A multi-parameter coupling-based acrylamide aqueous solution storage online regulation system

Abstract

The application discloses a kind of in-line regulation and control system based on multi-parameter coupling in the technical field of data acquisition and advanced process control, the application obtains temperature-pressure-concentration data to generate dimensionless measured parameter set, extracts external meteorological thermal disturbance feedforward prediction value, and based on historical fluctuation, extracts dynamic damping coefficient representing temperature control and pressure relief response time difference;The solving module uses the feedforward prediction value to dynamically shrink the safety threshold, calculates the spatial characteristic distance of the parameter set and the composite nonlinear out-of-limit gain, generates the comprehensive risk index;When out-of-limit, extract the dominant dimension feature, use the damping coefficient to smooth the action slope, output the decoupling control instruction of time sequence isolation.Effectively decouples the interaction of multi-dimensional parameters, eliminates system shock and physical fatigue under traditional control, and improves the adaptive optimization ability and global stability of the control system under complex storage conditions.

Description

Technical Field

[0001] This invention relates to the field of data acquisition and advanced process control technology, and in particular to an online control system for storing acrylamide aqueous solution based on multi-parameter coupling. Background Technology

[0002] As chemical fluid storage technology evolves towards greater automation and intelligence, data acquisition and advanced process control systems are increasingly being applied in complex and dynamic operating conditions. Modern industry not only requires basic safety monitoring for fluid storage, but also emphasizes online adaptive optimization and dynamic equilibrium under the coupled changes of external environmental disturbances and internal physicochemical properties.

[0003] For the storage of acrylamide aqueous solutions, which are highly sensitive to environmental conditions, existing technologies still face significant challenges in handling the interaction of multidimensional physical quantities and the timing of underlying hardware execution. For example, Chinese patent document CN119575897A proposes an intelligent production management system and method for highly concentrated polyacrylamide fracturing fluid, which achieves robust control of the production process through multi-source heterogeneous data acquisition and model prediction. However, in actual long-term storage, this macro-level production scheduling approach still has limitations: the traditional static single-dimensional threshold alarm mode fails to effectively decouple the nonlinear interaction between temperature, pressure, and concentration; and it fails to overcome the asymmetric response barrier between physical actuators. For example, when facing extreme heat intrusion such as summer sun exposure, there is a huge time scissor difference between the thermodynamic lag of the cooling water pump and the pneumatic speed of the pressure relief valve. Without the issuance of synchronous commands for spatiotemporal decoupling, it can easily lead to violent fluctuations in the internal pressure of the fluid and mechanical fatigue of the pipeline network. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an online control system for acrylamide aqueous solution storage based on multi-parameter coupling. By deeply integrating predictive feedforward bias of external meteorological heat flux, nonlinear topological distance of heterogeneous parameters within the system, and dynamic damping characteristics of underlying hardware response delay, a closed-loop collaborative control hub is constructed, consisting of "sensory alignment - dynamic boundary contraction - flexible suppression of command slope." This solves the problems in the background technology.

[0005] The objective of this invention is achieved as follows:

[0006] An online control system for storing acrylamide aqueous solution based on multi-parameter coupling, comprising:

[0007] The data collaborative sensing module is configured to: acquire the internal temperature, pressure and concentration data collected in real time by the on-site state sensing equipment, and generate a multi-dimensional measured state parameter set characterizing the current internal fluid state of the storage tank after dimensional unification processing;

[0008] Simultaneously acquire environmental meteorological forecast data and outer wall temperature from external sensors and network interfaces, and extract external thermal disturbance feedforward compensation forecast values ​​to characterize the trend of external heat intrusion.

[0009] Based on the acquisition of response delay characteristics of historical control fluctuations of pre-configured field actuators, dynamic damping compensation coefficients are extracted to characterize the differences in different physical response rates within the field actuators.

[0010] The adaptive feature calculation module is configured to: compare the multi-dimensional measured state parameter set with the preset working condition safety threshold range, and simultaneously introduce the external thermal disturbance feedforward compensation prediction value as the bias to perform compensation fusion calculation, thereby outputting a comprehensive over-limit risk assessment index that characterizes the overall system operation over-limit probability.

