-sample preprocessing data traceability quality control system performs cross-process continuous cross-sectional slicing operations on the preprocessing situation folding field, including: constructing a process evolution directed situation manifold based on the preprocessing situation folding field, embedding the situation evolution trajectories corresponding to the fragmentation, enrichment, extraction, and amplification processes into the process evolution directed situation manifold; performing continuous cross-sectional sampling operations based on the cross-process continuous hyperplane slicing structure on the process evolution directed situation manifold to generate a cross-sectional situation local structure tensor; extracting the perturbation amplitude distribution characteristics, temporal phase continuity characteristics, and perturbation direction consistency characteristics within the cross-section based on the cross-sectional situation local structure tensor, and embedding the extraction results into the cross-sectional constraint space corresponding to the cross-process continuous hyperplane slicing structure; performing cross-sectional correlation operations on the cross-sectional situation local structure tensor in the cross-sectional constraint space to generate an inter-sectional structure consistency description vector; and performing perturbation density normalization reconstruction of the cross-sectional situation local structure tensor based on the inter-sectional structure consistency description vector to generate a cross-sectional perturbation density spectrum that satisfies local topological conservation constraints.

[0011] In a possible implementation, the AI-assisted traceability quality control system for micro-sample preprocessing data further includes: an early warning module, used to perform threshold trigger analysis on the real-time updated value of the traceability credibility collapse index matrix, configure an early warning signal, and execute an early warning based on the early warning signal, wherein the early warning signal is used to indicate abnormal disturbance behavior occurring in the preprocessing process.

[0012] This application also provides an AI-assisted method for traceability and quality control of micro-sample preprocessing data. The method includes: during the preprocessing of micro-samples involving fragmentation, enrichment, extraction, and amplification, collecting transient evolution characteristics of the coupled perturbations of the microscale physical-chemical-rheological fields of the reaction sample; constructing a preprocessing state folding field with a time-recursive folding structure based on the transient evolution characteristics; performing cross-process continuous cross-sectional slicing operations on the preprocessing state folding field, introducing a self-constraint mechanism during the cross-sectional slicing process to generate a cross-sectional perturbation density spectrum that satisfies local topological conservation constraints; inputting the cross-sectional perturbation density spectrum into a latent source latent mapping model to generate a latent variable embedding vector set of perturbation latent sources; constructing a non-explicit causal topological skeleton for perturbation propagation based on the latent variable embedding vector set; calculating the topological convergence behavior of the perturbation propagation path in a high-dimensional collapse domain based on the non-explicit causal topological skeleton, and constructing a traceability credible collapse index matrix; injecting the traceability credible collapse index matrix into the traceability index control layer, driving the data lake storage engine to perform storage topology reconstruction based on topological convergence behavior, and generating a reversible traceability storage structure.

[0013] This application proposes an AI-assisted traceability and quality control system and method for micro-sample preprocessing data. The system comprises: a construction module for collecting transient evolution characteristics and constructing a preprocessing state folding field; a processing module for performing cross-process continuous cross-sectional slicing operations to generate cross-sectional perturbation density spectra; an analysis module for generating latent variable embedding vector sets and constructing an explicit causal topological skeleton; a calculation module for calculating the topological convergence behavior of perturbation propagation paths and constructing a traceability reliability collapse index matrix; and a management module for performing storage topology reconstruction to generate a reversible traceability storage structure. This system addresses the technical problems in existing technologies where the micro-sample preprocessing process lacks multi-dimensional dynamic monitoring and deep perturbation traceability capabilities, leading to a black-box processing process, difficulty in tracing the root causes of deviations, and insufficient data integrity and reliability. It achieves the technical effect of improving the reliability, repeatability, and data integrity of micro-sample analysis results. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0015] Figure 1 A schematic diagram of the structure of an AI-assisted micro-sample preprocessing data traceability and quality control system provided in this application embodiment.

[0016] Figure 2 This is a schematic diagram of the process for an AI-assisted micro-sample preprocessing data traceability and quality control method provided in an embodiment of this application.

[0017] Figure labeling: Module 10, Module 20, Module 30, Module 40, Module 50. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention in order to achieve the intended purpose, the following detailed description is provided in conjunction with the accompanying drawings and preferred embodiments, based on the specific implementation methods, structures, features and effects of the present invention.

[0019] This application provides an artificial intelligence-assisted traceability and quality control system for micro-sample preprocessing data, such as... Figure 1 As shown, the system includes:

[0020] The construction module 10 is used to collect transient evolution characteristics of the coupled perturbation of the microscale physical-chemical-rheological fields of the reaction sample during the pretreatment process of lysis, enrichment, extraction and amplification of micro-samples, and to construct a pretreatment situation folding field with a time recursive folding structure based on the transient evolution characteristics.

[0021] Preferably, the construction module performs multi-physics field synchronous monitoring based on micro-nano sensing and performs recursive deep encoding processing on high-dimensional temporal coupling signals to output a structured tensor that combines temporal dependence, multi-field coupling, and process spatiality. Specifically, through a micro-sensor array integrated into a microfluidic chip or reaction chamber, such as micro-pressure / temperature / conductivity sensors, micro-optical probes, and micro-shear force sensing units, the microscale physical-chemical-rheological parameters of the reaction samples in the pretreatment processes of pyrolysis, enrichment, extraction, and amplification are measured synchronously, continuously, and in situ. Among them, the physical field parameters include local temperature gradient, pressure fluctuation, flow rate / streamline distribution, and microbubble generation and collapse time series; the chemical field parameters include local pH transients, specific ion concentration gradients, redox potential fluctuations, and reaction interface chemical potential change rate; and the rheological field parameters include micro-region viscosity time-varying curves, shear thinning / thickening responses, thixotropic recovery rates, and frequency domain characteristics of viscoelastic modulus.

