3D printer multi-dimensional state data real-time acquisition system

Abstract

The application relates to the technical field of computers, in particular to a 3D printer multidimensional state data real-time acquisition system, which comprises a data acquisition module, a feature engineering module, a state prediction module and a self-adaptive control module; the data acquisition module is used for acquiring multidimensional state parameters in a printing process in real time; the feature engineering module is used for generating a comprehensive feature vector based on the multidimensional state parameters acquired by the data acquisition module; the state prediction module is used for predicting expected cell activity and expected geometric fidelity by adopting a physical information neural network model according to the comprehensive feature vector generated by the feature engineering module and a preset candidate printing speed; the self-adaptive control module is used for determining an optimal printing speed according to the expected cell activity and the expected geometric fidelity; and the instruction execution module is used for real-time regulation and control of the operation of the printer in response to the optimal printing speed. The application changes the contradictory relationship between the printing speed and the cell activity into a target that can be optimized in coordination, and realizes a fundamental change in the technical path.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a 3D printer multi-dimensional state data real-time acquisition system. BACKGROUND

[0002] In the application of in-situ bio 3D printing, the core challenge lies in the dynamic and nonlinear process environment. The cell activity of bio-ink naturally decays over time, while the shear force in the printing process causes damage to it; the fluctuations of patient physiological environment (such as body temperature, pH value) and the changes of ink rheological properties are coupled with each other, which jointly affect the geometric accuracy and biological functionality of the final printed scaffold. The existing technology uses fixed printing parameters, which cannot cope with this time-varying coupling relationship, resulting in a technical contradiction that sacrifices cell activity in pursuit of printing speed, or sacrifices printing efficiency in the protection of cell activity. SUMMARY

[0003] The present application provides a 3D printer multi-dimensional state data real-time acquisition system, which solves the problems in the background art.

[0004] To solve the above technical problems, the present application provides a 3D printer multi-dimensional state data real-time acquisition system, which comprises:

[0005] A data acquisition module for real-time acquisition of multi-dimensional state parameters in the printing process;

[0006] A feature engineering module for generating a comprehensive feature vector based on the multi-dimensional state parameters acquired by the data acquisition module;

[0007] A state prediction module for predicting the expected cell activity and the expected geometric fidelity by using a physical information neural network model according to the comprehensive feature vector generated by the feature engineering module and the preset candidate printing speed;

[0008] An adaptive control module for determining the optimal printing speed according to the expected cell activity and the expected geometric fidelity;

[0009] An instruction execution module for real-time regulation and control of the operation of the printer in response to the optimal printing speed.

[0010] Preferably, the data acquisition module is used to acquire biological metabolic parameters, rheological and mechanical parameters, and geometric and environmental parameters as multi-dimensional state parameters.

[0011] Preferably, the feature engineering module is used to:

[0012] Determine the metabolic rate feature based on the multi-dimensional state parameters;

[0013] Determine the cumulative stress damage feature based on the multi-dimensional state parameters;

[0014] determine a geometric deviation scalar based on the multi-dimensional state parameters and a preset numerical model;

[0015] generate a comprehensive feature vector by combining the metabolic rate feature, the cumulative stress damage feature, and the geometric deviation scalar.

[0016] Preferably, the physical information neural network model is trained by optimizing a composite loss function.

[0017] The composite loss function is composed of a data-driven loss term and a biological mechanism constraint loss term.

[0018] Preferably, the biological mechanism constraint loss term is constructed based on a preset ordinary differential equation of cell activity decay.

[0019] The ordinary differential equation of cell activity decay is used to describe the variation law of cell survival rate under the coupled influence of metabolic environment and mechanical shear force.

[0020] Preferably, the adaptive control module determines the optimal printing speed, including:

[0021] For each candidate printing speed, call the state prediction module to obtain its corresponding expected cell activity and expected geometric fidelity;

[0022] Compare the expected cell activity with a preset minimum cell survival rate threshold;

[0023] If the expected cell activity is not lower than the minimum cell survival rate threshold, the candidate printing speed is determined as an effective candidate printing speed;

[0024] If the expected cell activity is lower than the minimum cell survival rate threshold, the candidate printing speed is determined as an ineffective candidate printing speed;

[0025] For each effective candidate printing speed, calculate its comprehensive utility score;

[0026] Select the effective candidate printing speed with the maximum comprehensive utility score as the optimal printing speed.

