Clindamycin phosphate crystallization tracing method and system based on block chain

Abstract

The invention discloses a clindamycin phosphate crystallization tracing method and system based on a block chain, and belongs to the technical field of medicine manufacturing process control. The method comprises the following steps: collecting multi-modal data of a crystallization process in real time through a Raman spectrum sensor, an FBRM sensor and other process analysis technology sensors; constructing a dynamic space-time atlas based on the data to characterize a process state; deeply analyzing the map by using a time-space diagram neural network model so as to identify key crystallization events such as nucleation and crystal transformation; the AI oracle machine automatically generates an event certificate containing event details, an original data hash value and a digital signature for the identified CCE; and finally, submitting the certificate as a transaction to a block chain network for uplink evidence storage, and forming a tamper-resistant tracing record. According to the method, the AI and block chain technologies are fused, so that the problem that key microscopic events are lack of credible and non-tampering records in the crystallization process is solved, and a high-fidelity tracing record with cryptographic guarantee is provided for the pharmaceutical production process.

Description

Technical Field

[0001] The present invention relates to the field of information transmission in the drug manufacturing process, and in particular to a blockchain-based clindamycin phosphate crystal traceability method and system. Background Art

[0002] Clindamycin phosphate is an important broad-spectrum antibiotic widely used clinically. In its production process, crystallization is a key unit operation for achieving separation, purification, and obtaining specific physicochemical properties (such as purity, particle size, and crystal form). Existing clindamycin phosphate crystallization methods typically rely on pre-defined process recipes. For example, the crude product is dissolved in a specific solvent system and crystallization is induced through a series of fixed steps, including decolorization, concentration, and step-wise cooling, in order to optimize macroeconomic indicators such as crystallization yield and product purity.

[0003] However, clindamycin phosphate exhibits polymorphism, with at least Form I, Form II, and Form III known to exist, each exhibiting distinct physical properties and stabilities. For example, Form I is generally considered the thermodynamically stable form, while the metastable Form II tends to transform into Form I over time. This polymorphism makes process control based solely on static formulation unreliable. Any minor fluctuation in production conditions, such as localized supersaturation changes, impurity level differences, or slight deviations in cooling rate, can lead to the undesired nucleation of a metastable form or unintended polymorphic transformations. While the final product may pass pre-shipment quality control testing, such as final X-ray powder diffraction (XRPD) analysis, which only confirms the final polymorphic form, it cannot demonstrate whether an unstable polymorph was transiently generated during the crystallization process, nor can it confirm whether this polymorphic transformation has left subtle defects within the crystal lattice that could affect the long-term stability of the drug. Therefore, the existing technology lacks a mechanism that can generate high-fidelity, reliable, and tamper-proof records to prove that a specific batch of products strictly follows the expected crystal path and crystallization kinetics trajectory. Within the framework of Quality by Design (QbD), the verifiability of the process is insufficient.

[0004] To address this challenge, Process Analytical Technology (PAT) has been introduced into pharmaceutical processes to enable real-time online monitoring. Commonly used PAT tools include Focused Beam Reflectance Measurement (FBRM) for online monitoring of particle count and chord length distribution, online Particle Vision and Measurement (PVM) for real-time acquisition of crystal morphology images, and Raman spectroscopy for in-situ measurement of solute concentration and crystal structure. However, these PAT tools generate massive, high-dimensional, multimodal data streams, making it difficult for human operators to fully integrate and understand their inherent complex correlations in real time. This results in the collection of large amounts of data but the difficulty in extracting effective information for process control.

[0005] In recent years, artificial intelligence (AI) technologies, such as recurrent neural networks (RNNs), have been used to analyze complex PAT data for process modeling and control. Furthermore, blockchain technology, due to its decentralized, data-immutable, and traceable nature, has gained application in pharmaceutical supply chain management. It is primarily used to track finished products from manufacturer to patient, preventing counterfeiting and ensuring logistics transparency.