[0022] Preferably, the coupling effect of the three types of field parameters is calculated in real time through tensor coupling equations, outputting a multi-dimensional time-series signal set. This allows for the determination of the transient evolution characteristics of the three-field coupling perturbation, characterizing composite events such as "abnormal interfacial chemical mass transfer rate caused by local heating due to shearing". The multi-dimensional time-series signal set is then input into a deep feature encoder embedded with a time recursive unit to construct a high-dimensional tensor data structure. This structure undergoes recursive folding operations, including spatiotemporal convolution feature extraction, the introduction of cross-time step memory and state folding, and the output of a pre-processed situation folding field. Each time slice not only contains multi-field coupling information at the current moment but also encodes the situation evolution trajectory of multiple key time windows through the abnormal state encoding of the recursive unit, forming a dynamic feature map with temporal dependencies. This can be viewed as a continuous dynamic image of the entire pre-processing system in a high-dimensional feature space.

[0023] Processing module 20 is used to perform cross-process continuous cross-section slicing calculation on the preprocessed situation folding field, and introduce a self-constraint mechanism in the cross-section slicing process to generate a cross-section perturbation density spectrum that satisfies the local topological conservation constraint.

[0024] Furthermore, the specific configuration of the processing module 20 also includes: constructing a process evolution directed state manifold based on the preprocessed state folding field; embedding the state evolution trajectories corresponding to the fragmentation, enrichment, extraction, and amplification processes into the process evolution directed state manifold; performing continuous cross-section sampling operations based on a cross-process continuous hyperplane slice structure on the process evolution directed state manifold to generate a cross-section state local structure tensor; extracting the perturbation amplitude distribution characteristics, temporal phase continuity characteristics, and perturbation direction consistency characteristics within the cross-section based on the cross-section state local structure tensor, and embedding the extraction results into the cross-section constraint space corresponding to the cross-process continuous hyperplane slice structure; performing cross-section correlation operations on the cross-section state local structure tensor in the cross-section constraint space to generate an inter-section structure consistency description vector; and performing perturbation density normalization reconstruction of the cross-section state local structure tensor based on the inter-section structure consistency description vector to generate a cross-section perturbation density spectrum that satisfies local topological conservation constraints.

[0025] Preferably, the cross-process continuous cross-sectional slicing operation aims to extract a structured feature spectrum that can characterize the perturbation transmission law between processes from the dynamic situation folding field. Specifically, each sample processing process in the preprocessing situation folding field is formally defined as a directed curve in a high-dimensional feature space to determine the situation evolution trajectory, which is composed of situation points arranged in chronological order. The set of trajectories of all historical sample processing processes is regarded as a low-dimensional manifold embedded in the high-dimensional space, forming a process evolution directed situation manifold. The direction is given by the chronological order of the processing processes: fragmentation → enrichment → extraction → amplification. The situation evolution trajectories corresponding to the fragmentation, enrichment, extraction, and amplification processes are then embedded into the process evolution directed situation manifold. That is, on the process evolution directed situation manifold, fragmentation, enrichment, extraction, and amplification correspond to four continuous arc segments on the trajectory.

[0026] Preferably, in the high-dimensional space embedded in the process evolution directed state manifold, multiple continuously changing hyperplanes are defined and designed to be approximately orthogonal to the process evolution direction (i.e., the tangent direction of the trajectory). Then, continuous cross-section sampling operations based on the cross-process continuous hyperplane slicing structure are performed on the process evolution directed state manifold. That is, each hyperplane is used to cut the process evolution directed state manifold and its processing trajectory. Each time the cutting hyperplane intersects with the process evolution directed state manifold, a cross section is obtained, which contains the set of state points corresponding to different sample trajectories or the same trajectory in different running batches under the same processing progress. Then, local geometric analysis is performed on the set of state points on the cross section, such as calculating its local covariance matrix and fitting the tangent space to obtain the local structure tensor of the cross section state, which is used to describe the local distribution structure and variability of all possible states under a specific processing progress.

[0027] Preferably, feature decomposition or statistical analysis is performed on the local structural tensor of the cross-section situation corresponding to each cross-section to extract the perturbation amplitude distribution characteristics, temporal phase continuity characteristics, and perturbation direction consistency characteristics within the cross-section. Among them, the perturbation amplitude distribution characteristics refer to the distribution statistics of the Euclidean or Mahalanobis distances of each situation point within the cross-section from the ideal center trajectory, such as mean, variance, and skewness. The smoothness of the processing process is evaluated by calculating the displacement vectors or phase angles of corresponding situation points between adjacent cross-sections, and the presence of abrupt changes or lags in the temporal sequence is detected to obtain the temporal phase continuity characteristics. The alignment degree of each perturbation vector within the cross-section in the principal component direction is analyzed to determine whether the perturbation is isotropic noise or has a consistent deviation, thereby obtaining the perturbation direction consistency characteristics. The extracted results are embedded into the cross-section constraint space corresponding to the cross-process continuous hyperplane slice structure. One dimension of the cross-section constraint space is the processing progress, and the other dimensions correspond to three types of feature values, thereby compressing the local geometric information on the high-dimensional manifold into a semantically rich feature curve.

[0028] Preferably, cross-section correlation operations are performed on the local structural tensor of the cross-section situation in the cross-section constraint space to analyze the feature correlation between adjacent or nearby cross-sections along the processing progress direction. Specifically, mutual information, dynamic time warping, or cross-correlation analysis is used to calculate the similarity or causal correlation strength between the feature vector of cross-section i and the feature vector of cross-section j. For each cross-section, the pattern of its correlation strength with multiple cross-sections before and after it is calculated to generate a structural consistency description vector, which is used to describe the degree to which its local structural features are affected by the previous process under the current processing progress, and the degree to which it predicts the state of the subsequent process. A high value of the structural consistency description vector between cross-sections indicates that the cross-section is a stable joint for the state transfer between processes, while a low value of the structural consistency description vector between cross-sections indicates that there is a sudden change or decoupling in this stage.