[0027] Preferably, the comprehensive utility score is calculated, including:

[0028] Determine a cell activity weighted score based on the expected cell activity;

[0029] Determine a geometric fidelity weighted score based on the expected geometric fidelity;

[0030] Determine a normalized time deduction term based on the required time to complete the printing section and a preset reference time;

[0031] The cell activity weighted score is added to the geometric fidelity weighted score and subtracted by a normalized time deduction term to generate a comprehensive utility score.

[0032] Preferably, the minimum cell survival rate threshold is set according to the clinical medical requirements for ensuring that the printed tissue scaffold meets the preset biological functionality.

[0033] Compared with the prior art, the present application has the following beneficial effects:

[0034] The present application provides a 3D printer multi-dimensional state data real-time acquisition system, which has the following beneficial effects:

[0035] (1) The present application realizes deep and dynamic perception of the printing process state; the prior art usually uses fixed process parameters, and completely ignores key information such as biological ink activity attenuation, rheological property change, and microenvironment fluctuation during the printing process; the present application is different from the prior art, in that real-time multi-dimensional state parameters are obtained through a data acquisition module, and a comprehensive feature vector is further generated through a feature engineering module; more importantly, the comprehensive feature vector contains deep information such as metabolic rate features and cumulative stress damage features, which have time accumulation effects and dynamic change trends; as a direct result, the system no longer obtains isolated and instantaneous state snapshots, but a dynamic portrait with extremely rich information dimensions that can reflect the system evolution history and future trends; such high-quality input information is a logical prerequisite for subsequent accurate prediction and effective control, and is not possessed by the prior art.

[0036] (2) The present application constructs a prediction model with high confidence and physical constraints; for a complex biological printing process, a black-box prediction model that purely relies on data is often unstable and may violate basic biological or physical principles in real scenarios with limited data and serious noise interference; the state prediction module of the present application adopts a physical information neural network model trained by a composite loss function composed of a data-driven loss term and a biological mechanism constraint loss term, which fundamentally avoids this risk;

[0037] The actual significance of the composite loss function is that it forces the neural network to learn the experimental data while its output strictly obeys the biological laws defined by the cell activity attenuation ordinary differential equation; the ordinary differential equation can be constructed and implemented by those skilled in the art based on the parameters and functional relationships described in detail in the specification; this ensures that the model's prediction of expected cell activity and expected geometric fidelity still has extremely high accuracy and physical reality even under unseen working conditions; this capability enables the system to upgrade from passive response to active prediction, making it possible to achieve truly intelligent control.

[0038] The present application realizes real-time, safe and optimal adaptive control of printing strategy; the prior art adopts constant printing speed, which is essentially a global, non-optimal compromise solution; the adaptive control module of the present application realizes fine and personalized control of each printing section; in the process of determining the optimal printing speed, the module uses a double security mechanism; first, by comparing the expected cell activity with the minimum cell survival rate threshold set according to clinical medical requirements, all logical possibilities are exhausted, any candidate printing speed that may cause the cell survival rate to be lower than the safety baseline is determined as invalid candidate printing speed and directly excluded, thus building an insurmountable safety baseline; second, for all valid candidate printing speeds, by calculating their comprehensive utility scores , the three conflicting targets of printing speed, expected cell activity and expected geometric fidelity are unified into a quantifiable optimization framework;

[0039] The actual significance of the comprehensive utility score is that it provides a mathematical basis for the system on how to make the best trade-off between various performance targets under the premise of meeting safety constraints; for example, when printing a porous filling area inside a hip cup that requires rapid prototyping and has low precision requirements for internal structure, the function will calculate that a higher printing speed can bring greater comprehensive utility; conversely, when printing the edge of a hip cup that requires precise fitting, the function will point to a lower speed to ensure geometric fidelity, thus obtaining the highest comprehensive utility. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor;

