1. Introduction Fermentation of clean enzyme is the main industrial process for producing clean enzyme, which takes approximately 40 to 50 hours. Factors affecting clean enzyme yield include inoculum concentration, initial nutrient concentration (amino nitrogen, total sugar, pH), nutrient concentration during the reaction process, reaction temperature, and reactor rotation speed. These are referred to as environmental variables, and they are functions of reaction time. Figure 1 shows the curves of several key environmental variables changing over time in this "process," where "potency" reflects the cumulative increase in the clean enzyme product—"clean enzyme units." During clean enzyme fermentation, the process control system regulates these environmental variables, directing the fermentation process in a direction conducive to increasing "potency." [align=center] Figure 1 Schematic diagram of the changes in main parameters of the Jiemenzin fermentation [/align] [b]2. Establishing the Mathematical Model of the Process 2.1 Dynamic Mathematical Model[/b] The method for establishing a dynamic mathematical model of the Jiemenzin fermentation process is to start from the fermentation mechanism, construct a set of differential equations (parameters to be determined) reflecting its dynamic characteristics, and then perform simulation calculations on the parameters, initial conditions, and boundary conditions of the equation set. The calculation results are then compared with the observation data of the actual production process. This process is repeated multiple times until a set of calculation results that fits the actual data "well" is obtained. Substituting the model parameter values, initial values, and boundary values ​​corresponding to the "good" results into the original set of differential equations, the dynamic mathematical model of the process is obtained. [b]2.2 Static Mathematical Model 2.2.1 The Process of Establishing a Static Mathematical Model[/b] The method for establishing a static mathematical model of the Jiemenzin fermentation process is to use statistical principles to construct a set of algebraic equations reflecting its input/output characteristics. A prerequisite for establishing a static mathematical model is that the process has "stable" characteristics. The fermentation process of clean enzyme is essentially a dynamic process. Its fermentation reaction mechanism is very complex. The coupling degree between various variables in the reaction process is high and the time lag is large. However, since the fermentation process lasts for a long time and the reactants change slowly, from the perspective of actual production, it can be treated as a "stable" or "static" process in a macroscopic way. In this case, we only need to know the relationship between the steady state of its input/output. When considering static problems, regardless of how the output variable changes to a new steady state over time after the input changes, and regardless of what "fluctuations" the output process will produce when inputting, we assume that a certain input corresponds to a certain output [2]. This is the basic idea of ​​establishing a static mathematical model and optimizing its parameters. The general process of establishing a static mathematical model is to first construct a set of polynomial input/output equations (the order and parameters of which are to be determined), then input the observation data of the actual process, and have the system identification module "identify" the input/output relationship of the production system, and finally determine the parameters and order of the undetermined equations. The entire identification process described above was automatically completed on the "integrated digital simulation platform software." 2.2.2 Continuous Variable Processing To adapt to the simulation calculations of continuous production processes by digital computers, continuous variables need to be discretized, which is a key technology for achieving static modeling. The method is to divide the fermentation process of the clean enzyme into several time periods, such as the initial, middle, and late stages of fermentation, and take several feature points for each period. In this way, the continuous variables are transformed into a set of relatively independent discretized feature vectors, and a new coordinate space—a high-dimensional feature vector space—is constructed based on these feature vectors. Static modeling is then achieved using high-dimensional data analysis techniques in this coordinate space. The author used high-dimensional data analysis techniques to calculate 180 sets of data from actual production, found a hypersurface equation in a high-dimensional space, and used this equation to back-judge the original sample data. This hypersurface equation can distinguish between "good" and "bad" sample groups during the clean enzyme fermentation process. [align=center] Figure 2. Schematic diagram of nonlinear mapping [/align] Figure 2 shows the spatial distribution of 50 sets of sample data on a "plane" using integrated digital simulation software and a nonlinear mapping method. In the figure, "#" represents a "good" sample group, and "*" represents a "bad" sample group. This result indicates that using the aforementioned characteristic variables as model variables can well "explain" the fermentation process, thus it can be used to establish a static mathematical model of the process. 