Liquid preparation production system of oxytocin injection

Abstract

The invention relates to the technical field of production scheduling control, in particular to a liquid preparation production system of oxytocin injection, which comprises a multi-working-condition coupling modeling module for simulating the flowing and mixing conditions of fluid in a liquid preparation tank and generating a process scheduling instruction; the temperature-deviation adjusting module predicts the oxytocin degradation rate through an LSTM network and compensates for refrigeration power deviation; the illumination regulation and control module generates a light-shielding operation protocol based on a photosensitive feedback regulation strategy and in combination with an illumination accumulated dose so as to reduce the risk of light degradation; the pH value dynamic adjustment module constructs a buffer capacity prediction model, and adjusts the driving frequency of a metering pump through a fuzzy PID controller, thereby achieving the precise control of the pH value. And the PLC controls the collaborative decision-making module to calculate a QRI index by adopting a dynamic weight distribution strategy, perform nonlinear compensation on excess parameters, and generate a fault tree set to realize risk prediction and optimization control. The stability, uniformity and reliability of the liquid preparation process are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of production scheduling control, and specifically to a liquid preparation production system for oxytocin injection. Background Art

[0002] The production of oxytocin injection involves multiple processes such as raw material weighing, dissolution, liquid preparation, filtration, sterilization, filling, and sealing. In the existing automated liquid preparation production system, although programmed control technologies based on programmable logic controllers (PLCs) and distributed control systems (DCSs) are adopted, there are still technical bottlenecks in the deep integration of industrial Internet of Things and advanced process control technologies.

[0003] Firstly, there are limitations in the dynamic ratio precision control of highly active drug ingredients. The existing system does not construct a dynamic compensation module based on mass flow meters and multivariable PID controllers. Especially during continuous feeding, the PLC system cannot adaptively adjust the motion curve of the servo motor through real-time viscosity sensor data, resulting in the accumulation of volume measurement deviations of temperature-sensitive raw materials and affecting the uniformity of the final product content. Secondly, the multi-parameter coupling control mechanism is not yet perfect. When synchronously regulating key quality attributes such as pH value, osmotic pressure, and aseptic conditions, interactive interference between control loops is likely to occur, leading to process parameter drift beyond the scope specified in the pharmacopoeia. Thirdly, the adaptability of the existing program control architecture to complex production scenarios is insufficient. Especially when dealing with the switching of different specification batches, there is a lack of self-learning optimization algorithms based on real-time quality data, making it difficult to achieve dynamic correction of process parameters and intelligent matching of equipment operation modes. In addition, the integrity and real-time performance of the production data traceability system are defective, failing to fully meet the requirements of Good Manufacturing Practice (GMP) for continuous monitoring and audit tracking of key process parameters, and affecting the quality risk control efficiency. The above technical defects restrict the robustness and quality control level of the oxytocin injection production process and urgently need to be solved by an innovative system.

[0004] Therefore, a liquid preparation production system for oxytocin injection is proposed. Summary of the Invention

[0005] The object of the present invention is to provide a liquid preparation production system for oxytocin injection to improve the stability, uniformity and reliability of the liquid preparation process. Through a multi-condition coupling modeling module, the fluid flow and mixing conditions in the liquid preparation tank are simulated to generate process scheduling instructions; the temperature-deviation adjustment module uses an LSTM network to predict the oxytocin degradation rate and compensate for the deviation of the refrigeration power; the light control module generates a light-shielding operation protocol based on a photosensitive feedback adjustment strategy and in combination with the cumulative light dose to reduce the risk of photodegradation; the pH value dynamic adjustment module constructs a buffer capacity prediction model and adjusts the driving frequency of the metering pump through a fuzzy PID controller to achieve precise control of the pH value. The PLC control collaborative decision-making module calculates the QRI index using a dynamic weight allocation strategy, performs non-linear compensation on the excess parameters, and generates a fault tree set for risk prediction and optimal control.

[0006] To achieve the above object, the present invention provides the following technical solutions: A liquid preparation production system for oxytocin injection, comprising: A multi-condition coupling modeling module, which is used to simulate the fluid flow state of oxytocin and the material mixing situation in the liquid preparation tank based on a fluid dynamics model, and construct process scheduling instructions; calculate the uniformity index under the process scheduling instructions based on time-dependent simulation, and update the process scheduling instructions; A temperature-deviation adjustment module, which is used to predict the oxytocin degradation rate according to temperature time-series data using a long short-term memory network and compensate for the deviation of the refrigeration power of the process scheduling instructions; A light control module, which is used to calculate the illuminance distribution in the liquid preparation tank using a photosensitive feedback adjustment strategy and generate a light-shielding operation protocol in combination with the cumulative light dose; A pH value dynamic adjustment module, which is used to establish a buffer capacity prediction model to calculate the buffer capacity and adjust the driving frequency of the metering pump through a fuzzy PID controller; A PLC control collaborative decision-making module, which is used to calculate the QRI index using a dynamic weight allocation strategy according to temperature control parameters, pH value, the cumulative light dose and liquid preparation time, generate a fault tree set, and perform non-linear compensation on the excess parameters.