[0006] While each of the aforementioned technologies has achieved some progress, significant technical limitations remain. First, there is a lack of an integrated system capable of intelligently integrating multimodal data from various PAT sensors, such as FBRM, PVM, and Raman spectroscopy, in real time to achieve a deep, dynamic understanding of the crystallization process (particularly the dynamic evolution of polymorphs). Second, existing AI applications typically store their analysis and judgment results in a centralized, modifiable, internal database. This lacks cryptographic security mechanisms to ensure the permanence, integrity, and non-repudiation of these key findings, creating trust issues when facing rigorous regulatory audits. Finally, existing blockchain applications are limited to macro-level logistics traceability after drug production is completed, failing to penetrate the core stages of active pharmaceutical ingredient production and providing a reliable traceability record of critical micro-level events during the crystallization process that directly determine final product quality. Summary of the Invention

[0007] One aspect of the present invention is to provide a blockchain-based clindamycin phosphate crystallization traceability method and system, aiming to address the existing technical problem of being unable to record and verify dynamic microscopic events in a critical production process, such as pharmaceutical crystallization, with high fidelity, reliability, and immutability. This invention addresses the lack of a method that can intelligently interpret real-time process data and create immutable, auditable, and cryptographically secure digital evidence of critical crystallization events (CCEs) during the crystallization process of materials such as clindamycin phosphate, which undergo complex polymorphic transitions.

[0008] To achieve the above objectives, the present invention provides a blockchain-based clindamycin phosphate crystal traceability method, which comprises the following steps: Step a) Multimodal process data acquisition. During the clindamycin phosphate crystallization process, multiple process analytical technology (PAT) sensors deployed on the crystallization reactor are used to synchronously collect multimodal process data in real time. The PAT sensors include at least a Raman spectrometer for acquiring crystal chemistry and crystal form information, a focused beam reflectometry (FBRM) probe for acquiring particle chord length distribution and number information, and an online particle image analyzer (PVM) for acquiring particle images and morphology information.

[0009] Step b) Constructing a dynamic spatiotemporal map. The multimodal process data collected in step a) is preprocessed and time-aligned. Based on this data, a dynamic spatiotemporal map is constructed to characterize the instantaneous state of the crystallization process. In this map, nodes correspond to individual PAT sensors and other process parameters (such as temperature and stirring rate). Node features are composed of normalized data collected by the corresponding sensor at a specific time point (such as Raman spectral vectors, FBRM chord length distribution histograms, and PVM image feature vectors).

[0010] Step c) AI model state analysis. The dynamic spatiotemporal graph is fed into a pre-trained spatiotemporal graph neural network (ST-GNN) model. The ST-GNN model analyzes the input graph data in real time to learn and capture the complex dependencies between different sensor data in spatial (sensor correlation) and temporal (process dynamic evolution) dimensions, thereby generating in-depth analysis of the current crystallization process state.

[0011] The contribution of the present application is that the spatio-temporal graph neural network model abstracts the PAT sensors and process parameters of different physical positions and functions as nodes of a graph, captures the spatial correlation of each variable at the same time by using a graph convolution network, and processes the evolution law of each node on the time series by using a recurrent neural network unit, so that the complex dynamics of the crystallization process caused by the synergistic action of multiple variables can be more deeply understood than traditional time series models.

[0012] Step d), critical crystallization event identification. The ST-GNN model monitors and identifies the occurrence of a pre-defined critical crystallization event (CCE) in real time according to the deep analysis results of step c). The CCE refers to a process event with a specific multimodal data feature signature that has a decisive influence on the quality of the final product, such as the start of primary nucleation, crystal form transformation, end of crystal growth, or process abnormal deviation, etc.

[0013] Step e), critical crystallization event certificate generation. Once a CCE is identified to occur and its confidence is higher than a pre-set threshold, an AI oracle module automatically generates a digital critical crystallization event certificate (CCE Certificate) containing detailed information of the event. The CCE certificate is structured data, at least including event unique identifier, drug batch number, event occurrence timestamp, event type, event related parameters, AI model judgment confidence score, key feature values used for judgment, cryptographic hash value of the original PAT data slice used for judgment, and AI oracle digital signature.

[0014] Step f), on-chain storage on the blockchain. The CCE certificate generated in step e) is submitted as a transaction to a blockchain network after being cryptographically signed. The blockchain network verifies the validity of the transaction through a consensus mechanism and permanently and tamper-proof records it on a distributed ledger, thereby establishing a verifiable event chain for the clindamycin phosphate crystallization process of the batch.

[0015] The contribution of the present application is that an AI oracle module is innovatively proposed as a bridge connecting the crystallization process in the physical world and the blockchain ledger in the digital world. The AI oracle is not only an artificial intelligence model, but also an automated and trusted event notarization unit that encapsulates the results of intelligent analysis and submits them on-chain, thereby solving the data integrity and non-repudiation problems faced by process analysis results stored in traditional centralized databases.