[0041] Figure 1 is the logical block diagram of the system of the present application;

[0042] Figure 2 is the logical block diagram of the feature engineering module of the present application;

[0043] Figure 3 is the logical block diagram of the physical information neural network model of the present application;

[0044] Figure 4 is the logical block diagram of the present application for determining the optimal printing speed. DETAILED DESCRIPTION

[0045] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0046] Embodiment one

[0047] Please refer to Figure 1 The present application provides a 3D printer multi-dimensional state data real-time acquisition system, comprising:

[0048] A data acquisition module is configured to acquire multi-dimensional state parameters in a printing process in real time.

[0049] A feature engineering module is configured to generate a comprehensive feature vector based on the multi-dimensional state parameters acquired by the data acquisition module.

[0050] A state prediction module is configured to predict expected cell activity and expected geometric fidelity by using a physical information neural network model according to the comprehensive feature vector generated by the feature engineering module and a preset candidate printing speed.

[0051] An adaptive control module is configured to determine an optimal printing speed according to the expected cell activity and the expected geometric fidelity.

[0052] An instruction execution module is configured to real-time control the operation of the printer in response to the optimal printing speed.

[0053] The present embodiment aims to solve the technical problem that the success rate of surgery is limited due to the contradiction between printing speed and cell activity in in-situ bio-printing. The system realizes the fundamental reconstruction of the printing process through a complete closed-loop feedback architecture composed of a data acquisition module, a feature engineering module, a state prediction module, an adaptive control module and an instruction execution module. This architecture enables the system to perceive subtle changes in the printing environment in real time and make dynamic decisions based on accurate predictions of the future. The core of the system is that it performs parallel evaluation on a set of preset candidate printing speeds, and selects the optimal printing speed based on a comprehensive utility function that takes into account cell activity, geometric fidelity and printing efficiency, while meeting the biological safety bottom line. The system then real-time controls the operation of the printer through the instruction execution module, thereby transforming the contradictory relationship between printing speed and cell activity into a synergistic optimization goal, and realizing the fundamental change of the technical path.

[0054] Embodiment two

[0055] Please refer to Figure 2, which illustrates the parallel computing metabolic rate feature, the cumulative stress damage feature and the geometric deviation scalar, and finally combines to generate the comprehensive feature vector, the data acquisition module is used to obtain biological metabolic parameters, rheology and mechanical parameters and geometric and environmental parameters as multi-dimensional state parameters.

[0056] Specific acquisition methods:

[0057] Biological metabolic parameters (glucose , lactic acid ): Real-time detection by microfluidic chip sensor (model: BioSenSE2000), sampling frequency 1Hz;

[0058] Rheology and mechanical parameters (shear stress ): converted by the built-in pressure sensor of the printer nozzle (range 0~100kPa, accuracy ±0.5kPa) combined with the fluid mechanics model (k , is the calibration coefficient);

[0059] Geometric and environmental parameters (temperature , pH, printing coordinates): Infrared temperature sensor (accuracy ±0.1℃), pH electrode (range 6.0~8.0), laser displacement sensor (resolution 1μm) synchronous acquisition, sampling frequency 10Hz.

[0060] The feature engineering module is used to:

[0061] Determine the metabolic rate feature based on multi-dimensional state parameters;

[0062] Determine the cumulative stress damage feature based on multi-dimensional state parameters;

[0063] Determine the geometric deviation scalar based on multi-dimensional state parameters and a preset digital model;

[0064] Combine the metabolic rate feature, the cumulative stress damage feature and the geometric deviation scalar to generate a comprehensive feature vector.

[0065] Specifically, the calculation method of the geometric deviation scalar One non-limiting example is as follows: During printing, the laser displacement sensor real-time collects the coordinate point set of the actual deposition profile on the current printing path At the same time, the corresponding theoretical profile coordinate point set is extracted from the preset digital model (such as CAD model).