[0007] Further, the construction process of the process scheduling instructions includes: Introduce A turbulence model to simulate the stirring turbulence behavior, and use the Navier-Stokes equation to establish a fluid flow state model of oxytocin to calculate the velocity field and pressure field in the liquid preparation tank; use the convection-diffusion equation to model the material mixing situation to calculate the concentration distribution; Based on the finite volume method, the liquid preparation tank is divided into grid cells, and local grid refinement is performed in the stirring area and the feed port area; use an adaptive grid adjustment strategy to calculate the division error and update the grid cells; Based on the velocity field, the pressure field, and the concentration distribution, an integral equation is established for each grid cell based on the Gauss-Green theorem, and the SIMPLE algorithm is used to solve the velocity-pressure coupling, iteratively updating the velocity field, the pressure field, and the concentration distribution to generate the process scheduling instruction.

[0008] Further, the update process of the process scheduling instruction includes: At each time step, calculate the homogeneity index , expressed as: ; where is the concentration at grid cell , is the average concentration of the liquid preparation tank, is the actual mixing time, is the theoretical mixing time, is the dimensionless weight coefficient, is the total number of grid cells; If is less than the preset threshold, use the genetic algorithm to adjust the feeding order and update the process scheduling instruction.

[0009] Further, the implementation process of the genetic algorithm includes: Use the permutation coding method to set the oxytocin material order, and generate the initial population according to the oxytocin material order and the process scheduling instruction; Set maximizing the homogeneity index as the fitness function, and adopt roulette wheel selection operation, partially matched crossover, and swap mutation. When the maximum number of iterations is reached, output the updated process scheduling instruction.

[0010] Further, the process of generating the light avoidance operation instruction includes: Deploy virtual light sensors on the grid cells, and use the Beer-Lambert law to calculate the light attenuation to obtain the illuminance distribution; Multiply the illuminance distribution, the exposure time, and the photodegradation coefficient of the grid cell to obtain the light exposure cumulative dose; For each grid cell, compare the light exposure cumulative dose with the safe light threshold. If the light exposure cumulative dose is greater than the safe light threshold, generate the light avoidance operation instruction; otherwise, no additional operation; where the light avoidance operation instruction includes local light avoidance and global light avoidance.

[0011] Further, the process of adjusting the driving frequency of the metering pump includes: Collect the pH value in the liquid preparation tank through a pH sensor and record the amount of alkali input; Calculate the buffer capacity using the Henderson-Hasselbalch equation based on the pH value and the amount of alkali; Input the pH error, the buffer capacity, and the error change rate into the fuzzy PID controller; Design fuzzy rules based on expert experience, and the fuzzy PID controller outputs PID parameters according to the fuzzy rules; Calculate the control signal according to the PID parameters and convert the control signal into the driving frequency of the metering pump.

[0012] Further, the process of generating and analyzing the fault tree set includes: Define the QRI index exceeding the standard as the top event of liquid preparation, which is used to characterize the deviation of the QRI index from the final fault manifestation during the liquid preparation process; Define the temperature exceeding the standard, abnormal pH value, excessive light cumulative dose, and liquid preparation time exceeding the limit as intermediate events of liquid preparation, which are used to describe the key factors affecting the top event of liquid preparation; Define equipment failure, sensor error, and regulation strategy failure as basic events of liquid preparation, which are used to describe the physical factors and system factors at the bottom layer that cause the intermediate events of liquid preparation; Connect the top event of liquid preparation, the intermediate events of liquid preparation, and the basic events of liquid preparation using logic gates to generate the fault tree set; Through qualitative analysis, determine the key path of liquid preparation and the minimum cut set of liquid preparation; through quantitative analysis, calculate the probability of the top event of liquid preparation and identify the key risk points.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. Through multi-condition coupling modeling, the present invention uses computational fluid dynamics to simulate the flow state and material mixing situation of oxytocin fluid in the liquid preparation tank, and can accurately predict the fluid behavior. Based on the SIMPLE algorithm, velocity-pressure coupling is solved, and the process scheduling instructions are dynamically optimized. At the same time, the feeding sequence is adaptively adjusted by combining the genetic algorithm, improving the uniformity index, ensuring the stability and consistency of the liquid preparation process, and thus reducing the efficacy fluctuation of oxytocin injection caused by uneven mixing.

[0014] 2. The present invention uses a long short-term memory network to predict the degradation rate of oxytocin, dynamically compensates the refrigeration power, optimizes the temperature control strategy, can accurately regulate the temperature change during the liquid preparation process, and ensures the stable preparation of the liquid medicine in a suitable temperature environment. In addition, through the photosensitive feedback adjustment strategy, the illuminance distribution in the liquid preparation tank is calculated, and the light avoidance operation instruction is generated based on the cumulative light dose. Combining the virtual light sensor and the Beer-Lambert law to simulate the light attenuation, the local and global light avoidance regulation is realized, the risk of photodegradation is reduced, and the quality stability of the oxytocin injection is improved.

