Control system for synthesizing barium titanate by low-temperature hydrothermal method

Abstract

The invention provides a control system for synthesizing barium titanate by a low-temperature hydrothermal method, and relates to the technical field of barium titanate synthesis control systems, and the control system is characterized by comprising a multi-dimensional acquisition module, a control module, an execution module and a safety early warning module, the multi-dimensional acquisition module is used for acquiring parameters such as temperature, pressure, stirring rate, pH value of reaction liquid and raw material feeding rate in the reaction kettle in real time; the control module receives the collected parameters, compares the collected parameters with a preset process threshold value, and outputs a regulation and control instruction to the execution module; the regulation and control instruction comprises a temperature regulation and control instruction, a pressure regulation and control instruction, a stirring speed regulation and control instruction, a pH value regulation and control instruction, a feeding speed regulation and control instruction and an emergency disposal instruction; and the execution module receives the regulation and control instruction and regulates and controls the corresponding parameters. The control system for synthesizing barium titanate through the low-temperature hydrothermal method has the advantages that multi-stage parameters are precisely and dynamically adapted, and abnormal working conditions are classified and safely disposed.

Description

Technical Field

[0001] This application relates to the field of control system technology, and more specifically, to a control system for the low-temperature hydrothermal synthesis of barium titanate. Background Technology

[0002] Barium titanate (BaTiO3), a key ferroelectric and dielectric material, is widely used in electronic information fields such as multilayer ceramic capacitors, piezoelectric elements, and electronic ceramics. The precision of its preparation process directly determines the purity, particle size distribution, and crystal form of the product (tetragonal phase is the optimal crystal form for application). The low-temperature hydrothermal method has become the mainstream technology for barium titanate synthesis due to its advantages such as mild reaction conditions, high product purity, and environmentally friendly process. However, this process has technical limitations that restrict the stability of industrial-scale production and the consistency of product performance. 1. Poor adaptability to multi-stage processes and lag in parameter control: The low-temperature hydrothermal synthesis of barium titanate involves three core stages: precursor dissolution, crystal nucleation, and crystal growth. The requirements for temperature, pressure, and pH control differ significantly at each stage (e.g., strict pH stability is required during crystal nucleation, and precise pressure control is needed during crystal growth). Traditional control systems using fixed parameters or a single PID algorithm cannot dynamically adapt to the characteristics of multi-stage processes, easily leading to problems such as parameter adjustment lag and overshoot, resulting in a large coefficient of variation in product particle size and a high rate of crystal defects.

[0003] 2. The handling of abnormal operating conditions is crude, and it is difficult to balance safety and production: Low temperature hydrothermal synthesis depends on a high temperature and high pressure environment. Abnormal parameters may cause the reaction to run away from control. However, existing technologies mostly adopt a "one-size-fits-all" emergency shutdown strategy. Even minor abnormalities will lead to production interruption and waste of raw materials. In the event of severe abnormalities, only shutdown without rapid cooling poses safety hazards such as equipment overpressure and complete destruction of crystal form.

[0004] The above problems urgently need to be addressed. Summary of the Invention

[0005] The purpose of this application is to provide a control system for the low-temperature hydrothermal synthesis of barium titanate, which has the advantages of precise dynamic adaptation of multi-stage parameters and graded safe handling of abnormal operating conditions.

[0006] Firstly, this application provides a control system for the low-temperature hydrothermal synthesis of barium titanate, the technical solution of which is as follows: It includes a multi-dimensional data acquisition module, a control module, an execution module, and a security early warning module; The multidimensional acquisition module collects parameters such as temperature, pressure, stirring rate, pH value of reaction solution, and raw material feed rate in the reactor in real time. After receiving the collected parameters, the control module compares them with the preset process threshold and outputs control commands to the execution module. The control commands include temperature control commands, pressure control commands, stirring rate control commands, pH value control commands, feed rate control commands, and emergency response commands. The execution module receives control instructions and implements the corresponding parameter control. The safety early warning module monitors the sealing status of the reactor and the extreme values ​​of the system pressure. When the parameters exceed the preset process threshold, it triggers an audible and visual alarm and drives the execution module to perform an emergency shutdown.

[0007] Furthermore, in this application, the execution module includes a temperature regulation submodule, a pressure regulation submodule, a stirring drive submodule, a pH regulation submodule, a feed metering submodule, and an emergency response submodule; The temperature regulation submodule receives temperature regulation commands and implements temperature regulation. The pressure regulation submodule receives pressure regulation commands and implements pressure regulation. The stirring drive submodule receives stirring rate control commands and performs stirring rate adjustment. The pH adjustment submodule receives pH value adjustment commands and performs pH value adjustment; The feed metering submodule receives the feed rate control command and implements the feed rate adjustment; The emergency response submodule receives emergency response instructions and performs emergency buffer injection.

[0008] Furthermore, in this application, the control module includes an improved PID linkage control algorithm submodule, which dynamically adjusts the proportional coefficient, integral time, and derivative time of the PID parameters according to the reaction process. When the parameter deviation is greater than 15% of the preset process threshold, the rapid adjustment mode is activated. The rapid adjustment mode includes increasing the proportional coefficient, shortening the integral time, and shortening the derivative time. When the parameter deviation is less than 5% of the preset process threshold, the mode is switched to the stable adjustment mode. The stable adjustment mode includes maintaining the current proportional coefficient, integral time, and derivative time.

[0009] Furthermore, in this application, the improved PID linkage control algorithm submodule has a built-in fuzzy adaptive adjustment mechanism unit. The working logic of the fuzzy adaptive adjustment mechanism unit is as follows: a fuzzy rule library is pre-constructed, containing 5 fuzzy subsets of parameter deviation inputs (respectively, extremely small, small, medium, large, and extremely large) and 4 fuzzy subsets of deviation change rates (respectively, slowly, steadily, relatively quickly, and rapidly). Differentiated fuzzy rules are configured for the precursor dissolution stage, crystal nucleus formation stage, and crystal growth stage of low-temperature hydrothermal synthesis of barium titanate. During the precursor dissolution stage, when the temperature deviation is "large" and the rate of change is "rapid," the proportional coefficient is automatically adjusted to 1.2-1.5 times the initial value, and the integration time is shortened to 60%-80% of the initial value. During the crystal nucleus formation stage, when the pH deviation is "medium" and the rate of change is "relatively fast", the proportionality coefficient is maintained at its initial value and the derivative time is adjusted to 0.9-1.1 times the initial value; During the crystal growth stage, when the pressure deviation is "small" and the rate of change is "stable", the integral time is extended to 1.1-1.3 times the initial value, and the PID parameters for the corresponding reaction stage are precisely and dynamically adapted.

[0010] Furthermore, in this application, the pH adjustment submodule and the feed metering submodule establish a "feed-premix-pH real-time compensation" linkage logic: When the feed metering submodule delivers titanium source raw materials, the pH adjustment submodule controls the premixing buffer chamber to automatically draw in the corresponding volume of buffer solution according to the molar ratio of titanium source feed to buffer solution of 1:0.9-1.1. The buffer solution is then mixed in the chamber for 1-2 seconds to form a premixed liquid. The premixing buffer chamber is a subordinate control component of the pH adjustment submodule. The premixing buffer chamber is filled with 0.08-0.12 mol / L barium hydroxide buffer solution with the same composition as the barium titanate source. The pH adjustment submodule has a built-in pH sensing unit that collects the pH value of the premixed solution in real time. If the pH of the premixed solution deviates from the target range of 11±0.1 by more than 0.1, the pH adjustment submodule controls the premixed buffer chamber to immediately add buffer solution. After the pH of the premixed solution stabilizes within the target range of 11±0.1 for more than 0.5 seconds, it is then injected into the reaction vessel.

[0011] Furthermore, in this application, the control module also has a built-in reaction process prediction model submodule, which is generated based on historical reaction data. According to the real-time collected temperature and pH value change trends, it predicts the growth rate of barium titanate crystals and the purity of the product, and outputs parameter pre-adjustment instructions in advance.

[0012] Furthermore, in this application, the reaction process prediction model submodule uses complete process data of at least 50 batches of low-temperature hydrothermal synthesis of barium titanate as the training set. The data dimensions include the initial molar ratio of raw materials (Ti:Ba), temperature and pressure / pH time series data of each reaction stage, stirring rate curve, and final particle size distribution / purity of the product. The reaction process prediction model submodule adopts a fusion algorithm of BP neural network and random forest, and the output dimensions include: real-time predicted value of crystal growth rate, estimated value of crystal grain size distribution width, and estimated value of product purity deviation. When the predicted crystal growth rate is less than 10% of the preset baseline value, the reaction process prediction model submodule outputs a pre-adjustment command: increase the stirring rate by 5-8 r / min and fine-tune the reaction temperature by 0.3-0.5℃.

[0013] Furthermore, in this application, the reaction process prediction model submodule has a built-in iterative update unit. After accumulating 20 batches of valid reaction data, the iterative update unit automatically selects process optimization cases from the 20 batches of valid reaction data and retrains the model weights to keep the prediction accuracy stable within ±2%. The valid reaction data refers to the reaction data generated during the low-temperature hydrothermal synthesis of barium titanate, and must simultaneously meet the following requirements: process parameters meet preset thresholds, with deviation time ≤ 2 seconds and amplitude ≤ 3%; the full-process parameter acquisition frequency is ≥ 10 Hz, single batch data loss is ≤ 5 seconds and includes complete control records, status records and alarm records; product purity is ≥ 99.5%, particle size variation coefficient is ≤ 5% and the crystal form is tetragonal; the acquisition sensor has been calibrated within 30 days and is within its validity period, and the measurement error meets the preset accuracy.

[0014] Furthermore, in this application, the safety early warning module includes a built-in abnormal operating condition classification and handling submodule. This submodule classifies abnormalities where parameters exceed preset process thresholds into three levels and matches them with differentiated handling strategies: A parameter deviation of less than 5% from the threshold is considered a Level 1 anomaly: triggering each submodule of the control module to fine-tune the corresponding parameters, while simultaneously recording the anomaly node; A parameter deviation of 5%-10% from the threshold is considered a Level II anomaly: The emergency response submodule controls the injection of 0.1 mol / L potassium hydroxide emergency buffer solution into the reactor, with the injection volume being 0.5%-1.0% of the reaction liquid volume; the stirring drive submodule controls the reduction of the stirring rate to 80% of the original rate. If the parameter deviates from the threshold by more than 10%, it is a Level 3 anomaly: the safety warning module triggers an audible and visual alarm and drives the execution module to perform an emergency shutdown. At the same time, it drives the temperature regulation submodule to control the rapid circulation of the cooling medium in the reactor jacket, reducing the reaction temperature to room temperature within 10 seconds. The abnormal operating condition classification and handling submodule is linked with the reaction process prediction model submodule, and triggers the corresponding level of pre-handling strategy 1-2 minutes in advance based on the predicted parameter change trend.

[0015] Furthermore, in this application, the execution module also includes a data traceability and process replication submodule, which communicates bidirectionally with the control module to extract the complete process parameter sequence corresponding to historical valid reaction data and generate a standardized process parameter package; When a new reaction batch is started, the process parameter package is directly called. The control module automatically synchronizes the preset process threshold to each sub-module to achieve replication of the barium titanate synthesis process under the same process conditions. The data traceability and process replication submodule supports binding and labeling process parameter packages with the performance indicators of corresponding batches of products, and supports rapid process matching according to corresponding performance requirements; the performance indicators include barium titanate purity and particle size.

[0016] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing embodiments of this application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0017] Beneficial effects: 1) Achieve precise dynamic control in multiple stages and improve parameter stability and process adaptability: Through the built-in fuzzy adaptive adjustment mechanism of the improved PID linkage control algorithm submodule, differentiated fuzzy rules are configured for the three stages of precursor dissolution, crystal nucleus formation and crystal growth, and the PID parameters (proportional coefficient, integral time and derivative time) are dynamically adjusted. By switching between "fast adjustment mode (deviation > 15%)" and "stable adjustment mode (deviation < 5%)", precise control of "rapid correction of large deviations and stable maintenance of small deviations" is achieved.

[0018] 2) Stabilizing pH from the source to ensure the quality of crystal nuclei and crystal growth: Through the "feed-premix-real-time pH compensation" linkage logic of the pH adjustment submodule and the feed metering submodule, the acidity of the titanium source is neutralized by premixing with barium hydroxide buffer solution of the same composition as the barium source at a molar ratio of 1:0.9-1.1 before the raw materials are injected into the reactor. Then, real-time monitoring and replenishment ensure that the pH is stable in the range of 11±0.1. This design avoids the risk of sudden pH drop from the source, avoids the lag in adjustment in the reactor and the problem of local over-alkalization, improves the consistency of crystal nuclei formation, reduces the particle size variation coefficient, and ensures a high proportion of tetragonal phase crystals.