[0066] The expected geometric fidelity output by the state prediction module is a normalized scalar, and its relationship with the geometric deviation scalar can be defined by the following non-limiting formula: , wherein is a positive sensitivity coefficient calibrated by experiment to map distance deviation to the fidelity score of the interval.

[0067] In the initial stage of system running, the data acquisition module acts as the physical perception interface of the system, whose actions of acquiring biological metabolic parameters, rheological and mechanical parameters, and geometric and environmental parameters provide quantitative factual basis for the cognition and decision of the whole system. These original and instantaneous multidimensional state parameters, although reflecting the current situation, are not sufficient to reveal the deep trend of system evolution. In order to establish a more profound causal relationship, the construction of the feature engineering module is a logical prerequisite for effective prediction. Based on the acquired multidimensional state parameters, the module determines the metabolic rate feature, the cumulative stress damage feature, and the geometric deviation scalar through calculation. Among them, the metabolic rate feature reveals the dynamic health degree of the cell population; the cumulative stress damage feature quantifies the total mechanical stress history that the cell has suffered from leaving the ink cartridge to the current time; the geometric deviation scalar evaluates the physical accuracy of the printed structure in real time. Thereafter, the feature engineering module generates a comprehensive feature vector with extremely high information density by combining these deeply processed features. This vector is not only a description of the current state, but also a condensation of the system history and dynamics, which builds the necessary and high-quality input for the state prediction module to make accurate prediction.

[0068] Example Three

[0069] Please refer to Figure 3 , which shows the core idea of constructing a composite loss function composed of a data-driven loss term and a biology mechanism constraint loss term, and using it to optimize the training model; the physical information neural network model is trained by optimizing the composite loss function.

[0070] Among them, the composite loss function is composed of a data-driven loss term and a biology mechanism constraint loss term.

[0071] The specific structure of the physical information neural network is: input layer (12 dimensions, including 8-dimensional comprehensive feature vector + 4-dimensional candidate velocity parameter) → hidden layer 1 (64 nodes, ReLU activation) → hidden layer 2 (32 nodes, ReLU activation) → hidden layer 3 (16 nodes, ReLU activation) → output layer (2 nodes, Sigmoid activation, respectively output expected cell activity, expected geometric fidelity).

[0072] The training data comes from 3 groups of experiments: ① 10 kinds of biological ink (containing stem cell concentration 1 x 10 6 ~ 5 x 10 6 cells / mL); ② 5 kinds of printing speed (0.55 mm / s); ③ 3 kinds of environmental conditions (temperature 36-38℃, pH 7.2-7.6). The data-driven loss term The mean square error is adopted: is the sample size, is the measured value, is the predicted value). Training parameters: Adam optimizer, learning rate 0.001, 1000 iterations, batch size 32.

[0073] The construction of the loss term constrained by biological mechanism is based on a preset ordinary differential equation of cell activity attenuation;

[0074] Among them, the ordinary differential equation of cell activity attenuation is used to describe the change law of cell survival rate under the coupled influence of metabolic environment and mechanical shear force.

[0075] The predictive ability of the system is derived from the physical information neural network model adopted by the state prediction module, and the prediction performance of the model is determined by the composite loss function optimized in the training stage;

[0076]

[0077] The technical motivation of this composite loss function design is to overcome the insufficient generalization ability and the inherent defect of violating the physical intuition prediction of the pure data-driven model when facing complex biological systems due to limited training data; by introducing a constraint term based on the first principle of biology , the neural network is forced to learn the inherent law of the data while complying with the known cell activity mechanics; this greatly improves the accuracy and robustness of the model prediction in the small sample or high noise in situ printing scene;

[0078] Among them, is the total objective function that needs to be minimized in the system training;

[0079] is the data-driven loss term, which is in the form of mean square error, used to measure the gap between the model prediction and the real experimental data; is the biological mechanism constraint loss term, which is the core to ensure that the prediction result conforms to the specific physical law; is a dimensionless weight hyperparameter used to balance the relative importance of data fitting and physical constraints, and its value is determined by grid search or Bayesian optimization on the cross-validation set;