[0016] Preferably, step f) can be further improved in order to protect the business secrets while ensuring the process verifiability. After the AI Oracle generates the CCE certificate, instead of submitting the complete certificate content containing the specific process feature values (e.g. Evidentiary_Features field) directly on-chain, a cryptographic proof is generated using a Zero-Knowledge Proof (ZKP) protocol, such as zk-SNARKs (zero-knowledge succinct non-interactive arguments of knowledge). This proof can prove to the smart contract on the blockchain or the verifiers that “the AI Oracle indeed identified a certain CCE with a confidence higher than the threshold based on a set of PAT raw data that meets the preset specifications, through its legitimate ST-GNN model”, without revealing any specific process parameters or model judgment basis. Subsequently, this cryptographic proof instead of the complete CCE certificate is submitted as a transaction to the blockchain network for verification and recording. In this way, the compliance of the process can be verified without revealing the business secrets, thus ensuring the transparency of the process and the business secrets at the same time.

[0017] Further, the method of the present application also includes implementing adaptive process intervention using the smart contract deployed on the blockchain network. The process control rules are pre-encoded in the smart contract, and when a CCE indicating a process deviation (such as an unexpected polymorphic transformation) is received and recorded, the state of the smart contract is automatically updated and can trigger an off-chain operation. For example, if the smart contract records a CCE indicating an unexpected polymorphic transformation, it will automatically enter the “to be corrected” state and start a timer, and if a CCE indicating the success of the corrective measures submitted by the AI Oracle is not received within the preset time window, the smart contract can automatically trigger an off-chain action, such as sending a high-priority alarm to the quality control system or automatically locking the state of the production batch, through the Oracle or the application programming interface gateway. This builds an automated quality assurance closed-loop system based on blockchain consensus, which improves quality control from post-tracing to proactive intervention in the process.

[0018] Another aspect of the present invention is to provide a blockchain-based clindamycin phosphate crystallization traceability system for implementing the above method, the system comprising: a crystallization reactor for performing a crystallization operation of clindamycin phosphate; a PAT data acquisition module coupled to the crystallization reactor, comprising a Raman spectrometer, an FBRM probe, a PVM probe and other sensors for performing step a); an AI oracle server, communicatively connected to the PAT data acquisition module, wherein when the computer executable instructions stored therein are executed, the server is driven to perform steps b) to e); and a blockchain network, consisting of multiple distributed nodes, configured to receive transactions from the AI ​​oracle server and perform step f). BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a schematic flow chart of a blockchain-based clindamycin phosphate crystal traceability method provided by an embodiment of the present invention.

[0020] Figure 2 It is a structural diagram of a dynamic spatiotemporal graph in an embodiment of the present invention.

[0021] Figure 3 2 is a schematic diagram of the structure of the spatiotemporal graph neural network (ST-GNN) model in an embodiment of the present invention.

[0022] Figure 4 FIG. 4 is a schematic diagram of system hardware deployment in an embodiment of the present invention.

[0023] Figure 5 It is a schematic diagram of the overall layered architecture of the system in an embodiment of the present invention.

[0024] Figure 6 is an example diagram of multimodal PAT data in an embodiment of the present invention, wherein Figure 6 A is the Raman spectrum, Figure 6 B is the chord length distribution diagram of FBRM, Figure 6 C is the PVM crystal image. DETAILED DESCRIPTION

[0025] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0026] Example 1 This embodiment provides a method for tracing the origin of clindamycin phosphate crystals based on blockchain. Figure 1This method demonstrates the complete information flow and value chain from the crystallization process in the physical world to the blockchain ledger in the digital world. The method may include the following steps: Step S1, multimodal process data acquisition: During the crystallization process of clindamycin phosphate, multimodal process data is collected in real time and synchronously using multiple process analytical technology (PAT) sensors deployed on a 500 L crystallization reactor.