[0011] When it is determined that the comprehensive over-limit risk assessment index exceeds the upper limit of the working condition safety threshold range, the dominant weight dimension feature that leads to the increase in risk is automatically extracted. At the same time, the dynamic damping compensation coefficient is used to perform a smooth dimensionality reduction operation on the initial correction amount for the dominant weight dimension feature, and finally outputs the smooth decoupling control command parameters.

[0012] The decoupling strategy execution module is configured to generate drive instructions for the field execution component based on the smooth decoupling control instruction parameters. The drive instructions are configured to drive the field execution component to perform quantitative adjustments to the output power and start / stop frequency.

[0013] Furthermore, the data collaborative sensing module is specifically configured to extract the dynamic damping compensation coefficient as follows:

[0014] Extract the continuous numerical fluctuation curves of the internal temperature and pressure within a preset historical time window;

[0015] Extract the time lag characteristics between the moment when the internal temperature reaches its peak and the moment when the pressure begins to fluctuate accordingly;

[0016] The time hysteresis feature is mapped to a preset hardware response delay mapping table to generate the dynamic damping compensation coefficient, which characterizes the degree of misalignment between the action delay of the temperature control device and the pressure relief device in the field execution component.

[0017] Furthermore, the data collaborative sensing module is specifically configured to extract the feedforward compensation prediction value of the external thermal disturbance as follows:

[0018] Calculate the rate of temperature rise of the outer wall temperature within a predetermined time period;

[0019] By performing a heat conduction delay mapping between the temperature rise rate and the high temperature label in the environmental meteorological prediction data, the system exposure time node and the expected temperature rise limit of the external heat penetrating into the storage tank are determined;

[0020] Based on the system exposure time point and the expected temperature rise limit, a feedforward compensation prediction value for external thermal disturbance is generated to compensate for the lag in internal fluid temperature perception.

[0021] Furthermore, when the adaptive feature calculation module compares the multidimensional measured state parameter set with the preset operating condition safety threshold range and simultaneously introduces the external thermal disturbance feedforward compensation prediction value as a bias to perform compensation fusion calculation, it is specifically configured as follows:

[0022] Based on the predicted value of the external thermal disturbance feedforward compensation, a dynamic downward compensation calculation is performed on the preset operating condition safety threshold range to generate a strictly controlled safety threshold after dynamic downward compensation.

[0023] The multidimensional measured state parameter set is mapped to the working condition feature distribution model, and the spatial feature distance vector of the multidimensional measured state parameter set relative to the strictly controlled safety threshold after dynamic down-compensation is calculated within the model.

[0024] Based on the calculated spatial feature distance vector, the comprehensive over-limit risk assessment index is generated to configure the system to perform a pre-warning interception action before the internal fluid temperature reaches the critical upper limit.

[0025] Furthermore, when the adaptive feature calculation module performs smooth dimensionality reduction calculation on the initial correction amount for the dominant weight dimension feature using the dynamic damping compensation coefficient, it is specifically configured as follows:

[0026] Based on the hardware adjustment characteristics of the execution components corresponding to the dominant weight dimension features, a preset basic control step size is extracted as the initial correction amount;

[0027] The dynamic damping compensation coefficient is converted into a command smoothing suppression ratio and combined with the preset basic control step size to generate a target motion slope parameter with physical response suppression characteristics.

[0028] The control parameter execution data, which includes the target action slope parameter, is fused together to output the smooth decoupled control command parameters, so as to avoid control overshoot and frequent start-stop caused by inconsistent response speeds of different execution components.

[0029] Furthermore, the decoupling strategy execution module is specifically configured to generate drive instructions for the field execution component as follows:

[0030] The smooth decoupling control command parameters are analyzed, and the cooling target flow rate index and the pressure relief target opening index are extracted at the command level.

[0031] Based on the channel isolation and decoupling rules, a first control signaling is generated for the field cooling component included in the field execution component and a second control signaling is generated for the field pressure relief component included in the field execution component.

[0032] The first control signaling and the second control signaling are combined based on the dynamic damping compensation coefficient to form a drive command with a timing buffer isolation zone feature and issued, so that the field execution unit performs quantitative adjustment of the output power and start / stop frequency within the interleaved execution cycle with the time hysteresis feature compensation.