[0080] In specific applications, the state prediction module is precisely the physical information neural network model trained by this composite loss function, which can accurately predict the cell activity and geometric fidelity in the near future; this gives the system mechanism-based reasoning ability, and it no longer deals with isolated data points, but the biological-physical coupling law behind the data; as the core, the biological mechanism constraint loss term​ The construction is based on a pre-defined ordinary differential equation for cell activity decay;

[0081]

[0082] This equation is an extension of the classical apoptosis model, accurately describing the changes in cell viability under the coupled influence of the metabolic environment and mechanical shear forces. It constitutes... The physical kernel;

[0083] in, This refers to dimensionless cell viability. For time; This refers to the glucose concentration. This refers to the concentration of lactic acid. The ambient temperature under the Kelvin temperature scale; pH level; Shear stress; The basic apoptosis rate constant is derived from measurements obtained through in vitro cell culture experiments;

[0084] This is the shear force damage coefficient, whose dimension is the reciprocal of the Pascal-second (Pa). The physical meaning of shear stress is the rate of decay of cell viability per unit shear stress per unit time; its specific value is derived from calibration experiments using a rheometer coupled with cell counting. A non-limiting example of the calibration method is as follows: placing cell-containing bio-ink in a rheometer, and applying a series of known, constant shear stresses. At different times On cell survival rate Sampling and counting were performed, subtracting the number of cells that had undergone basal apoptosis ( After considering the impact of the term, the data points are fitted to the equation. Determine The value;

[0085] A custom dimensionless function to describe the effect of metabolic state on apoptosis;

[0086] A custom dimensionless function to describe the effect of environmental factors on exacerbating damage.

[0087] Among them, the metabolic influence function The specific form is:

[0088]

[0089] In the formula, The value is the metabolic sensitivity coefficient (calibrated by in vitro culture experiments of human bone marrow mesenchymal stem cells, with a range of 0.02~0.05). To avoid the minimum value of the denominator being 0;

[0090] environmental impact function The specific form is:

[0091]

[0092] In the formula, T is the physiological temperature (37℃), pH is the physiological pH value; (Temperature coefficient, determined by cell survival rate experiment in the range of 3242℃), (pH coefficient, calibrated by damage experiment in the range of pH 6.88.0), both are dimensionless coefficients.

[0093] Example Four

[0094] Please refer to Figure 4 , which shows in detail the logic flow of the adaptive control module determining the optimal printing speed, which is a set optimization method. The adaptive control module as the decision and optimization unit of the system, it determines the optimal printing speed with rigorous and logically complete process. For each candidate printing speed, call the state prediction module to obtain its corresponding expected cell activity and expected geometric fidelity; wherein, the candidate printing speed can be a discrete speed set covering from low speed to high speed according to experience, for example ;

[0095] The adaptive control module determines the optimal printing speed, including:

[0096] For each candidate printing speed, call the state prediction module to obtain its corresponding expected cell activity and expected geometric fidelity;

[0097] Compare the expected cell activity with the preset minimum cell survival rate threshold;

[0098] If the expected cell activity is not lower than the minimum cell survival rate threshold, the candidate printing speed is determined as a valid candidate printing speed;

[0099] If the expected cell activity is lower than the minimum cell survival rate threshold, the candidate printing speed is determined as an invalid candidate printing speed;

[0100] For each valid candidate printing speed, calculate its comprehensive utility score;

[0101] Select the valid candidate printing speed with the maximum comprehensive utility score as the optimal printing speed.

[0102] The calculation of the comprehensive utility score includes:

[0103] determine a cell viability weighted score based on the expected cell viability;

[0104] determine a geometric fidelity weighted score based on the expected geometric fidelity;

[0105] determine a normalized time deduction term based on the required time to complete the printing segment and a preset reference time;

[0106] add the cell viability weighted score and the geometric fidelity weighted score, and subtract the normalized time deduction term, to generate a comprehensive utility score.

[0107] The minimum cell survival rate threshold is set according to the clinical medical requirement of ensuring that the printed tissue scaffold meets the preset biological functionality.