[0027] Specifically, the PAT sensor includes at least one in-situ Raman spectrometer, the probe of which is inserted into the liquid level of the crystallization kettle through a flange interface, and uses a 785nm excitation light source to collect Raman spectrum data every 60 seconds, with a spectral range covering 200-2000 nm. , to capture molecular vibrational information of clindamycin phosphate and the differences in characteristic peaks of different crystal forms. A focused beam reflectometry (FBRM) probe, mounted on the side of the crystallization reactor, measures the chord length distribution and particle count of crystals in solution in real time, with a data acquisition frequency of once every 10 seconds. An online particle image analyzer (PVM), whose probe is installed adjacent to the FBRM probe, captures a high-resolution microscopic image of the crystal morphology every 30 seconds. In addition, other key process parameters are simultaneously collected, such as the reactor temperature (accurate to 0.1°C) acquired by a Pt100 temperature sensor, the stirring rate (accurate to 1 RPM) obtained from the inverter feedback signal, and the pH meter reading. These multimodal data streams together provide a comprehensive description of the crystallization process status. This step, through the collection of multi-source heterogeneous data, provides a rich and comprehensive raw data foundation for subsequent in-depth analysis, addressing the limited dimensionality of traditional single-parameter monitoring and providing a prerequisite for accurate traceability.

[0028] Step S2, dynamic spatiotemporal map construction: The multimodal process data collected in step S1 are preprocessed and time-aligned, and based on this, a dynamic spatiotemporal map is constructed to characterize the instantaneous state of the crystallization process.

[0029] Reference Figure 2At each time point t, this graph can be represented as G_t = (V, E, X_t). The graph's node set V represents different information sources, with V = {v_raman, v_fbrm, v_pvm, v_temp, v_stir} corresponding to the Raman spectrometer, FBRM, PVM, temperature sensor, and stirrer, respectively. The node's feature matrix X_t consists of the normalized data collected by the corresponding sensor at time point t. For example, Raman spectral data is baseline-corrected and normalized to form a feature vector; FBRM chord length distribution data is divided into 100 bins, forming a 100-dimensional histogram vector; and PVM images are extracted using a pretrained convolutional neural network (such as ResNet-18) to generate a fixed-dimensional feature vector. The feature vectors of all nodes are aligned at timestamp t to form X_t. The graph's edge set E represents the physical or logical connections between nodes, and its adjacency matrix A is predefined. For example, there are strong correlations between temperature and Raman spectroscopy (which affects solubility) and between stirring rate and FBRM (which affects crystal fragmentation and agglomeration), resulting in edges between their corresponding nodes. This step integrates discrete, multi-source data streams into a unified, structurally rich mathematical object: a spatiotemporal graph. This representation not only preserves the inherent information of each data source but, more importantly, explicitly defines the interactions between different process variables, laying the foundation for subsequent model learning of complex synergistic effects.

[0030] Step S3, AI model state analysis: The dynamic spatiotemporal graph sequence is used as input and fed into a pre-trained spatiotemporal graph neural network (ST-GNN) model.

[0031] Reference Figure 3The ST-GNN model analyzes input graph data in real time to learn and capture the complex dependencies between different sensor data in spatial (sensor correlation) and temporal (dynamic process evolution) dimensions, thereby generating in-depth analysis of the current crystallization process state. The model structurally comprises alternating stacked spatial convolution modules and temporal convolution modules. Specifically, the spatial convolution module can employ a graph convolutional network (GCN) layer, the operation of which can be formalized as H_t' = GCN(A, X_t) = ReLU(A_hat X_t W_gcn), where A_hat is the normalized adjacency matrix and W_gcn is the learnable weight matrix. This operation aggregates information within the neighborhood of each node at each time step t, thereby capturing the instantaneous spatial correlation between variables. The temporal convolution module, which can employ a gated recurrent unit (GRU) or a long short-term memory (LSTM) network, processes the sequence of node representations {H_1', H_2', ..., H_t'} output by the spatial convolution module to capture the temporal evolution and long-term dependencies of each node's state. This ST-GNN model is trained through supervised learning on a large amount of historical production batch data, which includes key crystallization events manually annotated by process experts. The model is optimized by minimizing the cross-entropy loss between predicted results and true labels. This model architecture provides a deep understanding of the complex dynamics of the crystallization process, driven by the synergistic effects of multiple variables. For example, it can learn how a small drop in temperature first induces a supersaturation signal in Raman spectroscopy, which then manifests as a nucleation burst in FBRM data, and ultimately leads to crystal growth observed in PVM images. This causal chain captures this causal chain far beyond that of traditional stand-alone time series analysis models. As an alternative, the spatial convolution module can also employ a graph attention network (GAT), which dynamically learns the importance of different neighboring nodes, providing the model with improved interpretability and adaptability.