[0019] 3) Proactively predicting reaction progress and intervening in advance to avoid performance deviations: The built-in reaction progress prediction model submodule in the control module, based on a BP neural network-random forest fusion algorithm trained with at least 50 batches of complete process data, can predict crystal growth rate, particle size distribution width, and product purity deviations in real time. Combined with an online iterative update function, the model weights are optimized every 20 batches of valid data, ensuring prediction accuracy remains stable within ±2%. Through a proactive intervention mode of "trend prediction - 1-2 minute advance adjustment" (e.g., fine-tuning the temperature by 0.3-0.5℃ and increasing the stirring rate by 5-8 r / min when the predicted growth rate is too low), the lag problem of passive correction in traditional systems is solved, thus improving product purity.

[0020] 4) Graded handling of abnormal operating conditions to balance production safety and continuity: The abnormal operating condition graded handling submodule of the safety early warning module classifies parameter anomalies into three levels and matches them with differentiated strategies: Level 1 anomalies (deviation <5%) involve fine-tuning and correction; Level 2 anomalies (5%-10%) involve buffer injection and slowing down the reaction to stabilize it; and Level 3 anomalies (>10%) involve shutdown and 10-second rapid cooling. This design avoids the production waste caused by the traditional "one-size-fits-all" shutdown, and at the same time, it achieves pre-treatment by linking with the predictive model, reducing the risk of anomaly escalation and balancing production continuity with extreme safety assurance.

[0021] 5) Enables traceable and reusable processes, improving batch consistency and industrial efficiency: Through the data traceability and process replication submodule, historical effective reaction data is encapsulated into standardized process parameter packages and bound to product performance indicators (purity, particle size). New batches can directly call the parameter packages to achieve precise process replication without repeated debugging. This function not only achieves bidirectional traceability of "performance-process" but also shortens the debugging cycle of new batches, reduces industrial production costs, and minimizes product performance deviations between batches, meeting the consistency requirements of large-scale mass production. Attached Figure Description

[0022] Figure 1 An architecture diagram of a control system for the low-temperature hydrothermal synthesis of barium titanate provided in this application; Figure 2 This is an architecture diagram of the execution module of this application; Figure 3 This is an architecture diagram of the control module of this application; Figure 4 This is an architecture diagram of the improved PID linkage control algorithm submodule of this application; Figure 5 This is an architecture diagram of the reaction process prediction model submodule of this application; Figure 6 This is an architecture diagram of the security early warning module in this application. Detailed Implementation

[0023] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0024] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0025] Please refer to Figures 1 to 6 As shown, this application provides a control system for the low-temperature hydrothermal synthesis of barium titanate, including a multi-dimensional acquisition module, a control module, an execution module, and a safety early warning module. The multi-dimensional acquisition module collects parameters such as temperature, pressure, stirring rate, pH value of the reaction solution, and feed rate of the raw materials in real time within the reactor. After receiving the collected parameters, the control module compares them with preset process thresholds and outputs control commands to the execution module. The control commands include temperature control commands, pressure control commands, stirring rate control commands, pH value control commands, feed rate control commands, and emergency response commands. The execution module receives the control commands and implements the corresponding parameter control. The safety early warning module monitors the sealing status of the reactor and the extreme pressure of the system. When the parameters exceed the preset process thresholds, it triggers an audible and visual alarm and drives the execution module to perform an emergency shutdown.

[0026] Specifically, the control system for the low-temperature hydrothermal synthesis of barium titanate provided in this application includes a multi-dimensional acquisition module, a control module, an execution module, and a safety early warning module; The multi-dimensional acquisition module collects real-time parameters such as temperature, pressure, stirring rate, pH value of the reaction solution, and raw material feed rate within the reactor. Its functions are: ① To clarify the core function of the multi-dimensional acquisition module: as the system's "data input terminal," it is responsible for collecting key process parameters of the reaction process, providing real-time and accurate data support for subsequent control logic; ② To limit the acquisition scope and parameter types: the acquisition scope is "within the reactor" (ensuring data relevance and excluding irrelevant environmental parameters), and the acquired parameters are temperature, pressure, stirring rate, pH value of the reaction solution, and raw material feed rate. These parameters are all core influencing factors in the low-temperature hydrothermal synthesis of barium titanate (temperature affects crystal growth, pH affects crystal nucleus formation, and feed rate affects raw material ratio, directly determining product purity and particle size), reflecting the scientific nature of parameter selection; ③ To emphasize "real-time acquisition": providing a prerequisite for subsequent "precise control" (only real-time data can ensure the control module responds promptly to parameter fluctuations), echoing the closed-loop control logic of the control module.

[0027] After receiving the collected parameters, the control module compares them with the preset process thresholds and outputs control commands to the execution module. These control commands include temperature control commands, pressure control commands, stirring rate control commands, pH value control commands, feed rate control commands, and emergency response commands. Their functions are: ① Defining the core logic of the control module: As the system's "core processing unit," it undertakes the core function of "data processing - command output," constructing a closed-loop control foundation of "collecting parameters → comparing thresholds → outputting commands." The preset process thresholds are benchmark values ​​determined based on the optimal synthesis process of barium titanate. After comparison, the output control commands achieve the core objective of "parameter deviation → precise correction." ② Clarifying the types of control commands: Six types of control commands are defined, covering both normal process parameter adjustments (temperature, pressure, etc.) and emergency response commands, achieving full coverage of commands for both routine control and emergency response. ③ Connecting module logic: It connects to the multi-dimensional acquisition module (receiving data) and the execution module (outputting commands), while also connecting to the emergency needs of the safety warning module through "emergency response commands," providing a data foundation for subsequent module linkage (such as safety warnings triggering emergency commands).

[0028] The execution module receives control commands and implements corresponding parameter control: its function is to ① clarify the core function of the execution module: as the "command execution end" of the system, it transforms the abstract commands output by the control module into specific process parameter control actions (such as adjusting the reaction temperature after receiving a temperature control command), and is the direct carrier for achieving "process parameter stability"; ② strengthen closed-loop control: it receives commands from the control module, completes the implementation of "command → action", and ensures that the control logic forms a closed loop (acquisition → processing → execution → re-acquisition).

[0029] The safety early warning module monitors the sealing status of the reactor and the extreme pressure of the system. When the parameters exceed the preset process threshold, it triggers an audible and visual alarm and drives the execution module to perform an emergency shutdown. Its functions are: ① To clarify the core function of the safety early warning module: as the "risk control end" of the system, it is responsible for monitoring key safety risk points in the reaction process to avoid reaction runaway (low-temperature hydrothermal synthesis must be carried out in a closed high-pressure environment; sealing failure and pressure exceeding limits may lead to safety accidents); ② To limit the monitoring scope: focusing on the sealing status (preventing raw material leakage and pressure instability) and the extreme pressure of the system (core safety risk parameters), reflecting the targeted nature of risk control; ③ To define an emergency response mechanism: clearly defining that exceeding the threshold triggers a dual emergency action of audible and visual alarm and emergency shutdown. The audible and visual alarm is used to warn operators, and the emergency shutdown is used to terminate the dangerous reaction. At the same time, the "driving the execution module to perform an emergency shutdown" achieves linkage with the execution module to ensure the implementation of emergency actions; ④ To complete the system logic loop: covering all scenarios of normal control and abnormal emergency, avoiding the neglect of safety risks due to only focusing on process parameter adjustment, making the logic of the entire control system more complete and rigorous.

[0030] In some preferred embodiments, the execution module includes a temperature regulation submodule, a pressure regulation submodule, a stirring drive submodule, a pH regulation submodule, a feed metering submodule, and an emergency response submodule; the temperature regulation submodule receives a temperature control command and performs temperature regulation; the pressure regulation submodule receives a pressure control command and performs pressure regulation; the stirring drive submodule receives a stirring rate control command and performs stirring rate regulation; the pH regulation submodule receives a pH value control command and performs pH value regulation; the feed metering submodule receives a feed rate control command and performs feed rate regulation; and the emergency response submodule receives an emergency response command and performs an emergency buffer solution injection operation.

[0031] Specifically, the execution module includes a temperature regulation submodule, a pressure regulation submodule, a stirring drive submodule, a pH regulation submodule, a feed metering submodule, and an emergency response submodule; The temperature regulation submodule receives and implements temperature control commands: its function is to clearly define the temperature regulation submodule as the sole executor of temperature control commands, focusing on temperature parameter adjustment, forming a closed loop with the temperature parameters collected by the multi-dimensional acquisition module (acquisition → command → adjustment); the execution module's implementation of corresponding parameter control is refined into a specific division of labor for temperature regulation; temperature is the core influencing factor in the low-temperature hydrothermal synthesis of barium titanate, directly determining the crystal growth rate and stability, and the functional definition of the temperature regulation submodule directly serves the core technical goal of precise temperature control.

[0032] The pressure regulation submodule receives pressure control commands and implements pressure regulation: its function is to define the exclusive function of the pressure regulation submodule as pressure parameter regulation in accordance with the pressure control commands, forming a closed loop with the pressure parameters collected by the multi-dimensional acquisition module; low-temperature hydrothermal synthesis needs to be carried out in a closed high-pressure environment, and pressure fluctuations will lead to uneven crystal size distribution. The function definition of the pressure regulation submodule addresses this process pain point, and ensures the accuracy of pressure regulation through a dedicated submodule.

[0033] The stirring drive submodule receives and implements stirring rate control commands: its function is to determine the stirring rate control command and form a closed loop with the stirring rate parameters of the multi-dimensional acquisition module. In low-temperature hydrothermal synthesis, uneven mixing of the reaction solution can lead to local raw material concentration imbalance, affecting the consistency of barium titanate crystal nuclei formation. The stirring drive submodule ensures uniform mixing of the reaction solution by precisely adjusting the stirring rate, thereby indirectly improving product performance.

[0034] The pH adjustment submodule receives and implements pH adjustment commands. Its function is to define the pH adjustment submodule's role as pH adjustment, forming a closed loop with the pH parameter of the reaction solution from the multi-dimensional acquisition module. pH is a decisive factor in the formation of barium titanate crystal nuclei (optimal range 11±0.1). The functional definition of the pH adjustment submodule directly addresses the core requirement of precise pH control, providing a core framework for the subsequent "feed-premix-real-time pH compensation" workflow. This enhances functional specificity by avoiding interference between pH adjustment and other parameter adjustments, thus improving control accuracy.

[0035] The feed metering submodule receives and implements feed rate control commands: its function is to regulate the feed rate in accordance with the feed rate control commands, forming a closed loop with the raw material feed rate parameters of the multi-dimensional acquisition module; ensuring accurate raw material ratios: in low-temperature hydrothermal synthesis, the stability of the Ti / Ba molar ratio directly affects the purity of the product. The feed metering submodule ensures that the raw materials are fed at a uniform rate according to the preset ratio by accurately adjusting the feed rate, avoiding ratio imbalance; supporting linkage logic: it provides a prerequisite for the linkage between the subsequent pH adjustment submodule and the feed metering submodule. Only by clarifying the function of the feed metering submodule can the accurate matching of feed amount and buffer injection amount be achieved.

[0036] The emergency response submodule receives emergency response instructions and performs emergency buffer injection. Its function is to clearly define the emergency response submodule's function as emergency buffer injection, corresponding to the emergency response instructions, and to link with the emergency needs of the safety early warning module. Focusing on safety emergency scenarios, unlike the other five conventional parameter adjustment submodules, the emergency response submodule specifically handles abnormal operating conditions (such as parameters exceeding thresholds). By injecting emergency buffer, it inhibits further deterioration of parameters and serves as the core execution carrier for subsequent three-level abnormal response strategies. Completing the system's functional dimensions, it achieves full coverage of the execution functions of conventional control (5 submodules) and emergency response (1 submodule), extending the logic of the entire control system from normal process adjustment to abnormal risk prevention and control, and more comprehensively meeting actual production needs.

[0037] In some preferred embodiments, the control module includes an improved PID linkage control algorithm submodule, which dynamically adjusts the proportional coefficient, integral time, and derivative time of the PID parameters according to the reaction process; when the parameter deviation is greater than 15% of the preset process threshold, a rapid adjustment mode is activated, which includes increasing the proportional coefficient, shortening the integral time, and shortening the derivative time; when the parameter deviation is less than 5% of the preset process threshold, a stable adjustment mode is switched to, which includes maintaining the current proportional coefficient, integral time, and derivative time.

[0038] Specifically, the control module includes an improved PID linkage control algorithm submodule. This submodule dynamically adjusts the proportional coefficient, integral time, and derivative time of the PID parameters according to the reaction process: ① It clarifies the algorithm type and core parameters: An improved PID linkage control algorithm is adopted, with the core adjustment objects being the three core PID parameters (proportional coefficient P, integral time I, and derivative time D)—these three parameters directly determine the control response speed, steady-state accuracy, and anti-interference capability; ② It emphasizes dynamic adjustment according to the reaction process: This overcomes the shortcomings of traditional fixed PID parameters and adapts to the multi-stage characteristics of low-temperature hydrothermal synthesis of barium titanate. The different response requirements for parameter control during stages such as precursor dissolution, crystal nucleus formation, and crystal growth (e.g., smooth control is needed during crystal nucleus formation to avoid damaging the crystal nucleus, while precise temperature control is needed during crystal growth) provide a foundation for subsequent staged fuzzy adaptive adjustment; ③ Echoing the logic of the preceding modules: The dynamic adjustment of the improved PID linkage control algorithm submodule relies on the real-time parameters of the multi-dimensional acquisition module (acquiring dynamic data of the reaction process), and the output precise control instructions directly correspond to the 6 submodules of the execution module (such as the temperature adjustment submodule and the pH adjustment submodule), realizing a complete closed loop of "data acquisition → dynamic algorithm → precise instructions → submodule execution".