[0108] The adaptive control module serves as the decision and optimization unit of the system, which determines the optimal printing speed in a rigorous and logically complete process. As a prerequisite for decision-making, for each candidate printing speed, the module calls the state prediction module for future deduction to obtain the corresponding expected cell viability and expected geometric fidelity. Based on the prediction results, a key logical judgment is performed: comparing the expected cell viability with the preset minimum cell survival rate threshold. The setting logic of the minimum cell survival rate threshold is strictly according to the clinical medical requirement of ensuring that the printed tissue scaffold can be successfully vascularized and have long-term biological functionality, for example, set to 85%. This is an insurmountable hard constraint. According to this comparison, the candidate printing speed is clearly divided into two categories: the expected cell viability is not lower than the threshold, which is judged as an effective candidate printing speed; otherwise, it is judged as an ineffective candidate printing speed and is directly excluded from the decision set. This division without gray zone ensures that the printing process will never sacrifice cell basic survival at the cost.

[0109] For all candidate printing speeds judged to be effective , the adaptive control module will further calculate its comprehensive utility score, aiming to unify the three conflicting goals of high activity, high precision, and high efficiency into a quantifiable optimization framework.

[0110] A preferred calculation method of the comprehensive utility score is to define it as the utility output per unit reference time, and its specific form is:

[0111] ;

[0112] The fundamental motivation of this utility function is to provide a unified quantitative evaluation scale for multiple performance objectives that are originally in conflict, that is, the weighted quality score obtained per unit of normalized time. It converts the complex multi-objective decision-making problem into a clearly defined, single-objective optimization problem.

[0113] wherein, is the printing speed of the printing path is the non-dimensional overall utility score at the printing speed ;

[0114] is the expected cell viability corresponding to the printing speed output by the state prediction module, which is a non-dimensional value; is the expected geometric fidelity corresponding to the printing speed output by the state prediction module, which is a non-dimensional value; and are user-defined non-dimensional weight coefficients corresponding to the cell viability and the geometric fidelity respectively, whose sum is usually normalized to 1, for example . These coefficients allow the adjustment of the emphasis on activity and accuracy according to different printing tasks, such as printing load-bearing structures or non-load-bearing structures; is the expected time required to complete the printing path at the printing speed , which is in seconds; is a preset reference time in seconds, for example, it can be set to 1 second, and the ratio of the two constitutes a normalized time term, representing how many reference time units are needed to complete the path.

[0115] In determining the optimal printing speed, the module converts the abstract performance advantages and disadvantages into specific numerical values by calculating the overall utility score of each effective candidate printing speed; this calculation process includes: determining the cell viability weighted score based on the expected cell viability; determining the geometric fidelity weighted score based on the expected geometric fidelity; combining the normalized time deduction term, and finally generating the overall utility score through weighted summation and deduction; on this basis, the adaptive control module selects the effective candidate printing speed with the maximum overall utility score and determines it as the optimal printing speed; this selection process enables the system to intelligently judge that, on the basis of ensuring the cell survival bottom line, it can speed up appropriately to exchange for higher activity when printing areas such as bone scaffold internal filling that do not require high precision, and automatically slow down to ensure geometric fidelity when printing load-bearing areas or fine tissue contours; this achieves a dynamic and optimal balance of the relationship between speed, activity and precision throughout the printing task.

[0116] The fundamental progress lies in that, by constructing a complete closed-loop feedback architecture composed of data acquisition, feature engineering, state prediction and adaptive control, the present application transforms in-situ bio-printing from an open-loop process relying on static, offline parameters into a closed-loop intelligent process capable of real-time sensing, accurate prediction and dynamic optimization; this paradigm shift completely solves the technical dilemma in the prior art that the printing speed and cell viability cannot be reconciled due to the inability to cope with dynamic, nonlinear coupling factors.

[0117] In summary, the series of interlocking technical features collectively achieve the overall beneficial effects of the present application: compared with the prior art, the present system can shorten the overall printing time by more than 40% while ensuring that the cell survival rate is not less than 85%, significantly improving the efficiency and success rate of in situ bio-printing, and clearing the key obstacles for the clinical application of this frontier technology.