[0032] Step S4: Identification of critical crystallization events: The ST-GNN model monitors and identifies the occurrence of predefined critical crystallization events (CCEs) in real time based on the in-depth analysis results of step S3.

[0033] CCEs are defined as process events with specific multimodal data signatures that have a decisive impact on the quality of the final product. The predefined CCEs in this embodiment include at least: CCE-01: The onset of primary nucleation, characterized by the increase in the number of small-size particles (<20 μm) monitored by FBRM exceeding a preset slope threshold within a short period of time (e.g., within 1 minute), and the simultaneous decrease in the intensity of the solute characteristic peak in the Raman spectrum. CCE-02: The unexpected transition from Form I to Form II, characterized by the increase in the characteristic peaks of Form I in the Raman spectrum (e.g., 850 μm). ) intensity decreased, and the characteristic peaks of Form II (such as 875 ) intensity increases, and the rate of this transition exceeds the allowable range. CCE-03: Significant secondary nucleation or agglomeration is characterized by the observation in the PVM image of a large number of small crystals attached to the surface of large crystals or a significant increase in the proportion of crystal aggregates, accompanied by an abnormal increase in the long chord end counts of the FBRM. The final output layer of the ST-GNN model is a Softmax classifier, which outputs the probability of each type of CCE occurring at the current moment based on the learned high-level features. When the output probability of a certain type of CCE is higher than the preset confidence threshold (for example, 0.95), the CCE is determined to have occurred. This step transforms the black-box analysis of the model into a key event with clear guiding significance for the production process, providing a specific and actionable anchor point for subsequent traceability and intervention.

[0034] Step S5: Generate a Critical Crystallization Event Certificate. Once a CCE is identified and its confidence level is above a preset threshold, an AI Oracle module automatically generates a digital Critical Crystallization Event Certificate (CCE Certificate) containing detailed information about the event.

[0035] The AI ​​oracle is a trusted software service deployed on an edge computing server or a central server. It is responsible for executing the calculations of steps S2 to S5 and serves as a bridge connecting the physical process and the blockchain. The generated CCE certificate is structured data in JSON format. For example, its content is as follows: { "eventID": "evt_..._xyz", "batchID": "CP-20230510-01", "timestamp": "2023-05-10T10:30:15Z", "eventType": "CCE-02:Unwanted Polymorph Transition", "eventDescription": "Transition from Form I to Form II detected.", "confidenceScore": 0.98, "evidentiaryFeatures": {"raman_peak_ratio_875_850": 1.2, "fbrm_fine_particle_slope": 150}, "rawDataSource": {"type": "slice", "start_ts": "...", "end_ts": "..."}, "rawDataHash": "0xabc...def", "oracleSignature": "0x123...789"}. rawDataHash is the hash value obtained by performing a SHA-256 operation on the raw PAT data slice used for this judgment (for example, data from 2 minutes before and after the event), while oracleSignature is the digital signature of the entire certificate content (excluding the signature itself) applied by the AI ​​oracle using its own private key. This step solidifies a fleeting process event into a verifiable, complete digital certificate. The hash locks the original evidence, and the digital signature ensures the identity of the issuer and the integrity of the certificate.

[0036] Step S6: Store the certificate on the blockchain. The CCE certificate generated in step S5 is cryptographically signed and submitted to a blockchain network as a transaction.

[0037] In this embodiment, a permissioned blockchain (PBC) can be employed, such as one built on the Hyperledger Fabric framework, whose nodes are jointly maintained by the manufacturer, quality assurance department, and possibly regulatory agencies. The AI ​​oracle server, acting as a client node, uses the serialized CCE certificate as the transaction payload and invokes a specific chaincode (smart contract) function (such as `createCCECertificate`) to initiate a transaction. After the transaction is verified and executed by the endorsing node, it is submitted to the sorting service, ultimately packaged into blocks and distributed to all nodes, permanently and immutably recorded on the distributed ledger. This establishes a multi-party auditable chain of events for the crystallization process of this batch of clindamycin phosphate, comprised of multiple CCE certificates. Compared to traditional manual record-keeping or centralized database storage, this method significantly improves process verifiability and data integrity, providing cryptographically secure "process evidence" for drug quality.