[0029] Preferably, the perturbation density normalization reconstruction of the local structure tensor of the cross-sectional situation is performed. That is, the cross-sectional structural consistency description vector is used as weights to reconstruct the original local structure tensor of the cross-sectional situation. This strengthens the cross-sectional structural information that plays a key role in the inter-process transmission and weakens random fluctuations and irrelevant noise. Then, the reconstructed cross-sectional structural information is standardized and vectorized, and the output is the cross-sectional perturbation density spectrum. The cross-sectional perturbation density spectrum is a two-dimensional matrix. One dimension is the continuous processing progress, and the other dimension may be the energy of the principal components of the perturbation. The matrix value represents the normalized perturbation intensity / density. The local topology conservation constraint means that the local topology of the manifold is not destroyed during the cross-sectional sampling and reconstruction process. That is, when slicing the cross-section and extracting features, the original proximity relationship of the data points in the high-dimensional space must be respected. This is achieved by ensuring the local isometry or differential homeomorphism of the sampling and mapping process, thereby ensuring that the cross-sectional perturbation density spectrum truly reflects the local connectivity relationship of the original situation field.

[0030] Analysis module 30 is used to input the cross-sectional disturbance density spectrum into the latent source latent mapping model, generate a set of latent variable embedding vectors for the disturbance latent source, and construct a non-explicit causal topological skeleton for disturbance propagation based on the set of latent variable embedding vectors.

[0031] Furthermore, the specific configuration of the analysis module 30 also includes activating a dual-channel latent space mapping structure, which includes a structure-preserving mapping channel based on perturbation morphology consistency and an energy-constrained mapping channel based on perturbation energy conservation constraints; using the structure-preserving mapping channel to perform nonlinear preservation projection on the local peak gradient distribution, frequency domain energy bandwidth distribution, and phase continuity characteristics of the cross-sectional perturbation density spectrum to generate a structure-constrained latent space feature matrix; using the energy-constrained mapping channel to perform multi-scale tensor quantization embedding on the time-frequency joint energy density distribution, amplitude decay curve, and time-series fluctuation entropy characteristics of the cross-sectional perturbation density spectrum to generate an energy-constrained latent space feature matrix; and performing cross-consistency regularization constraints on the structure-constrained latent space feature matrix and the energy-constrained latent space feature matrix to generate a latent variable embedding vector set.

[0032] Preferably, the cross-sectional perturbation density spectrum is input into the hidden source latent mapping model. This model is a neural network model with a dual-branch parallel encoder structure, used to infer the unobservable hidden sources driving these perturbations from the cross-sectional perturbation density spectrum and encode them as physically meaningful low-dimensional latent variable vectors. This model can explain all perturbation patterns observed in the cross-sectional perturbation density spectrum, with each latent variable corresponding to a perturbation source, such as "heater transient fluctuations," "microvalve opening and closing jitter," or "micro-uniformity of reagent concentration." Specifically, a dual-channel latent space mapping structure is activated, including a structure-preserving mapping channel based on perturbation morphology consistency and an energy-constrained mapping channel based on perturbation energy conservation constraints. The methods follow different physical constraints for encoding, capturing the structural morphology and dynamic energy information of disturbances respectively, thereby more robustly decoupling the hidden source. Among them, the structure-preserving mapping channel is designed based on the consistency of the disturbance morphology at the hidden source level. That is, the morphological patterns of disturbances caused by the same hidden source at different process stages should have self-similarity, so that the encoded latent features can retain the key morphological characteristics of the original disturbance to the greatest extent. The energy-constrained mapping channel is designed based on the physical law that the energy of the disturbance generally follows the conservation relationship during the generation, transmission and dissipation process, so that the encoded latent features can strictly abide by the macroscopic conservation laws that the disturbance conforms to during the propagation process, such as energy conservation.

[0033] Preferably, an encoder neural network with a structure-preserving mapping channel is used to perform nonlinear preservation projection on the local peak gradient distribution, frequency domain energy bandwidth distribution, and phase continuity characteristics of the cross-sectional perturbation density spectrum. That is, through multiple nonlinear transformations such as convolutional layers and attention layers, the high-dimensional morphological features are compressed and projected into a structure-constrained latent space feature matrix. The key constraint is to introduce morphological reconstruction loss into the loss function to ensure that the spectrum decoded from the structure-constrained latent space feature matrix can faithfully reproduce the input morphological features. Among them, the local peak gradient distribution analyzes the steepness distribution of the rising and falling edges of each peak in the spectrum, reflecting the sharpness and suddenness of the perturbation; the frequency domain energy bandwidth distribution is to perform short-time Fourier transform or wavelet analysis on the spectrum to observe the concentration of energy in the frequency dimension, reflecting whether the perturbation is a single-frequency resonance or a broadband impact; the phase continuity characteristic analyzes whether the phase change of the same frequency component at different processing stages is smooth and continuous, and phase jumps usually indicate the switching of different perturbation sources or mechanisms.

[0034] Preferably, an encoder neural network with energy-constrained mapping channels is used to perform multi-scale tensor embedding on the time-frequency joint energy density distribution, amplitude decay curve, and time-series fluctuation entropy features of the cross-sectional perturbation density spectrum. This involves using multi-scale convolutional kernels or wavelet transform layers to simultaneously capture the performance of the three features at different time scales, organizing the multi-scale features into tensors, and finally mapping them to an energy-constrained latent space feature matrix. The constraint is the introduction of an energy conservation regularization term into the loss function to ensure that the energy changes predicted by the latent variables match the energy consumption calculated by the physical model. The time-frequency joint... Energy density distribution is the time-frequency energy distribution of the calculated spectrum, accurately tracing the joint diffusion path of energy in time and frequency, reflecting the spatiotemporal migration pattern of disturbance energy; amplitude decay curve is the envelope of the main disturbance mode extracted from the spectrum, analyzing the decay rate of its amplitude with time / process, such as exponential decay and power-law decay, reflecting the damping characteristics and disturbance dissipation mechanism of the system; temporal fluctuation entropy is the permutation entropy or approximate entropy of the spectrum sequence in the time dimension, quantifying the complexity and randomness of disturbance dynamics, with high or low entropy values ​​corresponding to deterministic mechanisms or random noise dominance.