[0015] 3. The present invention calculates the QRI index through the dynamic weight task strategy, and combines the nonlinear compensation technology to optimize the excess parameters, improving the stability of the liquid preparation process of the oxytocin injection. The fault tree set analysis method is adopted to construct a fault tree set model with the QRI index exceeding the standard, identify key risk factors such as temperature, pH value, cumulative light dose, and liquid preparation time, and calculate the probability of the top event of the liquid preparation. Through qualitative and quantitative analysis, the liquid preparation process parameters are optimized, the system safety and fault diagnosis ability are improved, and thus the reliability of the oxytocin injection production process is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a schematic structural diagram of a liquid preparation production system for an oxytocin injection provided by the present invention; Figure 2 is a schematic flow diagram for generating and analyzing a fault tree set of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] Please refer to Figures 1 to 2 , the present invention provides a liquid preparation production system for an oxytocin injection, and the technical solution is as follows: Embodiment 1: Oxytocin is an important bioactive polypeptide drug, which is widely used in obstetric clinics to promote uterine contractions. However, the oxytocin injection is easily affected by factors such as temperature, pH value, light, and feeding order during the liquid preparation process, resulting in a decrease in biological activity, a decline in the stability of the preparation, and difficulty in ensuring production consistency.

[0019] At present, traditional liquid preparation systems mainly rely on fixed feeding sequences and empirical parameters for regulation, making it difficult to adapt to various working conditions. In addition, existing temperature control methods do not fully consider the degradation characteristics of oxytocin, the pH value adjustment lacks accurate prediction, and the management of the light environment is also relatively rough, resulting in fluctuations in the quality of the liquid medicine and affecting the safety and effectiveness of the final product. Therefore, it is necessary to develop an intelligent, multi-variable coupled control liquid preparation production system for oxytocin injection to optimize the feeding strategy, improve product stability, and ensure the consistency of production quality.

[0020] In order to optimize the feeding strategy and improve the stability of product quality, as Figure 1 shown, a liquid preparation production system for oxytocin injection is used, including: A multi-condition coupling modeling module with reference to Figure 1 is used to simulate the fluid flow state and material mixing situation of oxytocin in the liquid preparation tank based on the fluid dynamics model, and construct process scheduling instructions; calculate the uniformity index under the process scheduling instructions based on time-dependent simulation, and update the process scheduling parameters.

[0021] Among them, the process scheduling instruction refers to the specific material feeding sequence formulated according to the optimization goal during the liquid preparation process, which is used to guide key links such as equipment operation and time arrangement.

[0022] Furthermore, the construction process of the process scheduling instruction includes: Introduce a turbulence model to simulate the stirring turbulence behavior. This model calculates the turbulent viscosity by solving the transport equations of turbulent kinetic energy and dissipation rate , so as to describe the influence of turbulence on the flow field. Then, the turbulent viscosity is introduced into the Navier-Stokes equation to establish a fluid flow state model of oxytocin, which is used to calculate the velocity field and pressure field in the liquid preparation tank, expressed as: ; Among them, is the fluid density, is the dynamic viscosity of the fluid, is the external force, is the gradient operator, is the time variable, is the partial derivative symbol.

[0023] Use the convection-diffusion equation to model the material mixing situation, which is used to calculate the concentration distribution , expressed as: ; Among them, is the diffusion coefficient, and the convection term Velocity field solved based on the Navier-Stokes equation , realizing the coupling of the flow field and the concentration field.

[0024] For numerical solution, the liquid mixing tank is divided into grid cells based on the finite volume method, and local grid refinement is performed in the stirring area and the feeding port area to improve the calculation accuracy of key areas; an adaptive grid adjustment strategy is used to calculate the division error and update the grid cells.

[0025] Among them, the present invention adopts an adaptive grid adjustment strategy, effectively solving the limitations of static grids in traditional fluid simulations. Traditional methods often result in insufficient accuracy or computational redundancy due to their inability to adapt to the dynamic changes in flow characteristics. For this reason, the present invention introduces the concept of flow gradient, which reflects the rate of change of fluid physical quantities (such as velocity, pressure, and concentration, etc.) in space. Common flow gradients include velocity gradient and pressure gradient, etc. Based on the flow gradient or residual, the system can dynamically adjust the grid density. For example, in areas with complex flows such as near the agitator, the grid resolution is increased by 2 times to capture fine flow characteristics. For areas that need to be refined, by subdividing the existing grids, larger grid cells are divided into multiple smaller cells, thereby improving local accuracy. On the contrary, in the stable area far from stirring, the number of grids can be reduced by 50%. By merging adjacent grid cells (such as merging four small square grids into a larger square, or eight cubic grids into a larger cube), the computational burden is reduced. This dynamic adjustment strategy significantly improves efficiency while ensuring computational accuracy. Experiments show that compared with static grids, the calculation time is reduced by about 22.35%, achieving a balance between accuracy and efficiency and providing a better solution for complex fluid simulations.