2.2.3 Model Optimization The steps for parameter optimization of the static mathematical model are as follows: First, analyze the practically feasible range and characteristics of the model parameters. Define a parameter calculation table using uniform design or orthogonal design methods. Adjust the values ​​of the model parameters according to this calculation table and input them into the simulation model for calculation. During this process, carefully observe the changes in the "cleaning enzyme fermentation growth curve" corresponding to each calculation result and compare it with the actual cleaning enzyme fermentation growth curve in the production process. Select the "good growth curve," and the corresponding calculation result can be used as the optimized model parameters for the cleaning enzyme fermentation process. It is worth noting that the standard for a "good curve" should be the result after comprehensively evaluating the cleaning enzyme fermentation unit, fermentation cost, and output benefits. 2.2.4 Example of the Model The static mathematical model of the gentamicin fermentation process established in the "Optimization Study of Gentamicin Fermentation Process" of a pharmaceutical factory was obtained after multiple simulation calculations based on the mechanism of gentamicin fermentation and using "integrated digital simulation software [2,3]". The mathematical expression of a static model of the gentamicin fermentation process is as follows: Y=-2.48X1+2.24X2+1.80X3-3.27X4+0.02X5-0.04X6-0.04X7-0.07X8-0.12X9+0.05X10-0.56X11+0.59X12+0.17X13 Where Y represents the output (i.e., gentamicin fermentation unit), each Xi represents the environmental variables of the gentamicin fermentation process (continuous variables have been replaced by discrete characteristic variables), and the coefficients of each variable (model parameters) reflect their influence on the fermentation results. 3. Achieving Optimal Control A computer-simulated control system for biopharmaceutical processes integrates data acquisition, simulation calculation, and process control. Its core is establishing a mathematical model of the production process and analyzing, calculating, and dynamically "regulating" various control variables. 3.1 Integration of Simulation and Control During the production process, the simulation system is used to detect, analyze, adjust, and predict environmental variables. This series of processes is automated. That is, when the changes in certain environmental variables exceed set ranges, the system automatically triggers and starts a new round of simulation calculations. It automatically analyzes the "environment" of the control process and "finds" the optimal values ​​of environmental variables under the new environment, thereby establishing an optimal control model for the new environment. The specific calculation process involves arranging a virtual "control parameter table" based on a multi-factor, multi-level orthogonal design table or a uniform design table. Calculations are then performed according to the table, and the calculated values ​​are correlated with the actual process detection data. Based on a certain "degree of correlation," control variables are evaluated and filtered in real time to achieve dynamic parameter optimization. Finally, the new calculation results are transmitted to the process control system, providing a basis for adjusting the control variables and achieving optimal parameter control in the biopharmaceutical process. During system operation, process control personnel activate the on-site data acquisition and monitoring system. When the operating conditions (environmental variables) change, the simulation system is automatically triggered and started for simulation calculations. In this process, the front-end production control system and the back-end digital simulation system together form an integrated simulation-control system. 3.2 Simulation Step Size and Communication Regarding the step size in simulation calculations, a fixed-step clock control mechanism can be used for general production processes. This is already quite practical for stable systems (such as the example given in this paper). To achieve dynamic adjustment of the step size, additional external computing modules can be added to the system to further realize variable-step optimization. Communication between the simulation system and the external computing modules can generally be achieved using Microsoft's DCOM and MSMQ technologies. Synchronous and asynchronous messages can be sent through DCOM and MSMQ components to form an online and real-time simulation communication environment, realizing distributed computing between simulation and industrial control. The computer network supporting this distributed computing environment can utilize the enterprise's existing industrial control network or a VPN environment based on an Intranet or the Internet. 4. Conclusion The combination of digital simulation and process control systems is an example of using information technology to transform traditional process industries. The process control system collects production data, the simulation system analyzes the data, and establishes a mathematical model of the process; together, they achieve optimal control of the production process. As an intelligent core of the process control system, the computer digital simulation system is playing an increasingly important role in the traditional production control field.