[0033] Furthermore, the system also includes a fatigue-tolerant fault-tolerant control module, which is configured as follows:

[0034] Continuously monitor the long-term decay trend sequence of the dynamic damping compensation coefficient carried within the parameters of the smooth decoupling control command;

[0035] If the assessment determines that the long-term decay trend sequence continuously breaks through the hardware mechanical aging and sluggishness threshold that characterizes mechanical decay, then a cross-level backup resource scheduling signaling parallel to the drive command is automatically triggered and generated.

[0036] The bypass backup resource scheduling signaling is configured to trigger the plant's public power pipeline network to perform directional allocation of cooling capacity, thereby bypassing the conventional process and prioritizing the redirection of backup cooling flow in the public cooling pipeline network system to the acrylamide aqueous solution storage tank node that is currently in an abnormal state.

[0037] Furthermore, when the adaptive feature calculation module calculates the spatial feature distance vector of the multidimensional measured state parameter set relative to the dynamically down-compensated strict safety threshold within the model, it is further specifically configured as follows:

[0038] Perform difference stripping operation on the temperature, pressure and concentration dimensions included in the multidimensional measured state parameter set to extract the single-dimensional safety critical difference that represents the deviation between the current value of each dimension and its corresponding strict safety threshold.

[0039] Based on a preset nonlinear risk gain mapping benchmark, a numerical mapping is performed on the single-dimensional safety critical difference to allocate an exponentially amplified over-limit approximation gain coefficient to the dimension approaching the red line boundary.

[0040] The over-limit approximation gain coefficient is combined with the corresponding dimensional state parameter in the multidimensional measured state parameter set and a weighted fusion operation is performed to generate a multi-parameter coupled risk feature value containing weighted risk assessment indicators of each dimension. The multi-parameter coupled risk feature value is then output as the spatial feature distance vector.

[0041] Furthermore, when assessing and determining that the long-term decay trend sequence continuously exceeds the hardware mechanical aging sluggishness threshold characterizing mechanical degradation, the equipment fatigue fault-tolerant takeover module is further specifically configured as follows:

[0042] Capture the action response logs of the field actuators during multiple independent and discrete cooling cycle control cycles;

[0043] In the action response log, abnormal discrete points representing transient operating condition fluctuation error values ​​are removed, and the mechanical wear delayed change rate in the time dimension of the long-term decay trend sequence is extracted.

[0044] The mechanical wear-induced hysteresis rate is compared with the preset benchmark hardware aging hysteresis rate using a time-series variation gradient.

[0045] If the comparison determines that the mechanical wear-induced lag rate exceeds the tolerance upper limit boundary corresponding to the preset benchmark hardware aging lag rate, then it is confirmed that the field execution component has undergone irreversible mechanical aging, thereby triggering the generation of the bypass backup resource scheduling signaling.

[0046] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0047] This invention simultaneously acquires environmental meteorological forecasts and external wall temperature, extracts the predicted value of external thermal disturbance feedforward compensation, and uses it as an offset to dynamically adjust and compensate for the preset operating condition safety threshold. It transforms the traditional "passive cooling after overheating" into "active defense before heat penetration," and by compressing the safety red line, it provides sufficient physical ramp-up time for the lagging cooling water pump, effectively eliminating the risk of control failure caused by thermodynamic conduction lag.

[0048] This invention extracts dynamic damping compensation coefficients based on the historical time delay characteristics of on-site temperature control and pressure relief devices. When generating control commands, these coefficients are used to perform smooth dimensionality reduction and slope shaping on the initial correction amount of the dominant dimension. This mechanism compensates for the "time scissors difference" between slow cooling and rapid pressure relief, issuing commands within staggered, time-isolated cycles. This avoids drastic pressure drops and water hammer effects in the pipeline network caused by concurrent commands, extending the service life of mechanical actuators.

[0049] In solving the characteristics of a multidimensional set of measured state parameters (temperature, pressure, concentration), this invention not only unifies the dimensions of heterogeneous data, but also assigns an exponentially amplified nonlinear over-limit approximation gain coefficient to the single-dimensional safety critical difference that approaches the strictly controlled safety threshold. This enables extreme deterioration trends in a single dimension to instantly trigger macroscopic system-level interception, enhancing the system's sensitivity to capturing local abnormal operating conditions.