[0026] After the grid cell division is completed, based on the velocity field, the pressure field, and the concentration distribution, an integral equation is established for each grid cell based on the Gauss-Green theorem, expressed as: ; Among them, is a conserved quantity (such as velocity or concentration distribution ), is the operation of integrating the conserved quantity over the volume , is the flux, representing the flow rate of the conserved quantity through a unit area, is the unit normal vector, is the operation of performing an area integral of the flux over the surface of the grid cell, is the source term, representing the factor that generates or consumes the conserved quantity inside the grid cell , is the operation of integrating the source term over the volume of the grid cell .

[0027] To solve the coupling relationship between the velocity field and the pressure field , the SIMPLE algorithm is used for velocity-pressure coupling solution, and the velocity field, the pressure field and the concentration distribution are iteratively updated until the convergence condition is satisfied (such as the residual is less than ), forming a closed-loop coupling solution to generate the process scheduling instruction.

[0028] Among them, based on the obtained flow field (velocity field and pressure field ) and concentration field (concentration distribution ), the material mixing uniformity is analyzed. For example, if the concentration deviation is less than 5%, it is considered that the mixing is uniform. If the requirement is not met, the process scheduling instruction is readjusted, such as adjusting the stirring speed or changing the feeding order.

[0029] By accurately simulating fluid flow and material mixing, it is ensured that the concentration distribution of oxytocin injection is uniform, reducing the local concentration deviation, thereby improving the product quality. Based on the real-time calculation results, the system can dynamically adjust the stirring speed and feeding strategy. At the same time, the adaptive grid adjustment strategy reduces the number of grids in the steady area, effectively reducing the computational resource requirements. In addition, the SIMPLE algorithm and iterative solution ensure that a stable numerical solution can still be obtained under complex turbulent conditions, improving the system reliability.

[0030] Furthermore, the update process of the process scheduling instruction includes: At each time step of the liquid preparation production, the system obtains the concentration distribution of each grid cell in the liquid preparation tank through numerical simulation and calculates the uniformity index to quantify the degree of material mixing uniformity. At each time step, the uniformity index under the process scheduling instruction is calculated, expressed as: ; where is the concentration at the grid cell , is the average concentration of the liquid preparation tank, is the actual mixing time, is the theoretical mixing time (predicted based on experience or model, for example, set to 10 minutes), is the dimensionless weight coefficient used to adjust the influence of time on The influence (e.g., ), is the total number of grid cells; Set a preset threshold, e.g., 0.95, representing the goal of mixing uniformity. If is less than the preset threshold, use a genetic algorithm to adjust the feeding order and update the process scheduling instruction.

[0031] The update process of the process scheduling instruction ensures a uniform distribution of the material concentration in the dispensing tank and reduces the local concentration deviation by calculating in real time and dynamically adjusting the feeding order, thereby improving the quality of oxytocin injection. The application of the genetic algorithm avoids the inefficiency of traditional trial and error, quickly finds the optimal feeding order, and the system automatically adjusts the process parameters based on real-time data to form a closed-loop control, which can adapt to the dynamic changes in the production process and improve the system robustness and flexibility. In addition, the UI formula comprehensively evaluates the concentration deviation and mixing time, and through the flexible adjustment of the weight coefficient ensures the accuracy and scientificity of the mixing effect evaluation.

[0032] Furthermore, a genetic algorithm is used to optimize the feeding order of oxytocin materials to maximize the uniformity index , and the implementation process of the genetic algorithm includes: Set the oxytocin material order using permutation encoding. For example, if there are three materials A, B, and C, the feeding order can be represented as [A, B, C], that is, first feed material A, then feed material B, and finally feed material C. According to the oxytocin material order and the process scheduling instruction (such as [A, C, B]), randomly generate a group of feeding orders as the initial population.

[0033] Set maximizing the uniformity index as the fitness function, The higher it is, the better the mixing uniformity. And use roulette wheel selection operation, partially matched crossover, and swap mutation. When the maximum number of iterations is reached, e.g., 100 generations, find the best feeding order [B, C, A], whose is the largest, and output the updated process scheduling instruction [B, C, A]. The genetic algorithm successfully optimizes the feeding order of oxytocin materials through steps such as permutation encoding, initial population generation, fitness function, roulette wheel selection, partially matched crossover, and swap mutation. The system uses these steps for iterative optimization and can dynamically adapt to changes in the production process, such as fluctuations in material properties, so as to maintain the optimization effect and ensure that the feeding order of oxytocin injection is always in the optimal state.