[0035] Preferably, cross-consistency regularization constraints are applied to the structural constraint latent space feature matrix and the energy constraint latent space feature matrix. This involves minimizing the upper bound of their mutual information or maximizing their canonical correlation, while introducing adversarial training. This forces the structural preserving mapping channel and the energy constraint mapping channel to learn common latent source-related information and discard their respective channel-specific, irrelevant noise. This ensures that the latent source information contained in the latent feature matrices encoded by the structural preserving mapping channel and the energy constraint mapping channel is consistent. After fusion and filtering through cross-consistency constraints, the structural constraint latent space feature matrix and the energy constraint latent space feature matrix are aggregated and dimensionality reduced through a shared bottleneck layer. Finally, a set of latent variable embedding vectors is output, indicating that there are multiple independent latent sources, with each vector dimension representing a distributed representation of the decoupled latent source.

[0036] Furthermore, the specific configuration of the analysis module 30 also includes: performing the calculation of the delayed mutual information tensor and the conditional transition entropy tensor between the latent variable embedding vectors in the latent variable vector set; constructing a latent causal influence matrix based on the delayed mutual information tensor and the conditional transition entropy tensor; and performing topological sparsity constraints and loop structure suppression constraints on the latent causal influence matrix to generate a non-explicit causal topological skeleton for perturbation propagation.

[0037] Preferably, based on information theory and graph model optimization, a potential, nonlinear, and time-delayed causal influence network between latent sources is inferred from the latent variable vector set representing latent sources. Specifically, this involves determining whether the historical state of one latent source has a unique information contribution to the current state of another latent source that surpasses all other variables. Specifically, the nonlinear dependency strength of each pair of latent variable embedding vectors in the latent variable vector set under different time delays is calculated, and the delayed mutual information tensor is determined to characterize the strength of all possible lagged effects. Transfer entropy refers to the information gain about the current state of a latent source that can be obtained by knowing the past state of another latent source, given its own past state. The conditional transfer entropy tensor is calculated to eliminate the confusion effects of other latent variables. That is, it calculates the information contribution of the past state of another latent source to the current state of that latent source, given its own past state and the past states of all other latent variables. Similarly, the conditional transfer entropy is calculated for all pairs of latent variables and the delay that maximizes mutual information, forming a two-dimensional matrix to characterize the intensity of the directed direct information flow after eliminating confusion.

[0038] Preferably, the delayed mutual information tensor and the conditional transition entropy tensor are fused to construct a comprehensive latent causal influence matrix. The value of the latent causal influence matrix is ​​determined by the conditional transition entropy, representing the direct causal strength. The delayed mutual information tensor is used for verification weighting; only when significant mutual information exists at the optimal delay is the corresponding conditional transition entropy a reliable causal signal. Finally, the elements of the latent causal influence matrix represent the estimated causal influence strength of one latent source on another. Topological sparsity constraints are applied to the latent causal influence matrix, i.e., by adding regularization to a graph neural network, the sum of the absolute values ​​of the elements of the latent causal influence matrix is ​​minimized, forcing weak connections to zero and retaining only a few strong connections. Then, a cycle structure suppression constraint is applied, i.e., an acyclic constraint is introduced into the loss function, and this constraint is optimized to approach zero, thereby indirectly making the latent causal influence matrix an adjacency matrix of an acyclic graph. Finally, a directed graph model is constructed as the non-explicit causal topological skeleton for perturbation propagation, where graph nodes represent latent sources, and directed edges represent causal driving forces or significant influences between two latent sources.

[0039] The calculation module 40 is used to calculate the topological convergence behavior of the perturbation propagation path in the high-dimensional collapse domain based on the non-explicit causal topological skeleton, and to construct the source credible collapse index matrix.

[0040] Furthermore, the specific configuration of the computation module 40 also includes: constructing a high-dimensional embedded state space for the perturbation propagation path based on the non-explicit causal topological skeleton, using the node connectivity tensor of the high-dimensional embedded state space as the initial constraint boundary of the high-dimensional collapse domain; performing multi-scale path folding mapping on the perturbation propagation path, jointly mapping the path length distribution, propagation delay distribution, and path weight decay curve into a trajectory compression representation in the high-dimensional collapse domain; jointly constraining the trajectory compression representation and the node connectivity tensor to generate a collapse potential energy field for the perturbation propagation path; calculating the topological convergence gradient distribution between the perturbation propagation paths based on the collapse potential energy field, constructing a structural convergence matrix for the perturbation propagation path based on the topological convergence gradient distribution; and generating a source-tracing credible collapse index matrix using the topological consistency constraint relationship between the structural convergence matrix and the non-explicit causal topological skeleton.

[0041] Preferably, the state of each hidden source in the non-explicit causal topological skeleton, along with the influence of its causal neighbors, is embedded into a higher-dimensional vector space. That is, the hidden state of the graph neural network after multiple rounds of message passing is calculated for each node, or a node embedding algorithm is used to map the node and its structural role in the graph into a continuous vector. The embedding vectors of all nodes constitute a high-dimensional embedding state space, which is used to encode the functional state of the nodes and the network structural context. In the high-dimensional embedding state space, the causal connection strength between nodes is defined as a connectivity metric tensor to obtain the node connectivity tensor. For example, it is constructed as a weighted edge list in the embedding space as an induced constraint to define the allowed node state transitions and their difficulty. The node connectivity tensor of the high-dimensional embedding state space is used as the initial constraint boundary of the high-dimensional collapsed domain to specify the causal rules and strength that the perturbation propagation must follow.

[0042] Preferably, a multi-scale path folding mapping is performed on the perturbation propagation path. This involves analyzing the collective dynamics of all possible paths propagating along a causal network from any hidden source. Here, a perturbation propagation path refers to a sequence of paths on a non-explicit causal topological skeleton, starting from a source node, considering all lengths and possible endpoints, and propagating along directed edges to other nodes. For each propagation path, the path length distribution (i.e., the number of paths with different hop counts), the propagation delay distribution (i.e., the total time delay distribution of the perturbation propagation along the path), and the path weight decay curve (the product of the weights of each edge along the path, representing the degree of signal attenuation along that path) are calculated. Multi-scale path folding mapping uses an encoder model based on a recurrent neural network (RNN) to take multi-scale distribution features such as path length distribution, propagation delay distribution, and path weight decay curve as input, and jointly maps the output to a trajectory compression representation in a high-dimensional collapsed domain. This representation characterizes the overall behavioral pattern of all perturbation propagation paths originating from the source node, such as rapidly decaying local perturbations or global perturbations that can propagate over long distances. The high-dimensional collapsed domain is an abstract space used to compare and quantify different propagation patterns.