[0050] This invention continuously monitors the long-term mechanical wear-induced delayed change rate of the dynamic damping compensation coefficient through a fatigue-tolerant control module. When the conventional closed-loop system faces a failure boundary due to irreversible aging of the equipment (such as severe impeller wear delaying response), the system can overcome the local PID control dead zone, generate a cascading backup resource scheduling signal, and directly extract cooling capacity from the plant's public power pipeline network. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0052] Figure 1 This is a schematic diagram of the overall engineering architecture and business flow scenario of an online control system for storing acrylamide aqueous solution based on multi-parameter coupling, according to the present invention.

[0053] Figure 2 This is a flowchart of the process from multi-source sensing to threshold calculation and nonlinear weighting in this invention.

[0054] Figure 3 This is a flowchart of the process from indicator determination to time-series isolation execution and aging bypass takeover in this invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Please see Figure 1 and Figure 3 The present invention provides a technical solution:

[0057] This invention provides an online control system for storing acrylamide aqueous solution based on multi-parameter coupling, including a data collaborative sensing module, an adaptive feature calculation module, and a decoupling strategy execution module. The specific implementation details are as follows:

[0058] The data collaborative sensing module is configured to perform data extraction and dimension aggregation steps for the internal state. This module acquires internal temperature, pressure, and concentration data in real time through field state sensing devices deployed inside the acrylamide storage tank (e.g., explosion-proof PT100 RTD sensors, standard 4mA-20mA analog output piezoresistive transmitters, and online concentration probes). After acquiring the raw analog values, a buffer alignment queue based on the system's common clock is established. The buffer alignment queue is configured as a data container containing a timestamp recording area, a sensor physical channel index area, and a discrete sample value storage area. During data writing, the data acquisition process fills the corresponding sensor physical channel index area with the real-time values ​​returned by different field state sensing devices along with their received timestamps, according to the hardware port address. During synchronous sampling, the system uses a preset synchronization period. Using the absolute time node as the reading reference, the latest discrete sampled values ​​written by each physical channel in the queue are extracted, and these are then assembled to generate discrete parameters at the output terminal, mapped to the same time reference. The system synchronization period is also used. The meaning is the global time span of multi-channel analog-to-digital conversion and latching; in this embodiment, its preferred value is 500 milliseconds.

[0059] The data collaborative sensing module performs dimensional unification processing on the aforementioned heterogeneous raw data: the system retrieves the preset full-scale safety process reference values ​​for each physical parameter from the local memory, and performs a division ratio operation between the real-time acquired values ​​and their corresponding full-scale safety process reference values, thereby converting the unit-based engineering quantities into pure percentage figures. This is then used to generate a value denoted as... The multidimensional measured state parameter set; it represents the dimensionless normalized physical condition evaluation set of the fluid inside the storage tank in three dimensions: temperature, pressure, and concentration. The normalized calibration formula built into the data collaborative sensing module in this embodiment is as follows: It obtains the measured value of the heterogeneous sensor in the current i-th dimension through the standard analog input channel of the PLC. Retrieve the full-scale safety process reference value corresponding to the preset physical parameter from the local memory. The division ratio operation is performed, and the min function of the upper limit clamping operator is introduced. If the measured value of the heterogeneous sensor is less than or equal to the full-scale safety process reference value, the numerical result of the division ratio operation is directly output; if the comparison determines that the measured value of the heterogeneous sensor is greater than the full-scale safety process reference value, the upper limit clamping operator triggers a numerical overwrite operation, and the corresponding dimension channel is output to the multi-dimensional measured state parameter set. The parameter value is assigned the constant 1.

[0060] Among them, full-scale safety process reference value This characterizes the ultimate physical load that the storage tank can withstand under the material mechanics and chemical explosion-proof specifications. In this embodiment, for acrylamide fluid, the full-scale safety process reference value for the pressure dimension is preferably set to 1.6 MPa, and the full-scale safety process reference value for the temperature dimension is preferably set to 45°C. This ensures that the mapped values ​​of each parameter set are constrained within the dimensionless range of [0,1].