[0038] To ensure process verifiability while protecting trade secrets, step S6 can be further improved. After the AI ​​oracle generates a CCE certificate, it does not directly submit the full certificate content, including specific process feature values ​​(such as the 'evidentiaryFeatures' field), to the blockchain. Instead, it utilizes a zero-knowledge proof (ZKP) protocol, such as zk-SNARKs. The AI ​​oracle uses the model's reasoning process and input features as private inputs and the event type and the fact that the confidence level exceeds a threshold as public outputs, generating a concise zero-knowledge proof. This proof verifies to the smart contract on the blockchain that "the AI ​​oracle has indeed identified a specific CCE with a confidence level above the threshold, based on a set of PAT raw data that meets pre-defined specifications and using its legitimate ST-GNN model," without revealing any specific process parameters or the basis for the model's judgment. Subsequently, only this cryptographic proof and some metadata (such as the event ID, timestamp, event type, and data hash) are submitted to the blockchain as a transaction. This approach enables trusted verification of process compliance while effectively protecting the company's core process secrets.

[0039] Furthermore, the method of the present invention also includes implementing adaptive process intervention using smart contracts deployed on a blockchain network. These smart contracts pre-encode process control rules. For example, one smart contract stipulates that upon receiving and recording a CCE-02 indicating an unexpected crystal form transition, the contract automatically marks the corresponding batch's status as "pending correction" and triggers an event. Upon detecting this event, an off-chain worker immediately sends a high-priority alert to the Manufacturing Execution System (MES) or operator interface. If the smart contract fails to receive a new CCE submitted by the AI ​​oracle indicating a successful corrective action (e.g., crystal form transition back to the target Form I) within a preset time window (e.g., 15 minutes), the smart contract automatically updates the batch's status to "quality abnormality, locked," prohibiting it from proceeding to the next production stage. This creates an automated closed-loop quality assurance system based on blockchain consensus, achieving a transition from passive traceability to proactive, real-time intervention.

[0040] Example 2 This embodiment provides a blockchain-based clindamycin phosphate crystal traceability system, which is the physical carrier of the method described in Example 1. The system includes: Crystallization reactor, used to perform crystallization operation of clindamycin phosphate.

[0041] The PAT data acquisition module is coupled to the crystallization reactor, including but not limited to a Raman spectrometer, an FBRM probe, a PVM probe, a temperature sensor, etc., and is used to perform step S1 in Example 1 and obtain multimodal process data in real time.

[0042] The AI ​​oracle server, which is one or more high-performance computing servers, is connected to the PAT data acquisition module via a wired or wireless network. The server is equipped with an operating system, a database, and a specific application. The computer-executable instructions contained in the application, when executed by the server's processor, drive the server to perform steps S2 to S5 of Example 1, namely, dynamic spatiotemporal graph construction, ST-GNN model parsing, CCE identification, and CCE certificate generation and signing. The server has sufficient computing power (e.g., a GPU for AI model inference) and a secure key storage mechanism.

[0043] A blockchain network consists of multiple geographically or logically distributed computing nodes that collectively maintain a distributed ledger. An AI oracle server acts as a client of this network and communicates with it. This network is equipped with a consensus mechanism (e.g., PBFT) and a smart contract execution environment. It receives transaction requests from the AI ​​oracle server, verifies transaction validity, and executes step S6 in Example 1, recording the CCE certificate or its zero-knowledge proof on the blockchain. It can also execute automated state changes or trigger off-chain operations based on pre-set rules.

[0044] Example 3 This embodiment will take a specific production batch as an example to elaborate on the application of the method of the present invention in actual production scenarios, especially how to use blockchain evidence storage and smart contracts integrated with zero-knowledge proof to achieve closed-loop quality assurance and process intervention. Suppose a pharmaceutical company uses the system described in the present invention to crystallize clindamycin phosphate with batch number "CLP-20240315-B02". In the early stage of production, the crystallization process runs smoothly according to the preset process parameters. The AI ​​oracle continues to parse the real-time data streams from the Raman spectrometer, FBRM and PVM. The CCE probabilities of each item output by the ST-GNN model are all lower than the alarm threshold, and the system is silently monitoring.