[0039] When the parameter deviation exceeds 15% of the preset process threshold, a rapid adjustment mode is activated. This rapid adjustment mode includes increasing the proportional gain, shortening the integral time, and shortening the derivative time. When the parameter deviation is less than 5% of the preset process threshold, the mode switches to a stable adjustment mode. This stable adjustment mode maintains the current proportional gain, integral time, and derivative time. Its function is to quantify the mode switching threshold: using the preset process threshold as a reference, it clarifies the triggering conditions for the two adjustment modes (15% is the rapid adjustment threshold, and 5% is the stable adjustment threshold). The core objective of distinguishing between the two modes is: ① Rapid adjustment mode (deviation > 15%). ① **Stable Adjustment Mode (deviation < 5%):** For scenarios with significant parameter deviations, priority is given to ensuring response speed, quickly pulling parameters back to the threshold range to prevent the deviation from expanding and causing product quality deterioration (e.g., a sudden temperature rise may lead to abnormal crystal form). ② **Stable Adjustment Mode (deviation < 5%):** For scenarios with small parameter fluctuations, priority is given to ensuring stable control to avoid over-adjustment leading to new parameter fluctuations (e.g., during the crystal nucleation stage, gentle adjustment during small pH fluctuations can prevent crystal nucleus breakage). This deviation-level mode switching solves the core pain point of traditional PID algorithms: fast response leads to overshoot, while stable response leads to slow response, improving the control adaptability in complex reaction scenarios. Furthermore, it reserves space for emergency scenarios: when parameter deviation exceeds 15% and continues to expand, subsequent abnormal conditions can be linked for graded handling (e.g., deviation > 10% triggers level 2 / 3 anomalies), realizing a gradient control logic of routine algorithm adjustment and emergency safety warning; supporting product quality targets: the precise switching between the two modes directly serves the product requirements of "barium titanate purity ≥ 99.5%, particle size variation coefficient ≤ 5%" (the criteria for determining effective reaction data), ensuring process stability through algorithm optimization and indirectly improving product consistency.

[0040] For example, the improved PID linkage control algorithm submodule adopts a positional PID control model, which dynamically calculates and controls the output through the following core formula, and then dynamically adjusts the proportional coefficient, integral time and derivative time of the PID parameters according to the reaction process. ; Where u(t) represents the control output at time t, corresponding to the control signal of the execution module, such as the heating power of the temperature control submodule, the buffer injection flow rate of the pH control submodule, etc.; K p (t) represents the dynamic proportionality coefficient at time t (dynamically adjusted according to the reaction progress and parameter deviations, one of the core adjustable parameters); T i (t) represents the dynamic integral time at time t (dynamically adjusted according to the reaction progress and parameter deviations; one of the core adjustable parameters); T d(t) represents the dynamic differential time at time t (dynamically adjusted according to the reaction progress and parameter deviation, the third core adjustable parameter); e(t) represents the parameter deviation at time t (the difference between the real-time acquired value and the preset process threshold, e(t)=r(t)-y(t), where r(t) is the preset process threshold at time t, and y(t) is the real-time acquired value of the multi-dimensional acquisition module at time t. The acquired parameters include the temperature, pressure, stirring rate, pH value of the reaction solution and the feed rate of the raw materials in the reactor). This represents the intensity of the integral action per unit time. The larger the value, the faster the integral eliminates steady-state errors. The intensity of the integral action is dynamically adjusted according to the control requirements of different process stages to achieve a balance between accurately eliminating steady-state errors and avoiding parameter overshoot. For example, in the precursor dissolution stage, its value is increased to quickly eliminate temperature steady-state errors and ensure rapid precursor dissolution; in the crystal nucleation stage, its value is kept moderate to smoothly eliminate pH deviations and avoid crystal nucleus breakage or uneven particle size; in the crystal growth stage, its value is decreased to weaken the integral action and avoid pressure overshoot that could lead to abnormal crystal growth. This represents the integral term representing the parameter deviation before time t (used to eliminate steady-state error). This integral term is the accumulation of all parameter deviations from the start of the reaction to the current time. Its role in eliminating steady-state error can be analyzed from both the principle and process scenario perspectives: From the control principle perspective, the proportional term Kp(t)·e(t) can only output a control signal based on the current deviation and cannot completely eliminate steady-state error (for example, a small deviation of 0.2℃ may still exist after the temperature stabilizes). However, this integral term continuously accumulates historical deviations. As long as the deviation is not zero, the accumulated value of the integral term will continuously change, outputting a continuous control signal until the deviation is completely zero, thus eliminating steady-state error at its source. From the process scenario perspective, during the crystal growth stage, if the reaction pressure remains slightly below the preset threshold (a steady-state deviation exists), the integral term will accumulate this deviation, continuously driving the pressure regulation submodule to fine-tune the airflow until the pressure is precisely stabilized at the target value, avoiding problems such as uneven crystal growth rate and excessive particle size variation coefficient due to pressure steady-state deviation. In summary, this achieves the purpose of eliminating steady-state error. The differential term represents the parameter deviation at time t (used to predict the trend of deviation change and suppress overshoot). This differential term is the rate of change of the parameter deviation with respect to time, reflecting the "speed and direction of change" of the deviation. Its role in predicting the trend and suppressing overshoot is as follows: From the perspective of control principles, the differential term does not depend on the current magnitude of the deviation, but rather focuses on the trend of deviation change. When the deviation increases rapidly (e.g., the temperature rises rapidly from 160℃ to 165℃), the differential term will output a reverse control signal in advance (e.g., reduce the heating power in advance) to suppress further expansion of the deviation, thereby avoiding parameter "overshoot" (i.e., exceeding the preset threshold and then reverting back), and improving system stability. From the perspective of process scenarios, during the crystal nucleation stage, if the pH value rises rapidly (large rate of change of deviation), the differential term will predict that the pH will overshoot and drive the pH adjustment submodule in advance to reduce the amount of buffer solution injected, avoiding problems such as crystal nucleus breakage and uneven particle size distribution caused by excessive pH fluctuations, and ensuring the stability of crystal nucleus formation.

[0041] The integral term addresses the problem of "unable to eliminate steady-state error" by "accumulating historical deviations," while the derivative term addresses the problem of "parameter overshoot and excessive fluctuations" by "predicting deviation trends." Together with the proportional term, they achieve precise control of PID, thus meeting the stability requirements of the multi-stage process for low-temperature hydrothermal synthesis of barium titanate.

[0042] It should be noted that, considering the process characteristics of the low-temperature hydrothermal synthesis of barium titanate, initial baseline values ​​are first set: Initial value of dynamic scaling factor: K p0 =2.0~3.0 (adapted to temperature / pressure / pH control sensitivity of Ti / Ba systems); Initial value of dynamic integration time: T i0 =40~80s (equilibrium steady-state error elimination rate and system stability); initial value of dynamic derivative time: T d0 =10~20s (basic sensitivity for suppressing parameter overshoot).

[0043] Quantitative triggering of parameter adjustments is implemented for different process stages: During the precursor dissolution stage, the core control parameter is temperature. When e(t) > 0.15·r(t) and the rate of change is rapid, the corresponding deviation is "large," triggering a rapid adjustment mode. p (t) = 1.2~1.5Kp0 (strengthening the correction force), T i (t) = 0.6~0.8Ti0 (accelerates the elimination of steady-state error), T d (t) = 0.7~0.9Td0 (weakening the differential, prioritizing rapid pullback); During the crystal nucleation stage, the core control parameter is pH. When e(t) ≤ 0.15·r(t) and the rate of change is relatively fast, the corresponding deviation is "medium," and a stable adjustment mode is initiated.p (t)=K p0 (Maintain the initial value to avoid over-adjustment), T i (t) = 0.9~1.1T i0 (Smoothly eliminates error), T d (t) = 0.9 ~ 1.1T d0 (Fine-tuning the derivative to suppress fluctuations); During the crystal growth stage, the core control parameter is pressure. When e(t) ≤ 0.05·r(t) and the rate of change is stable, the corresponding deviation is "small," and a stable adjustment mode is initiated. p (t) = 0.9~1.1K p0 (Fine-tuning the ratio to avoid overshoot), T i (t) = 1.1 ~ 1.3T i0 (Extend the integral to mitigate error elimination), T d (t) = 0.9 ~ 1.2T d0 (Strengthen differentiation to suppress fluctuations in advance).

[0044] Among them, "rapid" deviation change rate means parameter change rate ≥ 0.5% / s; "relatively fast" deviation change rate means 0.2% / s ≤ parameter change rate < 0.5% / s; "stable" deviation change rate means 0.05% ≤ parameter change rate < 0.2% / s; and "slow" deviation change rate means parameter change rate < 0.05% / s.

[0045] When the parameter deviation |e(t)| > 0.15·r(t), that is, the parameter deviation is greater than 15% of the preset process threshold, the improved PID linkage control algorithm submodule starts the fast adjustment mode by increasing the proportional coefficient K. p (t), shortening the integration time T i (t) Improve the control response speed and quickly pull the parameters back to the preset process threshold range; When the parameter deviation |e(t)| < 0.05·r(t), that is, when the parameter deviation is less than 5% of the preset process threshold, the improved PID linkage control algorithm submodule switches to the smooth adjustment mode, and fine-tunes the proportional coefficient K. p (t), extending the integration time T i (t) and the optimized differential time T d (t) is used to avoid excessive fluctuations in parameters and ensure the stability of the reaction process.

[0046] In some preferred embodiments, the improved PID linkage control algorithm submodule has a built-in fuzzy adaptive adjustment mechanism unit. The working logic of the fuzzy adaptive adjustment mechanism unit is as follows: a fuzzy rule library is pre-constructed, containing 5 fuzzy subsets of parameter deviation inputs (respectively, extremely small, small, medium, large, and extremely large) and 4 fuzzy subsets of deviation change rates (respectively, slowly, steadily, relatively fast, and rapidly). Differentiated fuzzy rules are configured for the precursor dissolution stage, crystal nucleus formation stage, and crystal growth stage of low-temperature hydrothermal synthesis of barium titanate. In the precursor dissolution stage... When the temperature deviation is "large" and the rate of change is "rapid", the proportional coefficient is automatically adjusted to 1.2-1.5 times the initial value, and the integral time is shortened to 60%-80% of the initial value. During the crystal nucleation stage, when the pH deviation is "medium" and the rate of change is "relatively fast", the proportional coefficient is maintained at the initial value, and the derivative time is adjusted to 0.9-1.1 times the initial value. During the crystal growth stage, when the pressure deviation is "small" and the rate of change is "stable", the integral time is extended to 1.1-1.3 times the initial value, implementing precise dynamic adaptation of PID parameters for the corresponding reaction stage.

[0047] Specifically, the improved PID linkage control algorithm submodule has a built-in fuzzy adaptive adjustment mechanism unit. The working logic of the fuzzy adaptive adjustment mechanism unit is as follows: A fuzzy rule base is pre-constructed, containing five fuzzy subsets of parameter deviation inputs (minimum, small, medium, large, and maximum, respectively) and four fuzzy subsets of deviation change rates (slow, stable, relatively fast, and abrupt, respectively): ① Quantifying the parameter state dimension: The abstract parameter deviation is divided into five fuzzy subsets, and the deviation change rate is divided into four fuzzy subsets. This solves the defect of traditional PID that cannot accurately describe the dynamic change trend of parameters. Through multi-dimensional subset division, large deviations can be accurately captured. ① Combination of small and fast-changing states (e.g., "large deviation and rapid change" or "small deviation and stable change"); ② Establishment of the core carrier of the rule base: The fuzzy rule base is the core of the algorithm decision-making, which transforms process experience (e.g., "pH cannot fluctuate drastically during the crystal nucleation stage") into executable algorithm rules, providing a logical basis for subsequent staged adjustment; ③ The integration of fuzzy logic and PID breaks through the limitations of traditional PID "pure mathematical model control", introduces "experience rule drive", adapts to the process characteristics of "multi-factor coupling and nonlinearity" in low-temperature hydrothermal synthesis (e.g., temperature, pH, and pressure affect each other), and improves the algorithm's anti-interference ability and adaptability.