[0043] Preferably, the trajectory compression representation and the node connectivity tensor are jointly constrained by an energy function to generate a scalar field for the perturbation propagation path. This field describes the dynamics of the perturbation propagation tending towards a stable state or attractor in the embedded state space. The energy function represents the difference between the trajectory compression representation and the ideal stable propagation mode defined by the node connectivity tensor. For each point in the embedded state space or the collapsed domain, its energy value can be calculated to determine the collapse potential field. Regions with low energy correspond to states where the perturbation propagation mode converges, decays rapidly, and has limited influence; regions with high energy correspond to states where the propagation diverges, oscillates continuously, and has a far-reaching impact.

[0044] Preferably, the topological convergence gradient distribution between perturbation propagation paths is calculated based on the collapsed potential energy field, i.e., the state that evolves from the current state in the collapsed domain to reduce energy and tend towards stability. The direction and magnitude of the topological convergence gradient are used to define the "convergence direction" and "convergence driving force". Considering all hidden source node pairs, their trajectory representations and topological convergence gradients in the collapsed potential energy field are calculated, and the cosine similarity or cosine of the angle between the two topological convergence gradient vectors is calculated to obtain the structural convergence degree, which characterizes the propagation mode of perturbations starting from different hidden sources in dynamic convergence or divergence. The structural convergence degree is calculated for all node pairs to form a structural convergence matrix, whose diagonal elements are usually 1. The closer the off-diagonal elements are to 1, the more similar the perturbation propagation behavior of the hidden source pairs is, and the more likely they are to converge into the same system mode. Finally, the convergence metric is cross-validated with the original causal topology, i.e., the consistency between the structural convergence matrix and the non-explicit causal topology skeleton is checked, thereby generating the final credibility index. Based on the consistency principle, the structural convergence matrix and the non-explicit causal topology skeleton are fused to obtain the source tracing credibility collapse index matrix. Each element comprehensively reflects the existence of the causal influence structure of the two hidden sources and the stability, convergence, and predictability of the dynamic entropy. A high global source tracing credibility collapse index indicates that the entire disturbance propagation network structure is stable and the behavior is convergent, thus the entire preprocessing process has high controllability, the results have high credibility, and the source tracing clues are clear.

[0045] Furthermore, the specific configuration of the computation module 40 also includes: constructing a topologically stable domain for the perturbation propagation path based on the collapsed potential energy field; performing topologically stable domain constraint pruning on the topological convergence gradient distribution to construct a constrained convergence gradient tensor; calibrating the path coupling weights of the convergence gradient tensor using the node connectivity tensor to generate a path-credible convergence weight matrix; performing a joint tensor fusion operation on the path-credible convergence weight matrix and the structural convergence matrix to generate a collapse credibility factor tensor used to characterize the structural stability of the perturbation propagation; and outputting a source-tracing credibility collapse index matrix using the collapse credibility factor tensor and the high-order topological consistency mapping relationship of the non-explicit causal topological skeleton.

[0046] Preferably, in the collapsing potential energy field, not all low-energy regions are equally stable. By analyzing the positive definiteness of the second derivative matrix of the potential energy field or calculating its Morse exponent, the stable region is defined. The region that is a local minimum in energy and has an attracting basin characteristic in topological structure is used as the topological stable region for perturbation propagation path. Small perturbations at points within the topological stable region will not cause qualitative leaps in the propagation mode. The topological stable region constraint clipping is performed on the topological convergent gradient distribution, that is, each gradient vector is projected into the boundary of its stable region or its magnitude is directly restricted to a range that ensures that it does not jump out of the stable region, thus constructing a constrained convergent gradient tensor to ensure that the calculation is based on a physically realizable and structurally stable evolution direction.

[0047] Preferably, even if the propagation modes of two nodes show a convergence trend in the potential energy field, if they lack strong connections in the causal network or the connection paths do not conform to dynamics, the convergence may be false. The path coupling weights of the convergence gradient tensor are calibrated using the node connectivity tensor, i.e., the gradient convergence degree and structural connectivity are comprehensively calculated to determine the path reliability convergence weights, thus constructing a path reliability convergence weight matrix. This matrix is ​​used to quantify the degree of reliable convergence in the dynamics of perturbation propagation behavior of different hidden sources, while respecting the original causal structure. Then, a joint tensor fusion operation is performed on the path reliability convergence weight matrix and the structural convergence matrix, i.e., stacking the path reliability convergence weight matrix and the structural convergence matrix, followed by CP decomposition or Tucker decomposition to extract the common latent factors. These factors are then fused based on a cross-attention mechanism to generate a collapsed reliability factor tensor, used to characterize the overall structural stability of the perturbation propagation network. Finally, the stability pattern embodied in the collapsed credibility factor tensor is checked to see if it is consistent with the higher-order topological invariants contained in the non-explicit causal topological skeleton. These higher-order topological invariants include the principal eigenvalues ​​and eigenvectors of the graph, the spectral gaps of the graph, etc. Then, the consistency relationship between the collapsed credibility factor tensor and the higher-order features of the non-explicit causal topological skeleton is mapped out through a regression model. Finally, the source-tracing credibility collapse index matrix is ​​output, thereby ensuring that the stability assessment is consistent with the overall mathematical structure of the network, preventing local assessments from contradicting global properties, and making the final credibility index have extremely strong noise resistance.

[0048] The management module 50 is used to inject the traceability trust collapse index matrix into the traceability index control layer, drive the data lake storage engine to perform storage topology reconstruction based on topology convergence behavior, and generate a reversible traceability storage structure.

[0049] Preferably, the management module is a data management and physical storage implementation layer, used to transform traceability credibility metrics into a data storage architecture that supports efficient and intelligent traceability. Specifically, the traceability index control layer is an intelligent index management unit used to store and manage metadata indexes and rules for quickly locating, associating, and understanding data. Then, the traceability credibility collapse index matrix is ​​injected into the traceability index control layer as key quality and relation metadata and parsed into multiple index rules and storage strategy instructions. For example, a bidirectional pointer index is established for high-credibility causal paths, enabling traceability queries along these paths to be completed efficiently; the overall credibility level of the current processing is determined based on the sparsity pattern of the traceability credibility collapse index matrix, and the number of data replicas is determined, etc.