[0034] Refer to Figure 1 's temperature-deviation adjustment module, which is used to predict the oxytocin degradation rate using a long short-term memory network based on temperature time series data and compensate for the refrigeration power deviation of the process scheduling instruction; Specifically, by deploying temperature sensors at multiple key positions in the liquid dispensing tank, temperature time-series data is collected in real-time. After being collected by the PLC controller, these data are stored in the database, and at the same time, the corresponding actual oxytocin degradation rate is recorded to form a training dataset. Using this dataset to construct a long short-term memory network model, the temperature time-series data of the past n seconds is input into the network to predict the future temperature change trend and the future oxytocin degradation rate. By comparing the actual oxytocin degradation rate with the future oxytocin degradation rate, the loss is calculated using the mean square error to optimize the model parameters. Then, the current temperature time-series data is input into the trained LSTM model to predict the oxytocin degradation rate within the future time interval The oxytocin degradation rate within is predicted. Based on the prediction result, the PID control algorithm is used to calculate the cooling power deviation and output a new cooling power instruction. The PLC controller adjusts the cooling system accordingly to dynamically increase or decrease the cooling power to ensure the temperature stability during the liquid dispensing process, thereby reducing the risk of oxytocin degradation and improving the stability of the liquid medicine quality.

[0035] Reference Figure 1 The light control module is used to calculate the illuminance distribution in the liquid dispensing tank using a photosensitive feedback regulation strategy and generate a light avoidance operation protocol in combination with the light cumulative dose; Furthermore, the process of generating the light avoidance operation instruction includes: Deploy virtual light sensors on the grid cells, use the Beer-Lambert law to calculate the light attenuation, and obtain the illuminance distribution, expressed as: ; Among them, Is the illuminance distribution at depth , Is the incident light intensity, Is the absorption coefficient of the oxytocin liquid, Is the natural logarithm base.

[0036] Multiply the illuminance distribution, exposure time, and photodegradation coefficient of the grid cell to obtain the light cumulative dose , expressed as: ; Among them, Is the exposure time of this grid cell, Is the photodegradation coefficient of oxytocin.

[0037] Sample and calculate the light cumulative dose at different time points and accumulate it to the final cumulative light dose. Set a safety light threshold. For each grid cell, compare the light cumulative dose with the safety light threshold. If the light cumulative dose is greater than the safety light threshold, generate the light avoidance operation instruction; otherwise, no additional operation; The photosensitive feedback regulation strategy is a control method that dynamically adjusts the light avoidance operation indication based on the light measurement results. The light avoidance operation indication includes local light avoidance and global light avoidance. If only a few grid cells exceed the standard (for example, the cumulative light dose of 5 grid cells is 550 Lux·hr), a local light avoidance indication is generated, such as installing light shields in the area where the standard is exceeded. If multiple or most grid cells exceed the standard (for example, the cumulative light dose of 50 grid cells is greater than 500 Lux·hr), a global light avoidance indication is generated, such as turning off the light source or pulling down the light curtain.

[0038] The system accurately calculates the cumulative light dose of each grid cell through virtual light sensors and the Beer-Lambert law, realizes precise light avoidance operation, and avoids unnecessary interference of traditional global light avoidance methods. The system monitors the illuminance distribution in real time and dynamically feedbacks, quickly responding to light changes. When the light source intensity suddenly increases, it can detect the exceeding standard and generate a light avoidance indication within a few minutes, preventing drug degradation in time, thereby improving the product quality of oxytocin injection.

[0039] Reference Figure 1 The pH value dynamic regulation module is used to establish a buffer capacity prediction model to calculate the buffer capacity and adjust the driving frequency of the metering pump through a fuzzy PID controller; Furthermore, the process of adjusting the driving frequency of the metering pump includes: Collect the pH value in the liquid preparation tank through a pH sensor and record the amount of alkali input; According to the pH value and the amount of alkali, use the Henderson-Hasselbalch equation to calculate the buffer capacity, expressed as: ; ; Among them, is the acidity of the solution, is the negative logarithm of the acid dissociation constant, reflecting the strength of the acid, is the concentration of the base (i.e., the concentration of the conjugate base after acid dissociation), is the concentration of the acid (the concentration of undissociated acid molecules), is the logarithmic function, is the buffer capacity, measuring the ability of oxytocin injection to resist pH value changes, is the change in the amount of alkali added, is the change in oxytocin injection after adding alkali (or acid) of.

[0040] Input the pH error , the buffer capacity and the error change rate into the fuzzy PID controller; the pH error is the difference between the target pH value (e.g., 7.00) and the actual pH value, and the error change rate is the difference between the current error and the error at the previous moment.

[0041] Design fuzzy rules according to expert experience. The fuzzy PID controller outputs PID parameters according to the fuzzy rules; Calculate the control signal according to the PID parameters, and convert the control signal into the driving frequency of the metering pump, expressed as: ; where, is the driving frequency, is the reference frequency, is the conversion coefficient, is the control signal at time , and the control signal is an instruction output by the control system for adjusting the operation of the metering pump, which is calculated by the fuzzy PID controller according to the pH error, error change rate and buffer capacity.