[0061] Furthermore, the data collaborative sensing module sequentially executes the feedforward feature extraction step based on external thermal disturbances. The data collaborative sensing module synchronously acquires the outer wall temperature collected by external sensors (such as armored thermocouples attached to the sunlit side of the tank's outer wall) and receives accessed environmental meteorological forecast data via the factory Ethernet / standard API interface. When parsing the environmental meteorological forecast data, the system extracts the predicted environmental temperature value carried within the data packet and continuously compares this predicted environmental temperature value with a preset high-temperature warning baseline. This high-temperature warning baseline is an offline, fixed warning point based on local historical summer extreme weather records and the heat absorption conversion efficiency of the tank's outer wall material; as a specific embodiment, this high-temperature warning baseline is preferably configured as 35°C.

[0062] If the predicted ambient temperature is determined to be higher than the high-temperature warning baseline, the system generates a high-temperature tag in a high-level logic state and writes it to a local register as a Boolean trigger condition. After detecting that the local register is set, a time-window-based state latching operation is performed: the tag is given a valid holding time. During this holding period, the logic for determining external high-temperature intrusion remains true. The effective holding time of this state is... This refers to the reduced-dimensional survival period of a single network weather warning data in the control loop; in this embodiment, its preferred value is 60 minutes. If the value exceeds the upper limit of the range (e.g., greater than 120 minutes), the system will still maintain the high-temperature defense logic when the actual weather has turned into cloudy and rainy weather with cooling, which will lead to ineffective overcooling operations and energy waste; if the value is lower than the set lower limit (e.g., less than 15 minutes), the high-temperature status indicator is prone to frequent clearing and resetting within the continuous control cycle due to external network latency or occasional packet loss at the interface.

[0063] Extracting the feedforward compensation prediction value of external thermal disturbance, the data collaborative sensing module is specifically configured to: extract a predetermined time period (e.g., a sliding time window of the most recent 15 minutes of data), perform temporal feature difference calculation on the external wall temperature to obtain the rate of temperature rise, denoted as... Its feature extraction logic is expressed as follows: Sampling points are continuously pushed into the buffer queue using external patch-type armored thermocouples to extract the absolute temperature of the outer wall at the current moment. And compared with the outer wall temperature extracted from the tail of the queue at the initial moment of the time window. Perform the differential stripping operation; then, the obtained temperature difference is controlled within a preset sliding time window span. Perform dimensionality reduction by division, and output the rate of temperature rise characterizing the nonlinear heating acceleration caused by the absorption of external thermal radiation by the tank shell. The sliding time window span The preferred setting is 15 minutes, and it is designed to account for the rate of temperature rise. The hardware monitoring effective signal interception lower limit is set at 0.5℃ / minute. The rationale for this setting is that the rate of temperature rise during the alternation of dawn and dusk or the smooth rise of normal temperature in natural environments is less than 0.1℃ / minute; if small-amplitude temperature rise rates are not intercepted, frequent small fluctuations can easily cause the feedforward cooling system to fall into a fatigue start-stop state with high-frequency jitter; only when a steep temperature rise rate exceeding 0.5℃ / minute is it determined that an abnormal high-power heat intrusion condition such as heat wave exposure or surrounding fire is encountered, and then the warning characteristic parameter is allowed to be passed to the subsequent logic.