[0050] Preferably, topological convergence behavior refers to the convergence pattern of different perturbation propagation paths in high-dimensional space, driving the data lake storage engine to perform storage topology reconstruction. That is, the data lake storage engine receives instructions from the traceability index control layer and dynamically reorganizes the physical storage layout of the data according to the topological convergence behavior revealed by the traceability trust collapse index matrix. Specifically, this includes placing data columns that are tightly connected in causal topology and converge to the same pattern in dynamics on contiguous blocks of hard disks or the same erase block of SSDs to minimize I / O overhead during traceability queries; partitioning the data using stable communities or strongly convergent subgraphs identified in the traceability trust collapse index matrix as boundaries, with high data correlation within each partition and low correlation between partitions; and establishing special differential storage chains for the data versions associated with the key causal edges of the traceability trust collapse index matrix to optimize storage space to the extreme while ensuring reversible traceability. The final output supports a reversible source tracing storage structure with efficient bidirectional traversal. In forward tracing, given the final result, it can quickly locate and retrieve all upstream raw data that significantly contribute to it, such as temperature fluctuations in the pyrolysis stage and pressure anomalies in the extraction stage. The query path is guided by the high-weight causal links in the source tracing credibility collapse index matrix. In reverse attribution, given a source anomalous event, it can quickly assess and retrieve all downstream results and data that it may affect. The scope of its impact is defined by the propagation range defined by the source tracing credibility collapse index matrix. The logical associations of the data in the source tracing storage structure are directly encoded in the physical or logical structure of the storage.

[0051] Furthermore, the AI-assisted traceability quality control system for micro-sample preprocessing data also includes an early warning module, which is used to perform threshold trigger analysis on the real-time updated value of the traceability credibility collapse index matrix, configure an early warning signal, and execute an early warning based on the early warning signal. The early warning signal is used to indicate abnormal disturbance behavior that occurs in the preprocessing process.

[0052] Preferably, the early warning module is a real-time monitoring and proactive intervention terminal, used to transform static post-event analysis capabilities into dynamic in-event risk perception and alarm capabilities. The real-time updated value represents the stability and credibility status of the processed process reflected by the latest source tracing credibility collapse index matrix at the current moment. Preset multi-dimensional threshold rules are used to scan the real-time updated value in real time and perform threshold trigger analysis to determine whether it has triggered the early warning condition. The multi-dimensional threshold rules may include global stability threshold, critical path vulnerability threshold, and abnormal divergence detection. The main eigenvalue or matrix norm of the source tracing credibility collapse index matrix is ​​calculated to characterize the current convergence stability. If it is lower than the global stability threshold, it indicates that the performance of the entire processing network is declining. The module identifies the few causal paths with the highest weight in the source tracing credibility collapse index matrix and monitors their credibility index values. If they decrease significantly, it indicates that the critical transmission link is becoming unstable. The module monitors distant node pairs on the causal graph. If they increase abnormally, it indicates the emergence of new interference sources or abnormal couplings not covered by the existing causal model. Then, based on the results of threshold trigger analysis, structured early warning information is generated, including at least the early warning level, anomaly location, anomaly type, and confidence decay quantification value. The early warning signal is then configured to indicate abnormal disturbances in the pre-processing steps, and an early warning is issued based on the early warning signal. For example, in the control software interface, the abnormal process icon is highlighted and detailed information is displayed. Standardized alarm information is pushed to the laboratory operation and maintenance platform via API. In severe cases that trigger immediate intervention by the on-site audible and visual alarm devices, the complete early warning signal is recorded in the audit log, and automatic corrective measures are triggered to achieve early warning of hidden faults and ensure the operability of the alarm.

[0053] In the above text, refer to Figure 1 This paper describes in detail an artificial intelligence-assisted traceability and quality control system for micro-sample preprocessing according to an embodiment of the present invention. Next, reference will be made to... Figure 2 This invention describes an artificial intelligence-assisted method for traceability and quality control of micro-sample preprocessing data according to embodiments of the present invention. The artificial intelligence-assisted method for traceability and quality control of micro-sample preprocessing data, such as... Figure 2As shown, the method includes: during the preprocessing of micro-samples involving lysis, enrichment, extraction, and amplification, the transient evolution characteristics of the microscale physical-chemical-rheological three-field coupled perturbation of the reaction sample are collected; based on the transient evolution characteristics, a preprocessing state folding field with a time-recursive folding structure is constructed; cross-process continuous cross-sectional slicing operation is performed on the preprocessing state folding field, and a self-constraint mechanism is introduced during the cross-sectional slicing process to generate a cross-sectional perturbation density spectrum that satisfies local topological conservation constraints; the cross-sectional perturbation density spectrum is input into a latent source latent mapping model to generate a latent variable embedding vector set of perturbation latent sources, and a non-explicit causal topological skeleton of perturbation propagation is constructed based on the latent variable embedding vector set; based on the non-explicit causal topological skeleton, the topological convergence behavior of the perturbation propagation path in the high-dimensional collapse domain is calculated, and a source credible collapse index matrix is ​​constructed; the source credible collapse index matrix is ​​injected into the source credible collapse index control layer to drive the data lake storage engine to perform storage topology reconstruction based on topological convergence behavior, generating a reversible source credible storage structure.

[0054] In one possible implementation, the AI-assisted micro-sample preprocessing data traceability and quality control method further includes: activating a dual-channel latent space mapping structure, wherein the dual-channel latent space mapping structure includes a structure-preserving mapping channel based on perturbation morphology consistency and an energy-constrained mapping channel based on perturbation energy conservation constraints; using the structure-preserving mapping channel to perform nonlinear preservation projection on the local peak gradient distribution, frequency domain energy bandwidth distribution, and phase continuity characteristics of the cross-sectional perturbation density spectrum to generate a structure-constrained latent space feature matrix; using the energy-constrained mapping channel to perform multi-scale tensor quantization embedding on the time-frequency joint energy density distribution, amplitude decay curve, and time-series fluctuation entropy characteristics of the cross-sectional perturbation density spectrum to generate an energy-constrained latent space feature matrix; and performing cross-consistency regularization constraints on the structure-constrained latent space feature matrix and the energy-constrained latent space feature matrix to generate a latent variable embedding vector set.