[0048] Furthermore, differentiated fuzzy rules are configured for the precursor dissolution stage, crystal nucleation stage, and crystal growth stage of barium titanate synthesis at low temperature hydrothermal temperature. Anchoring the process stage, this achieves a deep binding between the algorithm and the process. The design basis of the fuzzy rule library is clearly the three core process stages of barium titanate synthesis at low temperature hydrothermal temperature (precursor dissolution, crystal nucleation, and crystal growth)—the reaction mechanisms and control requirements of these three stages are completely different (e.g., precursor dissolution requires rapid heating, crystal nucleation requires stable pH, and crystal growth requires precise pressure control). The differentiated rule design solves the pain point of traditional PID using a single set of parameters, which cannot adapt to multi-stage processes. Simultaneously, it echoes the temperature / pressure / pH time-series data of each reaction stage in the subsequent reaction process prediction model submodule, strengthening the linkage logic of the entire system's "process-algorithm-data". Secondly, the rule design logic is clarified, laying the groundwork for subsequent specific rules and limiting the applicable scenarios of the differentiated fuzzy rules, avoiding blind design of the rule library. It also provides a framework constraint for the specific rules of the three stages below. Each rule must revolve around the core control objectives of the corresponding stage (dissolution efficiency, crystal nucleation stability, and crystal uniformity), ensuring that the algorithm serves the process requirements.

[0049] During the precursor dissolution stage, when the temperature deviation is "large" and the rate of change is "rapid," the proportional coefficient is automatically adjusted to 1.2-1.5 times the initial value, and the integration time is shortened to 60%-80% of the initial value. This visualizes the control logic of the precursor dissolution stage, specifically: ① Locking in the core control parameters: The core objective of the precursor dissolution stage is to rapidly and fully dissolve the titanium / barium source raw materials. Temperature is a key influencing factor (insufficient temperature will lead to incomplete dissolution, affecting subsequent crystal nucleus formation). Therefore, temperature deviation is selected as the core control object, corresponding to the temperature parameters and temperature adjustment submodule of the multi-dimensional acquisition module; ② Designing parameter adjustment logic: A large deviation and rapid change mean that the temperature is far below the preset threshold and is still decreasing (or has not reached the target and the temperature rises slowly). At this time, the proportional coefficient (P×1.2-1.5) is amplified to improve the response speed, and the integration time (I×60%-80%) is shortened to accelerate the elimination of steady-state error, achieve rapid temperature rise to the target range, and ensure dissolution efficiency; ③ Quantifying the adjustment range: The adjustment multiples of the P and I parameters are clearly defined to avoid vague expressions and make the algorithm feasible.

[0050] During the nucleus formation stage, when the pH deviation is "medium" and the rate of change is "relatively fast," the proportional coefficient is maintained at its initial value, and the derivative time is adjusted to 0.9-1.1 times the initial value. This adapts to the stable control requirements of the nucleus formation stage. Specifically, it serves to: ① Lock in the core control parameters: The nucleus formation stage is the critical period for the crystal form and particle size of barium titanate products. pH fluctuations can lead to nucleus breakage or uneven size (optimal pH range 11±0.1). Therefore, pH deviation is chosen as the core control object, corresponding to the pH parameter of the reaction solution and the pH adjustment submodule; ② Design parameter adjustment logic: A moderate deviation and a relatively fast change mean that the pH deviates from the target range but does not exceed the standard. At this time, the proportional coefficient (P remains unchanged) is maintained to avoid over-adjustment that could cause a pH rebound. The derivative time (D×0.9-1.1) is finely adjusted to enhance the system's anti-interference ability, maintain pH stability, and ensure the consistency of nucleus formation; ③ Respond to the process objectives: The parameter adjustment logic directly serves the subsequent quality indicator of product particle size variation coefficient ≤5%, improving product uniformity through stable control.

[0051] During the crystal growth stage, when the pressure deviation is "small" and the rate of change is "stable," the integral time is extended to 1.1-1.3 times the initial value. This implements precise dynamic adaptation of the PID parameters for the corresponding reaction stage, matching the "precise pressure control" requirement of the crystal growth stage. Specifically, this process serves several purposes: ① Locking in core control parameters: The crystal growth stage requires maintaining a stable high-pressure environment (pressure fluctuations can lead to uneven crystal growth rates, affecting particle size distribution). Therefore, pressure deviation is chosen as the core control object, corresponding to the pressure parameters and the pressure adjustment submodule, forming a closed loop of "acquisition-algorithm-execution"; ② Designing parameter adjustment logic: Small deviation and stable change mean the pressure is close to the target threshold. Extending the integral time (I×1.1-1.3) slows down the integral action, avoiding frequent adjustments due to small deviations, maintaining a steady pressure state, and providing a stable environment for uniform crystal growth; ③ Summarizing the core value of the algorithm: Through a combination of staged, multi-parameter, and fuzzy rule design, precise dynamic adaptation of PID parameters is ultimately achieved, directly responding to the abstract logic of dynamically adjusting PID parameters according to the reaction process, and translating it into a concrete and executable technical solution.

[0052] In some preferred embodiments, the pH adjustment submodule and the feed metering submodule establish a "feed-premix-pH real-time compensation" linkage logic: when the feed metering submodule delivers titanium source raw materials, the pH adjustment submodule controls the premixing buffer chamber to automatically draw in the corresponding volume of buffer solution according to the molar ratio of titanium source feed to buffer solution of 1:0.9-1.1, and form a premixed solution through turbulent mixing in the chamber for 1-2 seconds; the premixing buffer chamber is a subordinate control component of the pH adjustment submodule, and the premixing buffer chamber is filled with 0.08-0.12mol / L barium hydroxide buffer solution with the same composition as the barium titanate source; the pH sensing unit built into the pH adjustment submodule collects the pH value of the premixed solution in real time. If the pH of the premixed solution deviates from the target range of 11±0.1 by more than 0.1, the pH adjustment submodule controls the premixing buffer chamber to immediately add buffer solution. After the pH of the premixed solution stabilizes within the target range of 11±0.1 for more than 0.5 seconds, it is then injected into the reactor.

[0053] Specifically, the pH adjustment submodule and the feed metering submodule establish a "feed-premixing-real-time pH compensation" linkage logic: This logic connects the functions of two independent submodules. Previously, only the "independent control functions" (pH adjustment and feed rate adjustment) of the two submodules were clarified. This section fills the logical gap regarding how the pH adjustment submodule and the feed metering submodule coordinate and control, achieving full-process coordination of "feed amount → pH pre-control → real-time compensation," reflecting the refined control of the execution module. Furthermore, the essence of this linkage logic is to proactively avoid pH fluctuations. In low-temperature hydrothermal synthesis, the addition of titanium source raw materials (mostly acidic) can easily cause a sudden drop in the pH of the reaction solution, directly affecting crystal nucleus formation (as previously stated, a stable pH is required during the crystal nucleus formation stage). Therefore, by linking feed and pH adjustment, pH pre-stabilization is completed before the raw materials enter the reactor, solving the pain points of delayed response and easy fluctuation in direct pH adjustment within the traditional reactor, providing a prerequisite for the stability of subsequent product purity and particle size.

[0054] When the feed metering submodule delivers the titanium source material, the pH adjustment submodule controls the premixing buffer chamber to automatically draw in the corresponding volume of buffer solution according to the molar ratio of titanium source feed to buffer solution of 1:0.9-1.1. The buffer solution is then turbulently mixed within the chamber for 1-2 seconds to form a premixed solution. The premixing buffer chamber is a subordinate control component of the pH adjustment submodule, filled with 0.08-0.12 mol / L barium hydroxide buffer solution of the same composition as the barium titanate source. This embodies the pre-control logic of "feed-premixing," specifically serving the following functions: ① Locking the linkage trigger condition: Using the delivery of the titanium source material by the feed metering submodule as a trigger signal, the feeding start and premixing are synchronized, preventing the raw material from entering the reactor alone and causing a sudden pH change. Since the titanium source is an acidic raw material (such as tetrabutyl titanate), direct injection would disrupt the reaction. ① pH balance: Premixing can neutralize acidity in advance and stabilize pH from the source; ② Quantify the premixing ratio and mixing parameters: Molar ratio of 1:0.9-1.1 ensures precise acidity matching between the buffer and the titanium source (neutralizing acidity without over-mixing), and 1-2 second turbulent mixing ensures premixing uniformity (turbulent mixing can quickly eliminate local concentration differences and avoid pH unevenness in the premixed solution); Limit the concentration of the same component (barium hydroxide) of the buffer and the barium source to +0.08-0.12 mol / L: Barium hydroxide is chosen to avoid introducing impurities (other buffers may introduce impurities such as sodium and potassium, affecting the purity of barium titanate), and the concentration range takes into account the buffering capacity (too low a concentration cannot stabilize pH, too high a concentration can easily lead to excessive barium source), directly serving the quality target of product purity ≥99.5% in subsequent steps.

[0055] The pH adjustment submodule's built-in pH sensor unit collects the premixed solution's pH value in real time. If the premixed solution's pH deviates from the target range of 11±0.1 by more than 0.1, the pH adjustment submodule controls the premixing buffer chamber to immediately add buffer solution. Once the premixed solution's pH stabilizes within the target range of 11±0.1 for more than 0.5 seconds, it is then injected into the reaction vessel. This constructs a closed-loop control system of "real-time compensation - stable injection," specifically: ① Defining the subject and range of pH detection: The pH adjustment submodule's built-in pH sensor unit is responsible for collecting the premixed solution's pH. This needs to be distinguished from the multi-dimensional acquisition module's pH collection within the reaction vessel: the former targets the premixing stage (outside the reaction vessel), while the latter targets the reaction stage (inside the reaction vessel). This clear division of labor avoids confusion regarding the collection functions; ② Quantifying the compensation trigger condition and stability standard: specifically, a deviation from the target range of 11±0.1 by more than 0.1. ① Define the trigger threshold for pH compensation (i.e., start compensation when pH < 10.9 or > 11.1) to avoid frequent adjustments caused by minor fluctuations, balancing control precision and efficiency; ② Ensure the premixed solution is completely stable before injection: This ensures the premixed solution enters the reactor only after pH is fully stable, preventing unstable premixed solution from disrupting the pH balance within the reactor. The crystal nucleation stage is extremely sensitive to pH fluctuations, and this standard directly guarantees the consistency of crystal nucleation, echoing the particle size variation coefficient ≤ 5% in subsequent steps; ③ Complete the linkage logic closed loop: From the feed metering submodule delivering raw materials → pH adjustment submodule controlling premixing → real-time detection and compensation → stable injection into the reactor, a complete closed loop of "feed-premixing-compensation-injection" is formed, transforming the independent functions of the feed metering submodule and pH adjustment submodule into collaborative control capabilities, significantly improving the accuracy and stability of pH control.

[0056] In some preferred embodiments, the control module also includes a reaction process prediction model submodule, which is generated based on historical reaction data. According to the real-time collected temperature and pH value change trends, it predicts the growth rate of barium titanate crystals and the purity of the product, and outputs parameter pre-adjustment instructions in advance.

[0057] Specifically, the control module also has a built-in reaction process prediction model sub-module: the introduction of the reaction process prediction model is different from the pure algorithm control of traditional control systems, which reflects a technological breakthrough. The reaction process prediction model realizes the prediction of process process and solves the pain point of the lagging adjustment of traditional PID algorithm (such as the adjustment is started only when the parameters have deviated from the threshold, which can easily lead to fluctuations in product performance).

[0058] The reaction process prediction model submodule is trained based on historical reaction data. It clarifies the model's training foundation and data source, specifically: ① Defining the training data type: Historical reaction data points to valid reaction data for subsequent steps, ensuring high-quality training data (no abnormal data, process compliance, product compliance), avoiding insufficient prediction accuracy due to low-quality data; ② Reflecting the data acquisition and storage logic: Acquiring historical reaction data relies on the real-time acquisition function of the multi-dimensional acquisition module (providing time-series data). Subsequent data tracing and process replication submodules can assist in data storage and retrieval, forming a closed loop of "data acquisition → storage → training → prediction". Furthermore, to reflect the model's process adaptability, the reaction process prediction model submodule is trained on specific historical data for the low-temperature hydrothermal synthesis of barium titanate, rather than general chemical reaction data. This ensures the model can accurately capture the nonlinear characteristics of this process (such as the coupling relationship between temperature / pH and crystal growth), avoiding the problem of poor adaptability of general models and directly serving the specific process requirements of barium titanate synthesis.

[0059] Based on real-time collected temperature and pH value trends, the growth rate and product purity of barium titanate crystals are predicted. The "input-output" logic of the reaction process prediction model submodule is clearly defined, binding it to the core parameters of the preceding process. Its function is twofold: ① To limit the model input: real-time collected temperature and pH value trends—temperature and pH are core process parameters (from acquisition to staged control, and then to precise stabilization). Their trends directly reflect the reaction process (e.g., stable pH → stable crystal nucleus formation, rising temperature → accelerated crystal growth). Selecting these two parameters as inputs reflects the design logic of a strong correlation between input parameters and process mechanism, ensuring prediction accuracy; ② To limit the model output: barium titanate crystal growth rate and product purity. These two indicators are the core product performance targets of low-temperature hydrothermal synthesis (claim 8 includes them in the criteria for valid reaction data). The reaction process prediction model submodule directly predicts the core performance, rather than indirect parameters, achieving a direct logic of "process parameter trend → product performance prediction," providing clear targets for advance control. To fill the gap in the correlation between process progress and product performance, traditional control systems can only monitor real-time parameters and cannot establish a correlation between parameter trends and product performance. This is achieved through the reaction process prediction model submodule, which upgrades the control system from focusing on parameters to focusing on product performance. This is more in line with the needs of actual production, which is centered on product quality, and improves the practicality of the technical solution.