[0045] At 15:15 on the afternoon of March 15, 2024 Beijing time, due to an instantaneous voltage instability in the external power supply network, a slight fluctuation of about 90 seconds occurred in the refrigerant flow of the cooling jacket that controls the temperature of the crystallization kettle, causing the cooling rate in the local area of ​​the kettle to be slightly faster than the set value. This slight process deviation caused a transient increase in local supersaturation, inducing the nucleation of a small amount of thermodynamically metastable clindamycin phosphate crystal form II. This event is completely invisible to traditional processes that rely on endpoint detection. However, the system of the present invention captured its multimodal characteristic signature immediately after the event occurred. Specifically, at 15:16:30, the ST-GNN model in the AI ​​oracle server comprehensively analyzed the changes in the dynamic spatiotemporal map: First, the data from the Raman spectrometer node showed that at 850 representing the target crystal form I, The characteristic peak intensity at 875 The characteristic peak representing Form II began to appear faintly but clearly at the 10-30μm range. Almost simultaneously, data from the FBRM node revealed a steep increase in the count of fine particles in the 10-30μm range, which did not conform to the normal growth curve. By learning complex spatiotemporal correlations, the ST-GNN model accurately attributed these two seemingly isolated phenomena to a common root cause and identified the "CCE-02: Unexpected Polymorph Transition" event with a confidence level of 99.2%.

[0046] Next, within milliseconds of identifying the event, the AI ​​oracle module automatically executed optimized versions of steps e) and f). It first generated a CCE certificate containing complete event details. However, to protect core trade secrets such as the precise correlation model between crystal transformation rate and temperature fluctuation, it did not upload the certificate to the blockchain. Instead, it utilized the zk-SNARKs protocol to generate a concise zero-knowledge proof, just a few hundred bytes in size, asserting that the AI ​​oracle identified the specific CCE based on the raw PAT data it acquired and using a valid ST-GNN model. This proof was encapsulated in a transaction and submitted to the company's internal Hyperledger Fabric consortium blockchain network, along with non-sensitive metadata such as the event ID, batch number, timestamp, event type (CCE-02), and the hash value of the raw data.

[0047] After receiving the transaction, the on-chain batch management smart contract first verified the validity of the zk-SNARK proof. Once verified, the smart contract's pre-set logic was automatically triggered: the contract detected that CCE-02 was a critical deviation requiring immediate intervention. It then automatically updated the on-chain status of batch "CLP-20240315-B02" from "In Production" to "Pending Correction" and initiated a 30-minute correction countdown. Simultaneously, the smart contract, through an on-chain event mechanism, was captured by an off-chain worker and immediately sent a high-priority alert to the Manufacturing Execution System (MES) and the quality manager's mobile device via the API. The alert read: "Unexpected Form II formation detected in batch CLP-20240315-B02 at 15:16."

[0048] The on-site operator received the alarm at 3:22 PM and immediately followed the alarm instructions, executing the pre-set corrective procedure: briefly raising the reactor temperature by 2°C to dissolve the unstable Form II, maintaining the temperature for 10 minutes, and then resuming cooling at a slower rate (70% of the original rate). During this period, the system of the present invention continued to monitor. At 3:45 PM, the AI ​​oracle analyzed the latest PAT data and found that the Raman spectrum of 875 The characteristic peak of Form II has completely disappeared, 850 The intensity of the characteristic peak of Form I recovered and steadily increased. FBRM and PVM data also showed that crystal growth had returned to normal. Based on this, the ST-GNN model identified a new event, "CCE-07: Process Correction Successful," with a confidence level of 99.8%. The AI ​​oracle again generated a zero-knowledge proof for this event and uploaded it to the blockchain.

[0049] After the smart contract receives the new CCE-07 transaction and verifies successfully, the state of the batch is updated from "to be corrected" to "corrected_process normal", and the correction countdown is stopped. Finally, the batch product successfully completes the crystallization, and its blockchain traceability record shows the whole process of process deviation, intelligent early warning, manual intervention and successful correction in a complete and tamper-proof manner. When the batch of drugs faces regulatory audit or quality investigation in the future, the enterprise can present this on-chain record containing two cryptographically verified key events to reliably prove the rigor of its process control and the inherent quality of the product, embodying the concept of "quality resulting from design and process control".