[0055] In one possible implementation, the AI-assisted micro-sample preprocessing data traceability quality control method further includes: performing the calculation of delayed mutual information tensors and conditional transition entropy tensors between latent variable embedding vectors in the latent variable vector set; constructing a latent causal influence matrix based on the delayed mutual information tensor and conditional transition entropy tensor; and applying topological sparsity constraints and loop structure suppression constraints to the latent causal influence matrix to generate a non-explicit causal topological skeleton for perturbation propagation.

[0056] In one possible implementation, the AI-assisted micro-sample preprocessing data tracing and quality control method further includes: constructing a high-dimensional embedded state space for the perturbation propagation path based on the non-explicit causal topological skeleton, using the node connectivity tensor of the high-dimensional embedded state space as the initial constraint boundary of the high-dimensional collapse domain; performing multi-scale path folding mapping on the perturbation propagation path, jointly mapping the path length distribution, propagation delay distribution, and path weight decay curve into a trajectory compression representation in the high-dimensional collapse domain; jointly constraining the trajectory compression representation and the node connectivity tensor to generate a collapse potential energy field for the perturbation propagation path; calculating the topological convergence gradient distribution between the perturbation propagation paths based on the collapse potential energy field, constructing a structural convergence matrix for the perturbation propagation path based on the topological convergence gradient distribution; and generating a credible collapse index matrix for tracing using the topological consistency constraint relationship between the structural convergence matrix and the non-explicit causal topological skeleton.

[0057] In one possible implementation, the AI-assisted micro-sample preprocessing data tracing quality control method further includes: constructing a topologically stable domain for the perturbation propagation path based on the collapsed potential energy field; performing topologically stable domain constraint pruning on the topological convergence gradient distribution to construct a constrained convergence gradient tensor; calibrating the path coupling weights of the convergence gradient tensor using the node connectivity tensor to generate a path-credible convergence weight matrix; performing a joint tensor fusion operation on the path-credible convergence weight matrix and the structural convergence matrix to generate a collapsed credibility factor tensor for characterizing the structural stability of the perturbation propagation; and outputting a tracing credibility collapse index matrix using the collapsed credibility factor tensor and the high-order topological consistency mapping relationship of the non-explicit causal topological skeleton.

[0058] In one possible implementation, the AI-assisted micro-sample preprocessing data traceability and quality control method further includes: constructing a process evolution directed state manifold based on the preprocessing state folding field, embedding the state evolution trajectories corresponding to the fragmentation, enrichment, extraction, and amplification processes into the process evolution directed state manifold; performing continuous cross-section sampling operations based on a cross-process continuous hyperplane slice structure on the process evolution directed state manifold to generate a cross-section state local structure tensor; extracting the perturbation amplitude distribution characteristics, temporal phase continuity characteristics, and perturbation direction consistency characteristics within the cross-section based on the cross-section state local structure tensor, and embedding the extraction results into the cross-section constraint space corresponding to the cross-process continuous hyperplane slice structure; performing cross-section correlation operations on the cross-section state local structure tensor in the cross-section constraint space to generate an inter-section structure consistency description vector; and performing perturbation density normalization reconstruction of the cross-section state local structure tensor based on the inter-section structure consistency description vector to generate a cross-section perturbation density spectrum that satisfies local topological conservation constraints.

[0059] In one possible implementation, the AI-assisted traceability quality control method for micro-sample preprocessing data further includes: an early warning module, used to perform threshold trigger analysis on the real-time updated value of the traceability credibility collapse index matrix, configure an early warning signal, and execute an early warning based on the early warning signal, wherein the early warning signal is used to indicate abnormal disturbance behavior occurring in the preprocessing process.

[0060] The AI-assisted traceability and quality control system for micro-sample preprocessing data provided in this embodiment of the invention can execute the AI-assisted traceability and quality control method for micro-sample preprocessing data provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0061] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An AI-assisted traceability and quality control system for micro-sample preprocessing data, characterized in that, The system includes: A construction module is used to collect transient evolution characteristics of the coupled perturbations of the three fields of physical-chemical-rheological fields at the microscale of the reaction sample during the pretreatment process of lysis, enrichment, extraction and amplification of trace samples, and to construct a pretreatment situation folding field with a time recursive folding structure based on the transient evolution characteristics. The processing module is used to perform cross-process continuous cross-section slicing calculations on the preprocessed situation folding field, and introduce a self-constraint mechanism during the cross-section slicing process to generate a cross-section perturbation density spectrum that satisfies the local topological conservation constraint. The analysis module is used to input the cross-sectional disturbance density spectrum into the latent source latent mapping model, generate a set of latent variable embedding vectors for the disturbance latent source, and construct a non-explicit causal topological skeleton for disturbance propagation based on the set of latent variable embedding vectors. The calculation module is used to calculate the topological convergence behavior of the perturbation propagation path in the high-dimensional collapse domain based on the non-explicit causal topological skeleton, and to construct the source credible collapse index matrix. The management module is used to inject the traceability trust collapse index matrix into the traceability index control layer, drive the data lake storage engine to perform storage topology reconstruction based on topology convergence behavior, and generate a reversible traceability storage structure.

2. The AI-assisted traceability and quality control system for micro-sample preprocessing data as described in claim 1, characterized in that, The analysis module, after inputting the cross-sectional perturbation density spectrum into the latent source mapping model, includes: Activate the dual-channel latent space mapping structure, which includes a structure-preserving mapping channel based on perturbation morphology consistency and an energy-constrained mapping channel based on perturbation energy conservation constraints; The structure-preserving mapping channel is used to perform nonlinear preservation projection on the local peak gradient distribution, frequency domain energy bandwidth distribution, and phase continuity characteristics of the cross-sectional perturbation density spectrum to generate a structural constraint latent space feature matrix; The energy constraint mapping channel is used to perform multi-scale tensor quantization embedding on the time-frequency joint energy density distribution, amplitude decay curve and time-series fluctuation entropy features of the cross-sectional perturbation density spectrum to generate an energy constraint latent space feature matrix. Cross-consistency regularization constraints are applied to the structural constraint latent space feature matrix and the energy constraint latent space feature matrix to generate a latent variable embedding vector set.