[0060] Furthermore, it outputs parameter pre-adjustment instructions in advance: clearly defining the core function of the model as early intervention, completing the closed loop of control logic. Parameter pre-adjustment instructions differ from the real-time adjustment instructions of PID algorithms: the former adjusts parameters in advance when it anticipates that the product performance may not meet the standards (e.g., raising the temperature in advance if the crystal growth rate is predicted to be too slow), while the latter corrects parameters in real time when they deviate from the threshold. The two form a dual control system of early prediction to prevent deviation and real-time correction to correct deviation, completely solving the pain point of lagging adjustment in traditional control systems. Connecting to the execution module and maintaining the consistency of the instruction system, parameter pre-adjustment instructions are an extension of control instructions (belonging to the same instruction type output by the control module as temperature, pressure, and other control instructions), directly pointing to the corresponding sub-module of the execution module (e.g., temperature pre-adjustment instructions correspond to the temperature control sub-module, and stirring rate pre-adjustment instructions correspond to the stirring drive sub-module), ensuring a logical closed loop of "model prediction → instruction output → sub-module execution".

[0061] In some preferred embodiments, the reaction process prediction model submodule uses complete process data of at least 50 batches of low-temperature hydrothermal synthesis of barium titanate as a training set. The data dimensions include the initial molar ratio of raw materials (Ti:Ba), temperature, pressure / pH time series data of each reaction stage, stirring rate curve, and final particle size distribution / purity of the product. The reaction process prediction model submodule adopts a fusion algorithm of BP neural network and random forest, and the output dimensions include: real-time predicted value of crystal growth rate, estimated value of crystal particle size distribution width, and estimated value of product purity deviation. When the predicted crystal growth rate is lower than 10% of the preset benchmark value, the reaction process prediction model submodule outputs a pre-adjustment instruction: increase the stirring rate by 5-8 r / min, and at the same time fine-tune the reaction temperature by 0.3-0.5℃.

[0062] Specifically, the reaction process prediction model submodule uses at least 50 batches of complete process data for the low-temperature hydrothermal synthesis of barium titanate as the training set. The data dimensions include the initial molar ratio of raw materials (Ti:Ba), temperature, pressure / pH time series data of each reaction stage, stirring rate curve, and final particle size distribution / purity of the product. The data training data volume and data integrity standards are clearly defined. The specific functions are as follows: ① Quantify the amount of training data: At least 50 batches is the key threshold for the reaction process prediction model submodule to have statistical significance. The low-temperature hydrothermal synthesis process has slight fluctuations between batches. Data from more than 50 batches can cover the process characteristics of different raw material batches and equipment conditions, avoiding model overfitting (prediction results cannot be generalized) due to insufficient data volume, and ensuring the model's adaptability to actual production; ② Define complete process data: Emphasize that the data covers the entire reaction cycle (from feed to discharge), rather than data from a single stage, to ensure that the reaction process prediction model submodule can capture the complete correlation link of "raw material ratio → process parameters → product performance", solving the problem of "prediction bias caused by data fragmentation" in traditional models.

[0063] Furthermore, precise matching of preceding process parameters and data acquisition logic is achieved, specifically in the correlation of data dimensions: Initial molar ratio of raw materials (Ti:Ba): This is a core raw material parameter for barium titanate synthesis (directly affecting product purity; claim 8 includes it in the determination of valid reaction data), corresponding to the control target of the feed metering submodule (precisely controlling the feed rate to maintain the molar ratio); Time-series data of temperature, pressure, and pH at each reaction stage: Echoing the real-time acquisition parameters (temperature, pressure, pH) of the multi-dimensional acquisition module and the three major process stages, the time-series data can reflect the dynamic change trend of parameters, providing support for the reaction process prediction model submodule to capture the "stage-parameter-performance" correlation; Stirring rate curve: Corresponding to the acquisition parameters and the control object of the stirring drive submodule, the stirring rate affects the mixing uniformity of the reaction solution and is an important influencing factor on the crystal growth rate; Final particle size distribution / purity of the product: This is the core target of the reaction process prediction model submodule and also a key indicator for determining valid reaction data (purity ≥99.5%, particle size variation coefficient ≤5%), serving as a label dimension for training data to ensure that the reaction process prediction model submodule learns the mapping relationship between input parameters and output performance.

[0064] The reaction process prediction model submodule adopts a fusion algorithm of BP neural network and random forest. The output dimensions include: real-time predicted value of crystal growth rate, predicted value of crystal grain size distribution width, and predicted value of product purity deviation. Its functions are: ① The technical rationality of choosing the fusion algorithm: BP neural network is good at capturing nonlinear relationships (adapting to the feature of "multi-parameter coupling affecting product performance" in low-temperature hydrothermal synthesis), and random forest has strong anti-interference ability and feature importance recognition ability (it can filter abnormal data and focus on key influencing parameters). The fusion of the two breaks through the limitations of single algorithms (such as BP neural network is prone to overfitting and random forest is insufficient for capturing time series data), significantly improving prediction accuracy and reflecting technological innovation; ② Different from traditional models: The fusion of AI / machine learning algorithms is different from traditional mechanism modeling or single regression models, echoing the core direction of active predictive regulation. In addition, it also has the following functions: expanding and clarifying the output dimensions of the model and improving the prediction function. Specifically, ① the progressive relationship of the output dimensions: on the basis of crystal growth rate and product purity, a new crystal particle size distribution width prediction value is added. The particle size distribution width directly affects the application performance of barium titanate (such as the need for narrow particle size distribution in the field of electronic ceramics). This new dimension makes the reaction process prediction model submodule more comprehensive and covers the core performance indicators of the product; ② quantifying the prediction target attributes: real-time prediction value, prediction value, and deviation prediction value clarify the functional positioning of the output data. Real-time prediction is used to dynamically track the reaction process, and deviation prediction (the deviation between product purity and the preset target) provides a direct basis for pre-adjustment instructions (such as adjusting parameters to improve purity if the deviation is negative), so that the output of the reaction process prediction model submodule is directly linked to the control logic.

[0065] When the predicted crystal growth rate is 10% lower than the preset baseline value, the reaction process prediction model submodule outputs a pre-adjustment command: increase the stirring rate by 5-8 r / min and fine-tune the reaction temperature by 0.3-0.5℃. Its function is to visualize the triggering conditions and execution parameters of the pre-adjustment command. Specifically, it quantifies the trigger threshold: 10% lower than the preset baseline value. The preset baseline value is a crystal growth rate standard determined based on the optimal process (such as a reasonable growth rate to ensure uniform product particle size). The 10% threshold avoids frequent adjustments caused by small fluctuations while allowing timely intervention for significant deviations, thus balancing the control effect. ① Efficiency and stability; ② Quantitative adjustment of parameters: increase stirring rate by 5-8 r / min and fine-tune temperature by 0.3-0.5℃. The adjustment range has been verified by the process (small temperature increase avoids abnormal crystal form, moderate increase of stirring rate promotes mixing and mass transfer of reaction solution and accelerates crystal growth). It ensures the adjustment effect and avoids over-adjustment that will cause new parameter fluctuations, reflecting the core idea of ​​precise pre-control; ③ Corresponding to the execution module sub-module: the pre-adjustment command directly points to the stirring drive sub-module (adjusting the stirring rate) and the temperature adjustment sub-module (adjusting the reaction temperature), realizing the closed loop of "model prediction → command output → sub-module execution".

[0066] For example, the reaction process prediction model submodule uses complete process data of at least 50 batches of low-temperature hydrothermal synthesis of barium titanate as a training set, and the single batch data of the training set is used to construct a feature vector X. k The data dimensions include the initial molar ratio of raw materials, M. Ti / Ba Temperature / pressure / pH time-series data matrix for each reaction stage T temp / pH (t), stirring rate curve n(t), final particle size distribution of product D k and purity P k Where k is the batch number (k=1, 2, ..., N, N≥50); The reaction process prediction model submodule employs a weighted fusion algorithm of BP neural network and random forest. First, the prediction results are output separately by the two sub-models. Then, weights are assigned based on the confidence scores from cross-validation of the training set. The core fusion formula is as follows: The BP neural network prediction formula is: Y BP (t)=f BP (W3·σ(W2·σ(W1·X(t)+b1)+b2)+b3); The random forest prediction formula is: ; The weighted fusion output formula is: Y pred (t)=α·Y BP (t)+(1-α)·Y RF (t); Where X(t) represents the real-time input feature vector at time t (dimension equal to or greater than the training set X). k Consistent (including current raw material molar ratio, collected temperature / pressure / pH time series segments, and real-time stirring rate); W1, W2, and W3 represent the input-hidden layer, hidden layer-hidden layer, and hidden layer-output layer weight matrices of the BP neural network, respectively; b1, b2, and b3 represent the bias vectors of each layer of the BP neural network; σ(·) represents the activation function, using the ReLU function, σ(x) = max(0, x), used to capture the nonlinear correlation of process parameters; f BP (·) represents the mapping function of the BP neural network, outputting a prediction vector with dimension 3; M represents the number of decision trees in the random forest (determined during training, typically M≥100); f RF，m (·) represents the prediction function of the m-th decision tree; Y RF (t) represents the ensemble prediction vector of the random forest (the mean of the predictions from all decision trees is taken to improve stability); α represents the fusion weight (0 < α < 1, determined by 5-fold cross-validation; if the prediction error of the BP neural network is smaller, α approaches 1, otherwise it approaches 0); Y pred (t) represents the final prediction vector of the fusion model, containing three output dimensions: v pred (t) represents the real-time predicted value of crystal growth rate (unit: nm / min); D pred (t) represents the estimated width of the crystal grain size distribution (unit: nm, expressed as the standard deviation of the grain size distribution); ΔP pred (t) represents the estimated deviation of product purity, ΔP pred (t)=P ref -P pred (t), where P ref To preset the target purity value, P pred (t) represents the predicted purity of the product at time t; When the predicted crystal growth rate v pred (t) < 0.9·v ref When, that is, below the preset baseline value v ref When the reaction progress prediction model submodule reaches 10%, it outputs a pre-adjustment command: controls the stirring drive submodule to increase the stirring rate from the current value n(t) to n′(t)=n(t)+△n, △n=5~8r / min, and controls the temperature regulation submodule to fine-tune the reaction temperature from the current value T(t) to T′(t)=T(t)+△T, △T=0.3~0.5℃, thereby accelerating crystal growth by improving mass transfer efficiency and ensuring that the product performance meets the standards.

[0067] In some preferred embodiments, the reaction process prediction model submodule has a built-in iterative update unit. After accumulating 20 batches of valid reaction data, the iterative update unit automatically selects process optimization cases from these 20 batches of valid reaction data and retrains the model weights to keep the prediction accuracy stable within ±2%. The valid reaction data refers to the reaction data generated during the low-temperature hydrothermal synthesis of barium titanate, and must simultaneously meet the following requirements: process parameters meet preset thresholds, with deviation time ≤2 seconds and amplitude ≤3%; the full-process parameter acquisition frequency is ≥10Hz, single batch data loss is ≤5 seconds and includes complete control records, status records and alarm records; product purity is ≥99.5%, particle size variation coefficient is ≤5% and the crystal form is tetragonal; the acquisition sensor has been calibrated within 30 days and is within its validity period, and the measurement error meets the preset accuracy.

[0068] Specifically, the reaction process prediction model submodule has a built-in iterative update unit. After accumulating 20 batches of valid reaction data, the iterative update unit automatically selects process optimization cases from these 20 batches of valid reaction data and retrains the model weights. Its function is to quantify the iteration triggering conditions and data selection logic. Specifically, it includes: ① Iteration batch threshold: Every 20 batches of valid reaction data are accumulated. 20 batches is a reasonable threshold for balancing iteration efficiency and data representativeness. Too few batches will result in insufficient new data to cover process fluctuations, while too many batches will lead to an excessively long model accuracy decay period. This threshold allows the model to adapt to subtle changes such as raw material batches and equipment status in a timely manner; ② Data selection criteria: Automatically selects process optimization cases and only selects optimized cases (i.e., effective data with better product performance and more precise parameter control) for iteration. This avoids diluting model performance with ordinary effective data or potentially abnormal data, ensuring that the prediction accuracy of the model only increases after iteration, rather than simply accumulating data for retraining.