[0042] The pH value dynamic regulation module realizes precise control of the pH value in the liquid preparation process by collecting the pH value and alkali amount in real time, calculating the buffer capacity using the Henderson-Hasselbalch equation, and dynamically regulating the driving frequency of the metering pump in combination with the fuzzy PID controller. It can adjust the acid-base addition amount in real time according to the acid-base balance characteristics of the solution, ensuring the stability and consistency of the oxytocin injection liquid preparation process.

[0043] Reference Figure 1 's PLC control collaborative decision-making module is used to calculate the QRI index using the dynamic weight allocation strategy according to the temperature control parameters, pH value, the cumulative light dose and the liquid preparation time, generate a fault tree set, and perform non-linear compensation for the excess parameters.

[0044] Among them, the excess parameter refers to the key parameter that exceeds the set threshold during the liquid preparation process. Calculating the QRI index is expressed as: ; where, is the QRI index, is the temperature deviation at time t, is the cumulative light dose at time t, is the liquid preparation time at time , , , and are the temperature weight, pH weight, light weight and time weight respectively.

[0045] where, , , and Weights are assigned according to the dynamic weight allocation strategy. As shown in Table 1, different weights corresponding to different stages are given. In the dissolution stage, the pH value has the greatest impact on QRI because the dissolution efficiency is most affected by the pH value. In the mixing stage, both the pH value and temperature jointly affect QRI, but the impact of light is relatively small. In the liquid retention stage, the weight of temperature is the largest because the stability of oxytocin is mainly affected by temperature. This staged weight allocation strategy can more accurately reflect the impact of various factors on QRI at different stages, thus optimizing the control of the liquid preparation process.

[0046] Table 1 Dynamic weights Process stage Temperature weight pH weight Light weight Time weight Dissolution stage 0.4 0.5 0.0 0.1 Mixing stage 0.3 0.4 0.1 0.2 Holding stage 0.5 0.3 0.1 0.1 Furthermore, as Figure 2 shown, the process of generating and analyzing the fault tree set includes: Defining the QRI index exceeding the standard as the top event of liquid preparation, which is used to characterize the final fault manifestation of the QRI index deviation (QRI > 2.0) during the liquid preparation process; Defining the temperature exceeding the standard, pH value abnormality, light cumulative dose exceeding the standard, and liquid preparation time exceeding the time limit as the intermediate events of liquid preparation, which are used to describe the key factors affecting the top event of liquid preparation; Defining equipment failures (mixer failure and refrigeration system failure), sensor errors (thermometer drift and pH sensor aging), and control strategy failures (PLC adjustment lag and parameter setting error) as the basic events of liquid preparation, which are used to describe the physical and system factors that trigger the intermediate events of liquid preparation at the bottom layer; Using logic gates to connect the top event of liquid preparation, the intermediate events of liquid preparation, and the basic events of liquid preparation to generate the fault tree set; Among them, an "OR gate" is used between the top event of liquid preparation (QRI exceeding the standard) and the intermediate events of liquid preparation (each parameter exceeding the standard), indicating that any parameter exceeding the standard may cause the QRI to exceed the standard. The connection between the intermediate events of liquid preparation and the basic events of liquid preparation selects the logic gate according to the specific situation. For example, the temperature exceeding the standard may be caused by a refrigeration system failure or a sensor error, and an "OR gate" is used to connect them; if a certain event requires multiple basic events to occur simultaneously to trigger, an "AND gate" is used.

[0047] Through qualitative analysis, determine the key path of liquid preparation and the minimum cut set of liquid preparation; through quantitative analysis, calculate the probability of the top event of liquid preparation and identify the key risk points.

[0048] Among them, in the fault tree set analysis, qualitative analysis identifies the minimal cut sets, that is, the minimal combinations of basic events that lead to the occurrence of the top event of liquid preparation. For example, "refrigeration system failure" alone can cause the temperature to exceed the standard and lead to the QRI exceeding the standard, so it constitutes a minimal cut set. The critical path of liquid preparation is used to identify the paths with higher occurrence probabilities in the fault tree set, and these high-risk paths are given priority attention. For example, historical data shows that "pH sensor drift" occurs frequently, so its path needs special attention.

[0049] In quantitative analysis, a failure probability is assigned to each basic event of liquid preparation, which can be based on historical data or expert evaluation. According to the fault tree set structure and logical gate relationships, the probability of the QRI exceeding the standard is calculated. For example, if the top event is multiple intermediate events of liquid preparation connected by an "OR gate", the probability of the top event of liquid preparation is , where is the probability of each intermediate event of liquid preparation, is the product. Then, the influence degree of each basic event of liquid preparation on the top event (such as Fussell-Vesely importance) is calculated to identify the key risk points. For example, "metering pump failure" contributes the most to the QRI exceeding the standard, so its importance is the highest.