[0064] The data collaborative sensing module performs a heat conduction delay mapping between the extracted temperature rise rate and the high-temperature label in the environmental meteorological forecast data. The specific processing of this mapping logic is as follows: It extracts the absolute thickness data of the current tank outer shell covering material and retrieves the preset specific heat capacity constant and thermal conductivity constant of the polyurethane material. Within the logic control unit, it constructs and generates a feature matrix of the thermal resistance and specific heat capacity of the tank insulation layer. This feature matrix is ​​a static topological data container used to characterize the spatial permeation state of heat flux. The specific construction method and data mapping relationship of this feature matrix are as follows: Based on the absolute thickness data of the outer shell covering material, the entire insulation layer is logically divided into M equidistant virtual heat transfer physical layers in the radial depth according to a preset spatial resolution, thereby establishing the feature matrix as an M×M order square matrix. In the mapping construction of this data container, the main diagonal elements of the feature matrix are configured to independently store the thermal resistance characteristic values ​​of each virtual heat transfer physical layer. These characteristic values ​​are calculated by quotienting the pre-set thermal conductivity constant of the polyurethane material with the discrete layer thickness and then normalized before being written into the matrix. The second diagonal elements of the feature matrix are configured to store the thermal capacity coupling transfer coefficient between two adjacent virtual physical layers. This coefficient is mapped and written into the matrix by multiplying the specific heat capacity constant of the polyurethane material with the material density parameter. The remaining elements in the matrix are set to zero to indicate that there is no direct heat conduction between non-adjacent physical layers. The system inputs the current outer wall heating rate into the "thermal resistance and specific heat capacity feature matrix of the storage tank insulation layer" solidified within the system to perform iterative state dot product derivation. The outer wall heating rate is transformed into an initial heat flux state column vector with M dimensions (where only the first row element is assigned an actual value, and the remaining elements are set to zero, indicating that the heat initially resides only at the outermost boundary). With a preset virtual time step as the control period, the feature matrix is ​​used to continuously perform right multiplication dot product operations on the heat flux state column vector. In each discrete cumulative iteration, the thermal energy data carried in the vector, under the suppression of the thermal impedance attenuation operator on the main diagonal of the feature matrix and the transmission effect of the elements on the secondary diagonal, undergoes numerical drift and penetration into the deeper dimensions of the column vector (the virtual physical layer index near the internal fluid). The system continuously monitors the numerical state of the Mth element in the state column vector (mapping the innermost physical layer); when the evaluation determines that the value of this terminal element crosses the preset internal thermal balance breaking threshold for the first time, the system immediately terminates the cumulative iteration loop and extracts the total number of matrix dot products executed so far; multiplying this total number by the virtual time step, the system outputs the theoretical time taken for the external heat flux to completely penetrate the insulation layer and reach the inner wall.

[0065] The internal thermal equilibrium disruption threshold is defined as the equivalent heat flux span required to trigger an irreversible temperature rise in the acrylamide fluid in the boundary layer adjacent to the inner wall. During system initialization, this threshold is generated by an offline calibration mechanism: the system retrieves the locally stored acrylamide specific heat capacity constant and the contact area with the inner wall of the tank, calculates the critical heat energy required to raise the fluid temperature by 0.1℃, and converts it into a value with the same dimensions as the end element, which is then directly solidified as the internal thermal equilibrium disruption threshold.

[0066] Further determine the system exposure time point when external heat penetrates into the interior of the storage tank, denoted as It represents the physical prediction of the critical moment when the internal acrylamide aqueous solution is about to produce a substantial chain reaction of temperature rise.

[0067] Before performing subsequent thermodynamic temperature rise limit calculations, the data collaborative sensing module performs scalarized conversion of the actual intruding thermal energy in this embodiment. Specifically, during the initialization phase, a static parameter benchmark characterizing the heat transfer properties of the current tank surface material is pre-fixed in the local memory. As a specific embodiment, the static parameter benchmark includes the natural convection heat transfer coefficient characterizing the airflow heat transfer resistance. For the outdoor polyurethane insulation layer in this embodiment, this coefficient is preferably calibrated to 15 W / (m²). 2 The system extracts solar radiation (·K) and solar shortwave radiation absorptivity (preferably calibrated to 0.65), which characterizes the shell's ability to absorb sunlight. The system performs parallel heat flux component extraction: it extracts the current outer wall temperature and performs a subtraction operation on the absolute air temperature value from the environmental meteorological forecast data to generate an internal and external environmental temperature difference feature term. This feature term is then multiplied and mapped with a preset natural convection heat transfer coefficient to generate a convective heat transfer component characterizing the intensity of temperature difference conduction. Based on the accessed environmental meteorological forecast data, it extracts the predicted solar radiation illuminance value (W / m²) for the current time period. 2 The radiative heat injection component is then multiplied and attenuated with the solar radiation absorptivity to generate a radiative heat injection component characterizing the intensity of photothermal conversion. The convective heat transfer component and the radiative heat injection component are then scalar superposition and fusion operations are performed at the system architecture layer to calculate the thermal energy density value penetrating the outer interface of the tank at the current moment, and this value is uniformly assigned and output as the outer wall heat flux.