[0069] Maintaining prediction accuracy within ±2% is the core objective of quantitative model iteration. This involves clearly defining the prediction error boundaries for indicators such as crystal growth rate and product purity after iteration (e.g., when the predicted purity is 99.5%, the actual purity is between 99.3% and 99.7%). This accuracy fully meets the industrial production requirements for low-temperature hydrothermal synthesis of barium titanate (product purity requirement ≥99.5%, as specified in claim 8). This avoids the failure of pre-adjustment instructions due to insufficient prediction accuracy. Simultaneously, it emphasizes the long-term effectiveness of the iteration function, ensuring that the model maintains high accuracy during long-term production (after multiple iterations), rather than only achieving the target after initial training.

[0070] The effective reaction data refers to the reaction data generated during the low-temperature hydrothermal synthesis of barium titanate, and must simultaneously meet the following requirements: process parameters meet preset thresholds, with deviation duration ≤2 seconds and amplitude ≤3%; the full-process parameter acquisition frequency is ≥10Hz, with single-batch data loss ≤5 seconds and containing complete control records, status records, and alarm records; product purity is ≥99.5%, particle size variation coefficient is ≤5%, and the crystal form is tetragonal; the acquisition sensor has been calibrated within 30 days and is within its validity period, and the measurement error meets the preset accuracy. Its purpose is to clearly define "effective reaction data," fill the gaps in the previous data standards, and define effective reaction data through the "four-in-one conditions" (process compliance, data integrity, product compliance, and data reliability), ensuring that the data used for model training / iteration is "process compliant, data complete, product high-quality, and measurement reliable," thus guaranteeing model accuracy from the source and avoiding model deviation caused by low-quality data.

[0071] Other functions include: ① Process compliance: Process parameters meet preset thresholds, with deviation duration ≤2 seconds and amplitude ≤3%, echoing preset process thresholds, quantifying parameter deviations from standards (avoiding severely abnormal data), and ensuring data reflects normal process status; ② Data integrity: Full-process parameter acquisition frequency ≥10Hz, single batch data missing ≤5 seconds and containing complete records, echoing real-time acquisition by multi-dimensional acquisition modules, acquisition frequency (≥10Hz) consistent with previous sequences, data missing threshold (≤5 seconds) and complete record requirements ensure data coverage of the entire reaction cycle, with no critical nodes omitted; ③ Product compliance: Product purity ≥99.5%, particle size variation... The coefficient is ≤5% and the crystal form is tetragonal, which corresponds to the final particle size distribution / purity of the product (model output index). At the same time, the core quality standard of barium titanate is clearly defined (tetragonal phase is the key crystal form of barium titanate as an electronic ceramic material), ensuring that the effective reaction data corresponds to the process characteristics of high-quality products, so that the model can learn the correlation between the parameters and performance of high-quality products; ④ Data reliability: The acquisition sensor has been calibrated within 30 days and is within the validity period. The measurement error meets the preset accuracy, which corresponds to the acquisition parameters of the multi-dimensional acquisition module. The sensor calibration requirements ensure the authenticity of the acquired data (avoid false data caused by sensor drift), providing reliable input for model training. The definition of effective reaction data is the key link connecting "data acquisition → model training / iteration → product quality": only data that meets this definition can train a high-precision model, and the pre-adjustment instructions output by the model can ensure that the product meets the standards. The data of the product meeting the standards will then serve as new effective data to feed back into the model iteration, forming a positive cycle of "data-model-product-data".

[0072] For example, the reaction process prediction model submodule has a built-in iterative update unit, which has an online incremental iterative update function. The iterative update unit accumulates 20 batches of valid reaction data (denoted as incremental dataset X). new Y trueAutomatically filter out process optimization cases (that meet △P) true ≤0.5% and D true ≤D ref That is, the actual purity deviation is ≤0.5% and the actual particle size distribution width is ≤the preset target value D. ref The model weights are updated using the following formula to keep the prediction accuracy stable within ±2%. The formula for weight update in a BP neural network (incremental gradient descent) is: ; ; Where W represents the BP neural network weight matrix before the update (including the weights of input layer-hidden layer, hidden layer-hidden layer, and hidden layer-output layer); W′ represents the BP neural network weight matrix after the iterative update; b represents the BP neural network bias vector before the update (corresponding to the biases of each layer); b′ represents the BP neural network bias vector after the iterative update; η represents the learning rate, with a value ranging from η=0.001 to 0.01, its function being to avoid excessive update amplitude leading to model oscillation; L(·) represents the loss function, using mean squared error (MSE), the specific expression of which is... , where Y pred，i Y represents the model-predicted product performance vector corresponding to the i-th incremental batch (including the predicted crystal growth rate, predicted grain size distribution width, and predicted product purity deviation for that batch); true，i This represents the actual product performance vector corresponding to the i-th incremental batch (including the actual crystal growth rate v of that batch). true Actual particle size distribution width D true Actual purity deviation △P true ); This represents the partial derivative of the loss function with respect to the weight matrix W (used to calculate the direction of weight updates). λ represents the partial derivative of the loss function with respect to the bias vector b (used to calculate the direction of bias update); λ represents the regularization coefficient, ranging from 10⁻⁵ to 10⁻⁴, which is used to prevent overfitting of the model; X new Y represents the input features of the incremental dataset (corresponding to process parameter data for the low-temperature hydrothermal synthesis of barium titanate); true This represents the actual product performance vector of the incremental dataset, including the actual crystal growth rate v. true Actual particle size distribution width D true Actual purity deviation △P true N new This indicates the number of incremental batches, with a value of 20.

[0073] The formula for updating weights in a random forest (using a resampling ensemble method) is: Xupdate =β·X opt +(1-β)·X old ; Where X update X represents the training dataset after iterative updates by the random forest; β represents the update weights of the random forest, with a value ranging from β=0.3 to 0.5, which balances the influence of new cases and historical data; X opt The selected process optimization case dataset (i.e., effective reaction data that yields better product performance); X old This is the original training set (complete process data of at least 50 batches of low-temperature hydrothermal synthesis of barium titanate originally used to train the model).

[0074] Valid reaction data refers to reaction data generated during the low-temperature hydrothermal synthesis of barium titanate, and must simultaneously meet the following quantification conditions: 1. Compliance of process parameters: maximum process parameter t∈[0，Ttotal] |y(t)-r(t)|≤0.03·r(t) and deviates from the duration ≤2s; Where y(t) is the real-time acquired value, r(t) is the preset process threshold, and T total This refers to the total reaction time for a single batch; 2. Data integrity: Acquisition frequency f s ≥10Hz, duration of missing data in a single batch ≤5s, and includes complete control command records, module status records, and alarm records (if any); 3. Product compliance: Actual product purity P true ≥99.5%, particle size variation coefficient ≤5% (the actual average grain size), and the crystal form is tetragonal; 4. Data Reliability: The calibration validity period t of the acquired sensors cal ≤30d, measurement error , (y std (t) is the standard value, δ ref (This is a preset accuracy threshold, such as temperature ±0.1℃, pH ±0.05).

[0075] In some preferred embodiments, the safety early warning module has a built-in abnormal condition classification and handling submodule. This submodule classifies abnormalities where parameters exceed preset process thresholds into three levels and matches them with differentiated handling strategies: Level 1 abnormality: parameters deviating from the threshold by less than 5% triggers each submodule of the control module to fine-tune the corresponding parameters and records the abnormal node; Level 2 abnormality: parameters deviating from the threshold by 5%-10% control the injection of 0.1 mol / L potassium hydroxide emergency buffer solution into the reactor, with the injection volume being 0.5%-1.0% of the reaction liquid volume; the stirring drive submodule controls the reduction of the stirring rate to 80% of the original rate; Level 3 abnormality: parameters deviating from the threshold by more than 10% trigger an audible and visual alarm and drive the execution module to perform an emergency shutdown, while simultaneously driving the temperature regulation submodule to control the rapid circulation of the cooling medium in the reactor jacket, reducing the reaction temperature to room temperature within 10 seconds; the abnormal condition classification and handling submodule is linked with the reaction process prediction model submodule, triggering the corresponding level of pre-treatment strategy 1-2 minutes in advance based on the predicted parameter change trend.

[0076] Specifically, the safety early warning module has a built-in abnormal condition classification and handling sub-module. The abnormal condition classification and handling sub-module classifies abnormalities where parameters exceed preset process thresholds into three levels and matches them with differentiated handling strategies: by adding classification and handling capabilities, the safety early warning is upgraded from a one-size-fits-all emergency shutdown to a gradient and precise handling, avoiding production interruptions due to minor abnormalities or safety accidents caused by untimely handling of serious abnormalities.

[0077] A parameter deviation of less than 5% from the threshold is considered a Level 1 anomaly: This triggers the fine-tuning of corresponding parameters in each sub-module of the control module, while simultaneously recording the anomaly node. Its purpose is to define the handling logic for "minor anomalies," balancing production and control. Specifically: ① Quantifying the Level 1 anomaly threshold: Deviations of less than 5% from the threshold are considered minor fluctuations (e.g., pH fluctuating from 11.0 to 11.4, a deviation of 4.5%), requiring no emergency shutdown and avoiding over-handling that could lead to production interruption; ② Clarifying the executing entity and action: Triggering the fine-tuning of each sub-module of the control module, corresponding to the improved PID linkage control algorithm sub-module, precisely corrects parameters through the PID's stable adjustment mode (switching when deviation < 5%), reflecting the principle of minimal intervention for minor anomalies; ③ Supplementing data traceability function: Recording anomaly nodes echoes subsequent data traceability and process replication sub-modules. Anomaly node data can serve as the basis for subsequent process optimization and model iteration, forming a closed loop of "anomaly handling → data recording → optimization feedback." The handling of level one anomalies relies on the PID fine-tuning capability of the control module, which works in conjunction with passive response control. Minor anomalies can be corrected through routine control without the need to initiate emergency measures or shut down the machine, thus ensuring production continuity.

[0078] A parameter deviation of 5%-10% from the threshold is considered a secondary anomaly: The emergency response submodule controls the injection of 0.1 mol / L potassium hydroxide emergency buffer solution into the reactor, with the injection volume being 0.5%-1.0% of the reaction liquid volume; the stirring drive submodule controls the reduction of the stirring rate to 80% of the original rate. Its role is to define a collaborative handling strategy for moderate anomalies and suppress the expansion of the anomaly. Specifically, this includes: ① quantifying the secondary anomaly threshold: 5%-10% is considered a moderate deviation (e.g., a temperature fluctuation from 180℃ to 200℃ exceeding the threshold by 11.1%, or a deviation of 8.3% from 195℃), requiring proactive intervention to prevent deterioration; ② clarifying the dual execution entities and quantified actions: The emergency response submodule injects 0.1 mol / L potassium hydroxide emergency buffer solution: corresponding to the emergency response... The buffer solution is selected as 0.1 mol / L potassium hydroxide (high concentration, strong buffering capacity, different from barium hydroxide buffer, priority for rapid error correction in emergency scenarios, no need to consider the same component as the barium source, avoid introducing impurities without affecting core performance), and the injection volume is 0.5%-1.0% to achieve quantification and ensure controllable buffering effect; the stirring drive submodule reduces the stirring rate to 80%: corresponding to the stirring drive submodule, the speed reduction can slow down the mixing and reaction rate of the reaction solution, buy time for parameter correction, and avoid the rapid expansion of abnormalities (such as when the pH drops sharply, the speed reduction can reduce the diffusion of acidic areas); ③ It reflects the synergistic treatment logic: through the combination of buffer neutralization and speed reduction to stabilize the reaction, moderate abnormalities are specifically suppressed, so as not to interrupt production and to quickly control risks.

[0079] A parameter deviation of more than 10% from the threshold constitutes a Level 3 anomaly: The safety warning module triggers an audible and visual alarm and drives the execution module to perform an emergency shutdown. Simultaneously, it drives the temperature regulation submodule to control the rapid circulation of the cooling medium in the reactor jacket, reducing the reaction temperature to room temperature within 10 seconds. Its function is to define the safety handling logic for severe anomalies and ensure ultimate safety, specifically including: ① Quantifying the Level 3 anomaly threshold: A deviation of more than 10% is considered a severe deviation (e.g., pressure fluctuating from 5MPa to 6MPa, a deviation of 20%), which may lead to reaction runaway (e.g., crystal form destruction, equipment overpressure), requiring priority to ensure safety; ② Replicating and strengthening the preceding safety logic: The full early warning module triggers audible and visual alarms and drives the execution module to shut down urgently, fully echoing the core safety early warning function and maintaining logical consistency; ③ A new rapid cooling safety action is added: the temperature regulation submodule is driven to start rapid circulation of the cooling medium. The corresponding temperature regulation submodule cools to room temperature within 10 seconds, which is a quantified safety standard (low-temperature hydrothermal synthesis relies on high temperature and pressure; rapid cooling can terminate the reaction and reduce equipment pressure), solving the safety hazard of only shutting down without cooling in the preceding process; ④ The execution subject level is clearly defined: the safety early warning module acts as the triggering subject, driving the temperature regulation submodule of the execution module to perform cooling, conforming to the linkage logic of early warning before execution. Adapting to process safety requirements, the high-pressure and high-temperature environment of low-temperature hydrothermal synthesis dictates that severe anomalies require both shutdown and cooling as a dual safeguard. The quantified 10-second cooling target ensures that the safety standard is implemented, avoiding secondary risks caused by incomplete handling.