[0050] For the case of severe overage, exponential decay compensation is used, expressed as: ; where is the compensation term, is the compensation coefficient, is the adjustment coefficient, is the overage duration, is the natural base.

[0051] For the case of slight overage, the compensation value increases slowly, expressed as: ; where is the logarithmic function.

[0052] Fault tree set analysis supports rapid fault diagnosis and the formulation of preventive measures by systematically identifying the potential causes of quality risks, thus ensuring the quality and stability of the oxytocin injection liquid preparation process.

[0053] Through the multi-condition coupling modeling module, the present invention optimizes the process scheduling instructions based on fluid dynamics simulation, improving the uniformity of the liquid dispensing process. The temperature-deviation adjustment module uses a neural network to predict the degradation rate and compensates for the deviation of the refrigeration power, ensuring temperature stability. The light control module adopts a photosensitive feedback adjustment strategy to generate a light-shielding operation protocol, improving light stability. The pH dynamic adjustment module combines a buffer capacity prediction model and fuzzy PID control to achieve precise adjustment. The PLC control collaborative decision-making module calculates the QRI index through a dynamic weight distribution strategy and uses a fault tree set analysis to identify key risk points, optimizing the non-linear compensation of parameters and enhancing the fault diagnosis ability. These modules work together to achieve precise control of the oxytocin injection liquid dispensing process, improving the stability, consistency, and safety of the process, reducing process deviations, and providing a reliable guarantee for high-quality drug production.

[0054] Example 2: Based on Example 1, a pharmaceutical company applied a liquid dispensing production system for oxytocin injection in a small-batch production line. The company used a stainless-steel mixing tank for liquid dispensing. Due to changes in the material flow state, mixing uniformity, and environmental factors in the production line, adaptive adjustments were required. A liquid dispensing production system for oxytocin injection includes: A multi-condition coupling modeling module, which is used to simulate the fluid flow state and material mixing situation of oxytocin in the liquid dispensing tank based on a fluid dynamics model and construct process scheduling instructions; calculate the uniformity index under the process scheduling instructions based on time-dependent simulation and update the process scheduling instructions; A temperature-deviation adjustment module, which is used to predict the degradation rate of oxytocin using a long short-term memory network according to temperature time-series data and compensate for the deviation of the refrigeration power of the process scheduling instructions; A light control module, which is used to calculate the illuminance distribution in the liquid dispensing tank using a photosensitive feedback adjustment strategy and generate a light-shielding operation protocol in combination with the cumulative light dose; A pH dynamic adjustment module, which is used to establish a buffer capacity prediction model to calculate the buffer capacity and adjust the driving frequency of the metering pump through a fuzzy PID controller.

[0055] Table 2 Fuzzy Sets of Input Variables

[0056] Table 3 Fuzzy Sets of Output Variables

[0057] Specifically, according to expert experience and process requirements, fuzzy rules are formulated. The fuzzy sets are shown in Tables 2 and 3, and the input variables are pH error , the buffer capacity and the error change rate , and Take 7 fuzzy subsets, Take 3 fuzzy subsets. The proportional gain of the output variable , integral gain and derivative gain all take 7 fuzzy subsets. Among them, NB is negative large, NM is negative medium, NS is negative small, Z is zero, PS is positive small, PM is positive medium, PB is positive large, S is small, M is medium, and L is large.

[0058] According to the number of fuzzy rules, the possible combination number of input variables is calculated as 7×7×3 = 147, so this fuzzy PID controller has a total of 147 fuzzy control rules. Of course, unnecessary rules can also be trimmed according to specific situations to make it more concise. The fuzzy rules are shown in Table 4. If the pH error is positive large and the buffer capacity is small, then increase the proportional gain and derivative gain for fast response; if the error change rate is large and the buffer capacity is large, then decrease the derivative gain to avoid overshoot. According to the fuzzy rules, calculate the adjustment amounts of the PID parameters ( , and ), and update the PID parameters. The fuzzy PID adaptively adjusts the acid-base addition rate in the liquid preparation process, avoiding over-adjustment or delayed reaction.

[0059] Table 4 Example of Fuzzy Rules

[0060] The PLC control collaborative decision-making module is used to calculate the QRI index using a dynamic weight allocation strategy according to the temperature control parameters, pH value, the cumulative light dose, and the liquid preparation time, generate a fault tree set, and perform non-linear compensation for excess parameters.