3. The AI-assisted traceability and quality control system for micro-sample preprocessing data as described in claim 2, characterized in that, The analysis module constructs a non-explicit causal topological skeleton for perturbation propagation based on the latent variable embedding vector set, including: Calculate the delayed mutual information tensor and the conditional transition entropy tensor between latent variable embedding vectors in the latent variable vector set; Construct a hidden causal influence matrix based on the aforementioned delayed mutual information tensor and conditional transition entropy tensor; Topological sparsity constraints and loop structure suppression constraints are applied to the implicit causal influence matrix to generate a non-explicit causal topological skeleton for perturbation propagation.

4. The AI-assisted traceability and quality control system for micro-sample preprocessing data as described in claim 1, characterized in that, In the computation module, the topological convergence behavior of the perturbation propagation path in the high-dimensional collapse domain is calculated based on the non-explicit causal topological skeleton, and a source-tracing credible collapse index matrix is ​​constructed, including: A high-dimensional embedded state space for perturbation propagation paths is constructed based on the aforementioned non-explicit causal topological skeleton, and the node connectivity tensor of the high-dimensional embedded state space is used as the initial constraint boundary of the high-dimensional collapse domain. A multi-scale path folding mapping is performed on the disturbance propagation path to jointly map the path length distribution, propagation delay distribution and path weight decay curve into a trajectory compression representation in a high-dimensional collapsed domain. By jointly constraining the trajectory compression representation and the node connectivity tensor, a collapse potential field for the perturbation propagation path is generated. The topological convergence gradient distribution between the perturbation propagation paths is calculated based on the collapsed potential energy field, and the structural convergence matrix of the perturbation propagation paths is constructed based on the topological convergence gradient distribution. By utilizing the topological consistency constraint relationship between the structural convergence matrix and the non-explicit causal topological skeleton, a source tracing credible collapse index matrix is ​​generated.

5. The AI-assisted traceability and quality control system for micro-sample preprocessing data as described in claim 4, characterized in that, In the computation module, the topological consistency constraint relationship between the structural convergence matrix and the non-explicit causal topological skeleton is used to generate a source tracing credible collapse index matrix, including: Based on the collapsed potential energy field, a topologically stable domain for the perturbation propagation path is constructed. Topological stability domain constraint pruning is performed on the topological convergent gradient distribution to construct a constrained convergent gradient tensor. The path coupling weights of the convergence gradient tensor are calibrated using the node connectivity tensor to generate a path reliable convergence weight matrix. Perform a joint tensor fusion operation on the path confidence convergence weight matrix and the structure convergence matrix to generate a collapsed confidence factor tensor used to characterize the stability of the perturbation propagation structure; Using the collapse credibility factor tensor and the high-order topological consistency mapping relationship of the non-explicit causal topological skeleton, the source tracing credibility collapse index matrix is ​​output.

6. The AI-assisted traceability and quality control system for micro-sample preprocessing data as described in claim 1, characterized in that, The processing module performs cross-process continuous section slicing calculations on the preprocessed situation folding field, including: Based on the preprocessing state folding field, a process evolution directed state manifold is constructed, and the state evolution trajectories corresponding to the splitting, enrichment, extraction, and amplification processes are embedded in the process evolution directed state manifold; The process evolution directed state manifold performs continuous cross-section sampling operations based on a cross-process continuous hyperplane slicing structure to generate a cross-section state local structure tensor. Based on the local structural tensor of the cross-section, the characteristics of disturbance amplitude distribution, temporal phase continuity and disturbance direction consistency are extracted within the cross-section, and the extraction results are embedded into the cross-section constraint space corresponding to the cross-process continuous hyperplane slice structure. In the cross-section constraint space, perform cross-section correlation operation on the local structure tensor of the cross-section situation to generate a cross-section structural consistency description vector; Based on the cross-sectional structural consistency description vector, the perturbation density normalization reconstruction of the local structural tensor of the cross-sectional situation is performed to generate a cross-sectional perturbation density spectrum that satisfies the local topological conservation constraint.

7. The AI-assisted traceability and quality control system for micro-sample preprocessing data as described in claim 1, characterized in that, The system also includes: The early warning module is used to perform threshold trigger analysis on the real-time updated value of the traceability credibility collapse index matrix, configure early warning signals, and execute early warning output based on the early warning signals. The early warning signals are used to indicate abnormal disturbance behaviors that occur in the pre-processing process.

8. An artificial intelligence-assisted method for traceability and quality control of micro-sample preprocessing data, characterized in that, The method is applied to the AI-assisted traceability and quality control system for micro-sample preprocessing data as described in any one of claims 1-7, and the method includes: During the pretreatment process of lysis, enrichment, extraction and amplification of trace samples, the transient evolution characteristics of the microscale physical-chemical-rheological three-field coupled perturbation of the reaction sample are collected, and a pretreatment situation folding field with a time recursive folding structure is constructed based on the transient evolution characteristics. The preprocessed situation folding field is subjected to cross-process continuous cross-section slicing operation. A self-constraint mechanism is introduced during the cross-section slicing process to generate a cross-section perturbation density spectrum that satisfies the local topological conservation constraint. The cross-sectional perturbation density spectrum is input into the latent source latent mapping model to generate a set of latent variable embedding vectors for the perturbation latent source. Based on the set of latent variable embedding vectors, a non-explicit causal topological skeleton for perturbation propagation is constructed. Based on the aforementioned non-explicit causal topological skeleton, the topological convergence behavior of the perturbation propagation path in the high-dimensional collapse domain is calculated, and a source-tracing credible collapse index matrix is ​​constructed. The source traceability credibility collapse index matrix is ​​injected into the source traceability index control layer, driving the data lake storage engine to perform storage topology reconstruction based on topology convergence behavior, generating a reversible source traceability storage structure.

Images