[0080] The abnormal condition classification and handling submodule is linked with the reaction process prediction model submodule. Based on the predicted parameter change trend, it triggers the corresponding level of pre-treatment strategy 1-2 minutes in advance. Its function is to realize the safety upgrade from passive response to active prediction, breaking through the limitations of traditional early warning. Specifically, it establishes cross-module linkage logic: linked with the reaction process prediction model submodule, it uses the reaction process prediction model submodule's ability to predict parameter change trends (such as predicting that pH will deviate from the threshold by 5% after 2 minutes) to trigger pre-treatment in advance, solving the pain point of the lag in response after the occurrence of traditional early warning; ② Quantify the advance amount of pre-treatment: 1-2 minutes in advance is a reasonable range to balance the accuracy of prediction and the timeliness of treatment. Too short a time will not be able to complete the pre-treatment, and too long a time may lead to misoperation due to prediction deviation; ③ Consistency of pre-treatment strategy: the pre-treatment strategy of the corresponding level refers to the early triggering of the first level (fine adjustment) and the second level (buffer solution and deceleration) treatment actions to prevent the abnormality from developing into a more serious level, reflecting the safety design of prevention first. Strengthening the logical closed loop of the entire system, this linkage logic connects the prediction model of the safety early warning module and the control module, enabling the system to form a full-chain safety protection of "predictive model to predict trends → graded disposal sub-module to pre-dispose → precise disposal after an anomaly occurs", which is completely consistent with the preceding "prediction-control" logic and has no logical gaps.

[0081] In some preferred embodiments, the execution module further includes a data traceability and process replication submodule. This submodule communicates bidirectionally with the control module to extract complete process parameter sequences corresponding to historical valid reaction data and generate a standardized process parameter package. When a new reaction batch is started, the process parameter package is directly invoked, and the control module automatically synchronizes preset process thresholds to each submodule, achieving replication of the barium titanate synthesis process under the same process conditions. The data traceability and process replication submodule supports binding and labeling process parameter packages with the performance indicators of the corresponding batch products, enabling real-time rapid matching of processes to meet performance requirements. These performance indicators include barium titanate purity and particle size.

[0082] Specifically, the execution module also includes a data traceability and process replication submodule. This submodule communicates bidirectionally with the control module to extract the complete process parameter sequence corresponding to historical valid reaction data and generate a standardized process parameter package. Its functions are: ① Communication method: Bidirectional communication with the control module. On the one hand, the data traceability and process replication submodule extracts historical valid reaction data from the control module (ensuring data quality). On the other hand, it can feed back the execution data after process replication to the control module, providing new data support for the online iteration of the reaction process prediction model, forming a positive cycle of data reuse and feedback optimization; ② Data extraction and transformation: First, the complete process parameter sequence is extracted, and then the standardized process parameter package is generated. The complete process parameter sequence refers to the full-cycle parameters such as temperature, pressure / pH time series data, stirring rate curve, and raw material molar ratio at each reaction stage, avoiding data fragmentation. The standardized process parameter package is a structured integration of scattered parameters (such as dividing parameter intervals according to reaction stages and marking threshold ranges), solving the pain point of relying on experience to record process parameters in traditional production and lacking unified standards, providing a carrier that can be directly called for subsequent replication.

[0083] When a new reaction batch is started, the process parameter package is directly invoked. The control module automatically synchronizes the preset process thresholds to each submodule, achieving replication of the barium titanate synthesis process under the same process conditions. Its function is to visualize the execution logic of the process replication and connect with the functions of the preceding modules. Specifically, ① the replication triggering and execution process is as follows: Start a new batch → Invoke the parameter package → Control module synchronizes thresholds → Each submodule executes, perfectly matching the control and execution logic of the preceding system. After receiving the parameter package, the control module synchronizes the preset process thresholds to submodules such as temperature regulation, pH regulation, and stirring drive, ensuring that each submodule works collaboratively according to a unified standard, achieving precise replication under the same process conditions. ① **Significant Reproducibility; ② **Solving Core Production Pain Points:** Traditional low-temperature hydrothermal synthesis relies on operator experience to replicate the process, which easily leads to batch-to-batch parameter deviations (such as inconsistent raw material molar ratios and reaction stage temperature thresholds), resulting in fluctuations in product purity and particle size. This process replication achieves batch-to-batch process consistency through standardized parameter packages, directly serving the quality targets of product purity ≥99.5% and particle size variation coefficient ≤5%; ③ **Quantitative Replication Precision Implicit Logic:** The process of synchronizing the parameter package to the sub-module relies on a pre-process improved PID algorithm and pH premixing compensation logic to ensure that the parameter control precision after replication is consistent with the original process, avoiding the problem of consistent parameter packages but execution deviations. Furthermore, it enhances the system's industrial applicability. The process replication function reduces the debugging cost of new batch production (eliminating the need for repeated parameter optimization), improves production efficiency, and is particularly suitable for the "multi-batch, high consistency" requirements of large-scale industrial production, upgrading the technical solution from laboratory-level precise control to industrial-grade reusable solutions.

[0084] The data traceability and process replication submodule supports binding and labeling process parameter packages with the performance indicators of corresponding batches of products, and supports rapid process matching based on corresponding performance requirements. The performance indicators include barium titanate purity and particle size. Its function is to establish a direct correlation between process parameters and product performance, and improve production flexibility. Specifically, the core value of binding and labeling is to bind standardized process parameter packages with the performance indicators of corresponding batches of products (e.g., parameter package A corresponds to a purity of 99.8% and a particle size of 1μm; parameter package B corresponds to a purity of 99.6% and a particle size of 2μm). This solves the problem of no clear correspondence between process and performance in traditional production. Operators can quickly match the corresponding process parameter packages according to downstream application requirements (e.g., electronic ceramics require narrow particle size distribution, and capacitors require high purity) without having to re-explore the process.

[0085] In addition, it has the value of supporting the entire chain of data traceability. The binding and labeling not only serves the purpose of rapid matching, but also provides a basis for process optimization: if the performance of a certain batch of products does not meet the standards, the corresponding process parameter sequence can be traced through the data traceability and process replication sub-module to locate the deviation node (such as excessive temperature fluctuation in a certain reaction stage), and the optimized parameters are updated to a new standardized parameter package to feed back into subsequent production and model iteration, forming a closed loop of "traceability-optimization-reuse".

[0086] For example, the specific working logic of the data traceability and process replication submodule is as follows: Step 1: Achieve data traceability, which involves transforming valid data into standardized parameter packages. The core is to convert fragmented historical valid response data into traceable and reusable structured parameter packages. Specific steps include: 1. Data extraction and filtering.

[0087] The data traceability and process replication submodule, through the control module, batch extracts complete process parameter sequences corresponding to historical valid reaction data (such as temperature / pressure / pH time series data for each reaction stage, stirring rate curves, and initial molar ratio of raw materials (Ti:Ba)). The filtering dimensions include: Time dimension: Covers the entire reaction cycle (process stages) from "precursor dissolution → crystal nucleus formation → crystal growth"; Parameter dimensions: including raw material feed rate curve, pH premixed buffer injection volume time series data, PID parameter dynamic adjustment record, and emergency response record (if any); Performance dimensions: purity, particle size (average particle size and particle size distribution width), and crystal form detection results for the corresponding batch of products.

[0088] 2. Data structuring and standardized encapsulation.

[0089] The data traceability and process replication submodule will generate standardized process parameter packages from the filtered complete parameter sequence according to the structure of "reaction stage, parameter type, threshold range, and execution logic". Each parameter package contains three core categories: The packaged contents include: Basic process parameters, such as the initial molar ratio of raw materials (Ti:Ba=1:1.05) and the preset temperature and pressure thresholds for each stage (e.g., temperature 160±2℃ and pressure 4MPa for the precursor dissolution stage). Dynamic control parameters, such as PID initial parameters (Kp=2.5, Ti=60s, Td=15s) and fuzzy rule thresholds for each stage; Configure the module parameters, such as the pH premixed buffer molar ratio (1:1) and the stirring rate curve (300 r / min for the precursor dissolution stage and 200 r / min for the crystal growth stage). Product performance labels, such as purity 99.8%, average particle size 1.2 μm, particle size variation coefficient 3.2%, and tetragonal crystal form.

[0090] 3. Data traceability function implemented.

[0091] The data traceability and process replication submodule has a built-in data storage unit (or is associated with an industrial database), assigns a unique ID to each standardized process parameter package, and supports two types of traceability operations: Trace by "Parameter Package ID": Enter the ID to query the full process parameter time-series curve, control record, and alarm log (if any) for the corresponding batch. Traceability by "Performance Indicators": Input the target purity (e.g., ≥99.7%) or particle size (e.g., 1-1.5μm), and the submodule will automatically match the corresponding parameter package ID and historical data to achieve reverse tracing from "performance to process".

[0092] Step 2: Achieve process replication, which involves accurately reproducing the process from calling the parameter package. The core is to ensure that the process conditions of the new reaction batch are completely consistent with those of historical high-quality batches by calling standardized process parameter packages. Specific steps: 1. Parameter package calls are synchronized with thresholds.

[0093] When starting a new reaction batch, the operator can directly select the target standardized process parameter package (or match and call it according to product performance requirements) through the interactive interface (such as a touch screen or host computer software) of the data traceability and process replication submodule. The data traceability and process replication submodule sends the parameter package to the control module through bidirectional communication, and the control module automatically performs "threshold synchronization". Synchronization objects: all execution submodules (i.e., temperature regulation, pressure regulation, pH regulation, stirring drive, feed metering, and emergency response submodules); Synchronized content: preset process thresholds for each submodule (such as the target temperature range of the temperature control submodule and the premixed buffer ratio of the pH control submodule), initial values ​​of PID parameters, and preset baseline values ​​of the reaction process prediction model (such as the baseline value of crystal growth rate of 5 nm / min). Synchronization verification: After the control module completes synchronization, it sends the "threshold matching results of each submodule" back to the data traceability and process replication submodule to ensure that there is no synchronization deviation (if the deviation is greater than 0.5%, a prompt will be triggered).

[0094] 2. Full-process process replication and execution.

[0095] After the new reaction batch starts, each execution submodule strictly follows the synchronized parameter package to perform operations. The core reproduction logic is as follows: During the feeding stage: the feeding metering submodule precisely controls the Ti / Ba raw material feeding according to the "raw material feeding rate curve" in the parameter package, and the pH adjustment submodule simultaneously controls the injection of barium hydroxide buffer into the premixed buffer chamber according to the "1:1 molar ratio". Reaction phase: The control module dynamically adjusts the control commands according to the PID initial parameters and fuzzy rules in the parameter package, and the reaction process prediction model outputs pre-adjustment commands according to the reference values ​​in the parameter package; Anomaly Handling: The safety warning module performs tiered handling according to the anomaly thresholds in the parameter package to ensure that the safety control during the reproduction process is consistent with historical batches.

[0096] 3. Verification and feedback of replication results.

[0097] After the new batch reaction is completed, the data traceability and process replication submodule automatically collects product performance indicators (purity, particle size) and compares them with the performance labels in the parameter package: If the deviation is ≤2% (echoing the prediction accuracy of claim 8), the replication is considered successful, and the new batch of data is included in the "effective response data" for subsequent model iterations; If the deviation is greater than 2%, the deviation point (such as excessive temperature fluctuation at a certain stage) will be automatically recorded to provide a basis for process optimization.

[0098] Step 3: Performance binding and fast matching implementation.

[0099] The data traceability and process replication submodule supports binding and labeling standardized process parameter packages with the performance indicators (purity, particle size) of corresponding batch products. Specific implementation details are as follows: 1. Bind annotation logic.

[0100] The product performance label for each standardized process parameter package is permanently bound to the parameter package itself (associated with the same ID during storage), and the label content includes: Key performance indicators: barium titanate purity (e.g., 99.6%), average particle size (e.g., 0.8 μm); Derivative performance indicators: particle size variation coefficient (e.g., 4.5%), crystal form (tetragonal phase); Application scenario tags (optional, to improve practicality): such as "for electronic ceramics (high purity, narrow particle size)" and "for capacitors (medium particle size, high density)".

[0101] 2. The quick matching function has been implemented.

[0102] Operators can input target performance requirements (such as "purity ≥ 99.5%, particle size 1-2μm") through the interactive interface of the data traceability and process replication submodule. The submodule will then quickly match these requirements using the following logic: Screening by performance indicators: From all standardized process parameter packages, select candidate parameter packages whose performance labels meet the target requirements; Sort by matching degree: Calculate the deviation between the performance label of the candidate parameter package and the target requirement (e.g., if the target purity is 99.5% and the candidate package purity is 99.8%, the deviation is 0.3%), and sort them by deviation from smallest to largest; Recommend the optimal parameter package: Recommend the parameter package with the highest matching degree to the operator, and support direct call to start a new batch, so as to realize the rapid implementation of "requirement → matching → replication" without repeated process debugging.