[0050] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A clindamycin phosphate crystal traceability method based on blockchain, characterized in that: The following steps are involved: Step a), during the crystallization process of clindamycin phosphate, using multiple process analysis technology (PAT) sensors deployed on the crystallization reactor to synchronously collect multimodal process data in real time; Step b) preprocessing and time-aligning the multimodal process data, and constructing a dynamic spatiotemporal graph to characterize the instantaneous state of the crystallization process, wherein the nodes of the graph correspond to individual PAT sensors and other process parameters, and the features of the nodes are composed of data collected by the corresponding sensors at specific time points; Step c) feeding the dynamic spatiotemporal graph into a pre-trained spatiotemporal graph neural network (ST-GNN) model. The ST-GNN model analyzes the input graph data in real time to learn and capture the dependencies between different sensor data in spatial and temporal dimensions, generating an analysis of the current state of the crystallization process. Step d), the ST-GNN model monitors and identifies the occurrence of predefined critical crystallization events (CCEs) in real time based on the analysis results, wherein the critical crystallization events are events that characterize changes in the microscopic state of the clindamycin phosphate crystallization process and include at least one of primary nucleation, polymorphic transition, secondary nucleation, agglomeration, or process abnormal deviation; Step e) When a CCE is identified, the AI ​​oracle module automatically generates a digital key crystallization event certificate containing the event information. The CCE certificate contains at least the event type, the cryptographic hash value of the original PAT data slice used for judgment, and the digital signature of the AI ​​oracle. Step f) submitting the CCE certificate as a transaction to the blockchain network and recording it on the distributed ledger.

2. The method according to claim 1, characterized in that The multiple PAT sensors in step a) include at least: a Raman spectrometer for obtaining crystal chemistry and crystal form information, a focused beam reflectance measurement (FBRM) probe for obtaining particle chord length distribution and quantity information, and an online particle image analyzer (PVM) for obtaining particle images and morphology information.

3. The method according to claim 1, characterized in that In step b), in the dynamic spatiotemporal graph, the edge set of the graph represents the physical or logical association between nodes, and the spatiotemporal graph neural network ST-GNN model uses the graph structure defined by the edge set to capture the spatial association of each PAT sensor and process parameter at the same time.

4. The method according to claim 1, wherein The spatiotemporal graph neural network ST-GNN model in step c) structurally includes alternately stacked spatial convolution modules and temporal convolution modules, wherein the spatial convolution module is used to aggregate information within the neighborhood of each node to capture the instantaneous spatial correlation between variables, and the temporal convolution module is used to process the node representation sequence to capture the evolution of each node state over time.

5. The method according to claim 1, wherein The CCE certificate generated in step e) is structured data and also includes the event unique identifier, drug batch number, event occurrence timestamp, AI model judgment confidence score and key feature values ​​used for judgment.

6. The method according to claim 1, characterized in that The critical crystallization event (CCE) in step d) refers to a process event that has a decisive influence on the quality of the final product, including at least one of primary nucleation, crystal transformation, secondary nucleation, agglomeration or abnormal process deviation.

7. The method according to claim 1, characterized in that Said step f) further comprises: Before submitting the CCE certificate to the blockchain network, a cryptographic proof is generated based on the CCE certificate using a zero-knowledge proof protocol; The cryptographic proof, rather than the complete CCE certificate, is submitted as a transaction to the blockchain network for verification and recording, thereby proving the occurrence of CCE without revealing the specific process parameters contained in the certificate.

8. The method according to claim 1, characterized in that The method further includes: Adaptive process intervention is achieved using a smart contract deployed on the blockchain network. The process control rules are pre-encoded in the smart contract. When a CCE indicating a process deviation is received and recorded, the smart contract automatically triggers an off-chain operation.

9. The method according to claim 1, characterized in that The blockchain network in step f) is a consortium chain, and its nodes are jointly maintained by the manufacturing enterprise, quality assurance department or regulatory agency.

10. A clindamycin phosphate crystal traceability system based on blockchain, characterized in that: include: A PAT data acquisition module, coupled to the crystallization reactor, is used to synchronously acquire multimodal process data during the crystallization process of clindamycin phosphate in real time; an AI oracle server, in communication with the PAT data acquisition module, configured to: construct the multimodal process data into a dynamic spatiotemporal graph; and parse the dynamic spatiotemporal graph using a spatiotemporal graph neural network (ST-GNN) model to identify a critical crystallization event (CCE) comprising at least one of primary nucleation, polymorphic transition, secondary nucleation, or agglomeration; After identifying the CCE, it generates a key crystallization event certificate containing the cryptographic hash value of the original PAT data slice and its own digital signature; and a blockchain network, composed of a plurality of distributed nodes, for receiving a transaction containing the CCE certificate from the AI ​​oracle server and recording the transaction on a distributed ledger.

Images