[0118] The above merely describes the preferred embodiments of the present application, and is not intended to limit the present application in other forms, and any person skilled in the art can modify or change the above disclosed technical content to equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification made according to the technical essence of the present application to the above embodiments without departing from the technical solution content of the present application still belongs to the protection scope of the technical solution of the present application.

Claims

1. A multi-dimensional state data real-time acquisition system of a 3D printer, characterized in that, The method comprises the following steps: a data acquisition module is used to acquire multi-dimensional state parameters in real time during the printing process; a feature engineering module is used to generate a comprehensive feature vector based on the multi-dimensional state parameters acquired by the data acquisition module; a state prediction module is used to predict the expected cell activity and the expected geometric fidelity by using a physical information neural network model according to the comprehensive feature vector generated by the feature engineering module and a preset candidate printing speed; an adaptive control module is used to determine the optimal printing speed according to the expected cell activity and the expected geometric fidelity; an instruction execution module is used to real-time control the operation of the printer in response to the optimal printing speed. The data acquisition module is used to acquire biological metabolic parameters, rheological and mechanical parameters, and geometric and environmental parameters as multi-dimensional state parameters.

2. The system according to claim 1, characterized in that The data acquisition module acquires biological metabolic parameters by using a microfluidic chip sensor, acquires rheological and mechanical parameters by using a pressure sensor built in a printer nozzle, and acquires geometric and environmental parameters by using a laser displacement sensor.

3. The system according to claim 1, characterized in that, The feature engineering module is used to: determine a metabolic rate feature based on the multi-dimensional state parameters; determine a cumulative stress damage feature based on the multi-dimensional state parameters; determine a geometric deviation scalar based on the multi-dimensional state parameters and a preset digital model; generate a comprehensive feature vector by combining the metabolic rate feature, the cumulative stress damage feature, and the geometric deviation scalar.

4. The system according to claim 3, characterized in that, When determining the geometric deviation scalar, the feature engineering module is specifically used to calculate the distance between the coordinate point set of the actual deposition profile collected by the laser displacement sensor and the coordinate point set of the theoretical profile extracted from the preset digital model.

5. The system according to claim 1, characterized in that, The physical information neural network model is trained by optimizing a composite loss function; wherein the composite loss function is composed of a data-driven loss term and a biological mechanism constraint loss term.

6. The system according to claim 5, characterized in that The construction of the biological mechanism constraint loss term is based on a preset cell activity attenuation ordinary differential equation; wherein the cell activity attenuation ordinary differential equation is used to describe the change rule of the cell survival rate under the coupled influence of the metabolic environment and the mechanical shear force.

7. The system according to claim 1, characterized in that, The adaptive control module determines the optimal printing speed, including: for each candidate printing speed, calling the state prediction module to obtain its corresponding expected cell activity and expected geometric fidelity; comparing the expected cell activity with a preset minimum cell survival rate threshold; if the expected cell activity is not lower than the minimum cell survival rate threshold, the candidate printing speed is determined as a valid candidate printing speed; if the expected cell activity is lower than the minimum cell survival rate threshold, the candidate printing speed is determined as an invalid candidate printing speed; for each valid candidate printing speed, calculating its comprehensive utility score; selecting the valid candidate printing speed with the maximum comprehensive utility score as the optimal printing speed.

8. The system according to claim 7, characterized in that The calculation of the comprehensive utility score includes: determining a cell activity weighted score based on the expected cell activity; determining a geometric fidelity weighted score based on the expected geometric fidelity; determining a normalized time deduction term based on the required time to complete the printing section and a preset reference time; adding the cell activity weighted score and the geometric fidelity weighted score, and subtracting the normalized time deduction term to generate the comprehensive utility score.

9. The system according to claim 7, characterized in that, The minimum cell survival rate threshold is set according to clinical medical requirements to ensure that the printed tissue scaffold meets preset biological functionality.

Images