[0103] In summary, the functions of the data traceability and process replication submodule are as follows: Traceability: Enables forward traceability of "process parameters → product performance" and reverse traceability of "performance requirements → process parameters", solving the pain points of traditional production such as "no process records and difficulty in locating problems"; At the replication level: By standardizing parameter packages and synchronizing thresholds across modules, we ensure consistent process conditions across multiple batches, thereby improving product consistency. In terms of matching: performance binding annotations make process selection more efficient, adapt to the performance requirements of different downstream applications, and reduce the debugging costs of large-scale production.

[0104] The control system for the low-temperature hydrothermal synthesis of barium titanate provided in this application, based on the above technical solutions, has the following advantages: 1) Achieve precise dynamic control in multiple stages and improve parameter stability and process adaptability: Through the built-in fuzzy adaptive adjustment mechanism of the improved PID linkage control algorithm submodule, differentiated fuzzy rules are configured for the three stages of precursor dissolution, crystal nucleus formation and crystal growth, and the PID parameters (proportional coefficient, integral time and derivative time) are dynamically adjusted. By switching between "fast adjustment mode (deviation > 15%)" and "stable adjustment mode (deviation < 5%)", precise control of "rapid correction of large deviations and stable maintenance of small deviations" is achieved.

[0105] 2) Stabilizing pH from the source to ensure the quality of crystal nuclei and crystal growth: Through the "feed-premix-real-time pH compensation" linkage logic of the pH adjustment submodule and the feed metering submodule, the acidity of the titanium source is neutralized by premixing with barium hydroxide buffer solution of the same composition as the barium source at a molar ratio of 1:0.9-1.1 before the raw materials are injected into the reactor. Then, real-time monitoring and replenishment ensure that the pH is stable in the range of 11±0.1. This design avoids the risk of sudden pH drop from the source, avoids the lag in adjustment in the reactor and the problem of local over-alkalization, improves the consistency of crystal nuclei formation, reduces the particle size variation coefficient, and ensures a high proportion of tetragonal phase crystals.

[0106] 3) Proactively predicting reaction progress and intervening in advance to avoid performance deviations: The built-in reaction progress prediction model submodule in the control module, based on a BP neural network-random forest fusion algorithm trained with at least 50 batches of complete process data, can predict crystal growth rate, particle size distribution width, and product purity deviations in real time. Combined with an online iterative update function, the model weights are optimized every 20 batches of valid data, ensuring prediction accuracy remains stable within ±2%. Through a proactive intervention mode of "trend prediction - 1-2 minute advance adjustment" (e.g., fine-tuning the temperature by 0.3-0.5℃ and increasing the stirring rate by 5-8 r / min when the predicted growth rate is too low), the lag problem of passive correction in traditional systems is solved, thus improving product purity.

[0107] 4) Graded handling of abnormal operating conditions to balance production safety and continuity: The abnormal operating condition graded handling submodule of the safety early warning module classifies parameter anomalies into three levels and matches them with differentiated strategies: Level 1 anomalies (deviation <5%) involve fine-tuning and correction; Level 2 anomalies (5%-10%) involve buffer injection and slowing down the reaction to stabilize it; and Level 3 anomalies (>10%) involve shutdown and 10-second rapid cooling. This design avoids the production waste caused by the traditional "one-size-fits-all" shutdown, and at the same time, it achieves pre-treatment by linking with the predictive model, reducing the risk of anomaly escalation and balancing production continuity with extreme safety assurance.

[0108] 5) Enables traceable and reusable processes, improving batch consistency and industrial efficiency: Through the data traceability and process replication submodule, historical effective reaction data is encapsulated into standardized process parameter packages and bound to product performance indicators (purity, particle size). New batches can directly call the parameter packages to achieve precise process replication without repeated debugging. This function not only achieves bidirectional traceability of "performance-process" but also shortens the debugging cycle of new batches, reduces industrial production costs, and minimizes product performance deviations between batches, meeting the consistency requirements of large-scale mass production.

[0109] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A control system for the low-temperature hydrothermal synthesis of barium titanate, characterized in that, It includes a multi-dimensional data acquisition module, a control module, an execution module, and a security early warning module; The multidimensional acquisition module collects parameters such as temperature, pressure, stirring rate, pH value of reaction solution, and raw material feed rate in the reactor in real time. After receiving the collected parameters, the control module compares them with the preset process threshold and outputs control commands to the execution module. The control commands include temperature control commands, pressure control commands, stirring rate control commands, pH value control commands, feed rate control commands, and emergency response commands. The execution module receives control instructions and implements the corresponding parameter control. The safety early warning module monitors the sealing status of the reactor and the extreme values ​​of the system pressure. When the parameters exceed the preset process threshold, it triggers an audible and visual alarm and drives the execution module to perform an emergency shutdown.

2. The control system for the low-temperature hydrothermal synthesis of barium titanate according to claim 1, characterized in that, The execution module includes a temperature regulation submodule, a pressure regulation submodule, a stirring drive submodule, a pH regulation submodule, a feed metering submodule, and an emergency response submodule; The temperature regulation submodule receives temperature regulation commands and implements temperature regulation. The pressure regulation submodule receives pressure regulation commands and implements pressure regulation. The stirring drive submodule receives stirring rate control commands and performs stirring rate adjustment. The pH adjustment submodule receives pH value adjustment commands and performs pH value adjustment; The feed metering submodule receives the feed rate control command and implements the feed rate adjustment; The emergency response submodule receives emergency response instructions and performs emergency buffer injection.

3. The control system for the low-temperature hydrothermal synthesis of barium titanate according to claim 2, characterized in that, The control module includes an improved PID linkage control algorithm submodule, which dynamically adjusts the proportional coefficient, integral time, and derivative time of the PID parameters according to the reaction process. When the parameter deviation is greater than 15% of the preset process threshold, the rapid adjustment mode is activated. The rapid adjustment mode includes increasing the proportional coefficient, shortening the integral time, and shortening the derivative time. When the parameter deviation is less than 5% of the preset process threshold, the system switches to a stable adjustment mode, which includes maintaining the current proportional coefficient, integral time, and derivative time.

4. The control system for the low-temperature hydrothermal synthesis of barium titanate according to claim 3, characterized in that, The improved PID linkage control algorithm submodule has a built-in fuzzy adaptive adjustment mechanism unit. The working logic of the fuzzy adaptive adjustment mechanism unit is as follows: a fuzzy rule library is pre-constructed, containing 5 fuzzy subsets of parameter deviation inputs (represented as minimal, small, medium, large, and maximal) and 4 fuzzy subsets of deviation change rates (represented as slow, stable, relatively fast, and abrupt). Differentiated fuzzy rules are configured for the precursor dissolution stage, crystal nucleation stage, and crystal growth stage of barium titanate synthesis at low temperature hydrothermal conditions. During the precursor dissolution stage, when the temperature deviation is "large" and the rate of change is "rapid," the proportional coefficient is automatically adjusted to 1.2-1.5 times the initial value, and the integration time is shortened to 60%-80% of the initial value. During the crystal nucleation stage, when the pH deviation is "medium" and the rate of change is "relatively fast", the proportionality coefficient is maintained at its initial value and the derivative time is adjusted to 0.9-1.1 times the initial value; During the crystal growth stage, when the pressure deviation is "small" and the rate of change is "stable", the integral time is extended to 1.1-1.3 times the initial value, and the PID parameters for the corresponding reaction stage are precisely and dynamically adapted.

5. The control system for the low-temperature hydrothermal synthesis of barium titanate according to claim 4, characterized in that, The pH adjustment submodule and the feed metering submodule establish a "feed-premix-pH real-time compensation" linkage logic: When the feed metering submodule delivers titanium source raw materials, the pH adjustment submodule controls the premixing buffer chamber to automatically draw in the corresponding volume of buffer solution according to the molar ratio of titanium source feed to buffer solution of 1:0.9-1.

1. The buffer solution is then mixed in the chamber for 1-2 seconds to form a premixed liquid. The premixing buffer chamber is a subordinate control component of the pH adjustment submodule. The premixing buffer chamber is filled with 0.08-0.12 mol / L barium hydroxide buffer solution with the same composition as the barium titanate source. The pH adjustment submodule has a built-in pH sensing unit that collects the pH value of the premixed solution in real time. If the pH of the premixed solution deviates from the target range of 11±0.1 by more than 0.1, the pH adjustment submodule controls the premixed buffer chamber to immediately add buffer solution. After the pH of the premixed solution stabilizes within the target range of 11±0.1 for more than 0.5 seconds, it is then injected into the reaction vessel.

6. The control system for the low-temperature hydrothermal synthesis of barium titanate according to claim 5, characterized in that, The control module also has a built-in reaction process prediction model submodule, which is trained and generated based on historical reaction data. Based on the real-time collected temperature and pH value change trends, it predicts the growth rate of barium titanate crystals and the purity of the product, and outputs parameter pre-adjustment instructions in advance.

7. The control system for the low-temperature hydrothermal synthesis of barium titanate according to claim 6, characterized in that, The reaction process prediction model submodule uses complete process data of at least 50 batches of low-temperature hydrothermal synthesis of barium titanate as the training set. The data dimensions include the initial molar ratio of raw materials (Ti:Ba), temperature, pressure / pH time series data of each reaction stage, stirring rate curve, and final particle size distribution / purity of the product. The reaction process prediction model submodule adopts a fusion algorithm of BP neural network and random forest, and the output dimensions include: real-time predicted value of crystal growth rate, estimated value of crystal grain size distribution width, and estimated value of product purity deviation. When the predicted crystal growth rate is less than 10% of the preset baseline value, the reaction process prediction model submodule outputs a pre-adjustment command: increase the stirring rate by 5-8 r / min and fine-tune the reaction temperature by 0.3-0.5℃.

8. The control system for the low-temperature hydrothermal synthesis of barium titanate according to claim 7, characterized in that, The reaction process prediction model submodule has a built-in iterative update unit. After accumulating 20 batches of valid reaction data, the iterative update unit automatically selects process optimization cases from the 20 batches of valid reaction data and retrains the model weights to keep the prediction accuracy stable within ±2%. The valid reaction data refers to the reaction data generated during the low-temperature hydrothermal synthesis of barium titanate, and must simultaneously meet the following requirements: process parameters meet preset thresholds, with deviation time ≤ 2 seconds and amplitude ≤ 3%; the full-process parameter acquisition frequency is ≥ 10 Hz, single batch data loss is ≤ 5 seconds and includes complete control records, status records and alarm records; product purity is ≥ 99.5%, particle size variation coefficient is ≤ 5% and the crystal form is tetragonal; the acquisition sensor has been calibrated within 30 days and is within its validity period, and the measurement error meets the preset accuracy.

9. The control system for the low-temperature hydrothermal synthesis of barium titanate according to claim 8, characterized in that, The safety early warning module has a built-in abnormal operating condition classification and handling submodule. This submodule classifies abnormalities where parameters exceed preset process thresholds into three levels and matches them with differentiated handling strategies: A parameter deviation of less than 5% from the threshold is considered a Level 1 anomaly: triggering each submodule of the control module to fine-tune the corresponding parameters, while simultaneously recording the anomaly node; A parameter deviation of 5%-10% from the threshold is considered a Level II anomaly: The emergency response submodule controls the injection of 0.1 mol / L potassium hydroxide emergency buffer solution into the reactor, with the injection volume being 0.5%-1.0% of the reaction liquid volume; the stirring drive submodule controls the reduction of the stirring rate to 80% of the original rate. If the parameter deviates from the threshold by more than 10%, it is a Level 3 anomaly: the safety warning module triggers an audible and visual alarm and drives the execution module to perform an emergency shutdown. At the same time, it drives the temperature regulation submodule to control the rapid circulation of the cooling medium in the reactor jacket, reducing the reaction temperature to room temperature within 10 seconds. The abnormal operating condition classification and handling submodule is linked with the reaction process prediction model submodule, and triggers the corresponding level of pre-handling strategy 1-2 minutes in advance based on the predicted parameter change trend.

10. The control system for the low-temperature hydrothermal synthesis of barium titanate according to claim 9, characterized in that, The execution module also includes a data traceability and process replication submodule, which communicates bidirectionally with the control module to extract the complete process parameter sequence corresponding to historical valid reaction data and generate a standardized process parameter package. When a new reaction batch is started, the process parameter package is directly called. The control module automatically synchronizes the preset process threshold to each sub-module to achieve replication of the barium titanate synthesis process under the same process conditions. The data traceability and process replication submodule supports binding and labeling process parameter packages with the performance indicators of corresponding batches of products, and supports rapid process matching according to corresponding performance requirements; the performance indicators include barium titanate purity and particle size.

Images