[0061] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A system for preparing oxytocin injection, characterized in that: include: The multi-condition coupling modeling module is used to simulate the oxytocin fluid flow state and material mixing in the liquid preparation tank based on the fluid dynamics model, and to build process scheduling instructions; Calculating a uniformity index under the process scheduling instruction based on a time-dependent simulation, and updating the process scheduling instruction; A temperature-deviation adjustment module, used to predict the degradation rate of oxytocin using a long short-term memory network according to the temperature time series data, and compensate for the refrigeration power deviation of the process scheduling instruction; A light control module, used to calculate the illumination distribution in the liquid dispensing tank using a photosensitive feedback adjustment strategy, and generate a light avoidance operation protocol in combination with the accumulated light dose; pH value dynamic adjustment module, used to establish a buffer capacity prediction model to calculate the buffer capacity, and adjust the driving frequency of the metering pump through the fuzzy PID controller; The PLC controlled collaborative decision-making module is used to calculate the QRI index using a dynamic weight allocation strategy according to the temperature control parameters, pH value, the cumulative light dose and the liquid preparation time, generate a fault tree set, and perform nonlinear compensation for excess parameters.

2. The system for preparing oxytocin injection according to claim 1, characterized in that: The process of constructing the process scheduling instruction includes: Introduction The turbulence model simulates the stirring turbulence behavior, and the Navier-Stokes equation is used to establish the oxytocin fluid flow state model for calculating the velocity field and pressure field in the liquid preparation tank; the convection-diffusion equation is used to model the material mixing situation for calculating the concentration distribution; The liquid dispensing tank is divided into grid units based on the finite volume method, and local mesh encryption is performed in the stirring area and the feed inlet area; the division error is calculated using an adaptive grid adjustment strategy, and the grid units are updated; According to the velocity field, the pressure field and the concentration distribution, an integral equation is established for each grid unit based on the Gauss-Green theorem, and the SIMPLE algorithm is used to solve the velocity-pressure coupling, the velocity field, the pressure field and the concentration distribution are iteratively updated, and the process scheduling instructions are generated.

3. The system for preparing oxytocin injection according to claim 1, characterized in that: The updating process of the process scheduling instruction includes: At each time step, the homogeneity index is calculated , expressed as: ； in, For grid cells The concentration at is the average concentration in the liquid distribution tank, is the actual mixing time, is the theoretical mixing time, is the dimensionless weight coefficient, is the total number of grid cells; like If the value is less than a preset threshold, a genetic algorithm is used to adjust the feeding sequence and update the process scheduling instructions.

4. The system for preparing oxytocin injection according to claim 3, characterized in that: The implementation process of the genetic algorithm includes: Using an arrangement coding method to set the order of oxytocin materials, and generating an initial population according to the order of oxytocin materials and the process scheduling instructions; The maximization of the homogeneity index is set as the fitness function, and a roulette wheel selection operation, partial matching crossover and exchange mutation are adopted. When the maximum number of iterations is reached, the updated process scheduling instruction is output.

5. The system for preparing oxytocin injection according to claim 1, characterized in that: The process of generating light avoidance instructions includes: Deploy virtual light sensors on the grid cells, calculate light attenuation using the Beer-Lambert law, and obtain the illumination distribution; Multiplying the illumination distribution, exposure time and light degradation coefficient of the grid unit to obtain the light cumulative dose; For each of the grid units, the accumulated light dose is compared with the safe light threshold, and if the accumulated light dose is greater than the safe light threshold, the light avoidance operation instruction is generated; otherwise, no additional operation is performed; The light-avoidance operation instructions include local light-avoidance and global light-avoidance.

6. The system for preparing oxytocin injection according to claim 1, characterized in that: The driving frequency adjustment process of the metering pump includes: The pH value in the liquid preparation tank is collected by a pH sensor, and the amount of alkali added is recorded; Calculating the buffer capacity using the Henderson-Hasselbalch equation based on the pH value and the amount of alkali; Inputting pH error, the buffer capacity and the error change rate into the fuzzy PID controller; Design fuzzy rules according to expert experience, and the fuzzy PID controller outputs PID parameters according to the fuzzy rules; A control signal is calculated according to the PID parameters, and the control signal is converted into a driving frequency of the metering pump.

7. The system for preparing oxytocin injection according to claim 1, characterized in that: The process of generating and analyzing the Fault Tree Set includes: The QRI index exceeding the standard is defined as the top event of liquid preparation, which is used to characterize the deviation of the QRI index from the final fault performance during the liquid preparation process; Temperature exceeding the limit, pH value abnormality, light cumulative dose exceeding the limit and liquid preparation time exceeding the limit are defined as liquid preparation intermediate events, which are used to describe the key factors affecting the liquid preparation top event; Equipment failure, sensor error and control strategy failure are defined as basic events of liquid preparation, which are used to describe the physical factors and system factors at the lowest level that cause the intermediate events of liquid preparation; Use logic gates to connect the liquid dispensing top event, the liquid dispensing intermediate event and the liquid dispensing basic event to generate the fault tree set; Through qualitative analysis, the critical path of liquid preparation and the minimum cut set of liquid preparation are determined; through quantitative analysis, the probability of liquid preparation top event is calculated and key risk points are identified.

Images