[0068] In this embodiment, the foreseeability period for this time point ranges from 30 to 120 minutes into the future, allowing sufficient mechanical response time for the actuator. The expected temperature rise limit is calculated based on the aforementioned external wall heat flux and denoted as... Its characteristic is that if no pre-cooling intervention is taken, the fluid inside the storage tank may passively approach the extreme limit of the highest temperature.

[0069] The system extracts the external wall heat flux calculated in the preceding steps and retrieves the current rated liquid level volume of the storage tank and the comprehensive specific heat capacity physical constant of the acrylamide aqueous solution from the local static parameter configuration table. The local static parameter configuration table contains entries for the physical dimensions of the engineering equipment and the basic physical property constants of the medium. The data values ​​in these entries are statically written offline during the system initialization phase by the implementers based on the container's internal cavity volume parameters measured on-site and the corresponding physical property data from the standard chemical handbook. In the computational mapping, the external wall heat flux is integrated over a time within the estimated heat penetration time window to extract the expected total heat flux load that may be injected into the system from the external high-temperature environment. This expected total heat flux load is used as the numerator characteristic term, and the product of the liquid level volume and the comprehensive specific heat capacity physical constant is used as the denominator characteristic term, performing a ratio dimensionality reduction operation. The final output is the theoretical maximum positive shift increment of the fluid temperature inside the storage tank caused by this expected intrusion heat energy under the assumption of no active cooling intervention. This increment limit is used as the expected temperature rise limit. The numerical output.

[0070] Based on the system exposure time point and the expected temperature rise limit derived above, the data collaborative sensing module performs an exponential decay weighted mapping operation to generate a feedforward compensation prediction value for external thermal disturbances to compensate for the lag in internal fluid temperature sensing, denoted as The determining logic is as follows:

[0071] It obtains the predicted warming range limit derived from the simulation. This is compared with the pre-configured system thermal response sensitivity weights. Perform multiplication and composition to generate the theoretical maximum thermal shock load; simultaneously extract the system exposure time points. The structural thermal inertia constant, which characterizes the barrier properties of the outer polyurethane insulation material of the storage tank, is introduced. As the denominator, a natural logarithmic decay operation is performed on the exposure time to extract the time-domain approximation factor; finally, the time-domain approximation factor is composited onto the theoretical maximum thermal shock load to output the feedforward compensation prediction value of the external thermal disturbance. This characterizes the defensive cooling preset equivalent that the cooling water pump inverter needs to establish in reverse before heat enters the interior. The system thermal response sensitivity weight is also included. The preferred value range is [0.65, 0.85], and the structural thermal inertia constant is... The system needs to be calibrated during the initial installation phase based on the thickness of the external insulation layer (in this embodiment, the value is 120 minutes). This embodiment introduces an exponential time-domain approximation factor. Achieving a dynamic balance between cooling efficiency and energy consumption: If the external heat wave is still in its early stages (system exposure time point) (Large, far from the current moment), this exponential term will approach 0 infinitely, outputting a weak compensation bias to prevent premature startup of the high-power chilled water pump and energy waste; once the system exposes the time node... As the temperature rapidly converges to near zero (the heat wave is about to penetrate the outer wall), the index jumps to 1, representing the boundary between the full release of heat and the expected temperature rise. Matching cooling signals enable effective compensation for the internal fluid temperature rise dead zone.

[0072] The data collaborative sensing module, attached to the aforementioned front-end sensing network, extracts dynamic damping characteristics based on the acquisition of response delay features of historical control fluctuations of pre-configured field execution components. In the temperature-pressure coupled control of chemical storage tanks, the cooling water pump (temperature control equipment) changes the fluid temperature with thermal inertia lag, while the release of pressure after the pressure relief valve (pressure relief equipment) opens is an instantaneous physical process. If the control system uses traditional synchronous concurrent commands, it is prone to instantaneous pressure drop, causing system oscillations.

[0073] This allows the data collaborative sensing module to extract continuous numerical fluctuation curves within a preset historical time window from the system's local historical database, denoted as... The system's local historical database is a structured time-series data container, internally configured with a control command issuance timestamp field and a sensor physical channel array storage field mapped to it. During the system's daily operation, it records the action enable level transition events of the process monitoring and issuance to the field execution components, writing the absolute time of the transition into the control command issuance timestamp field. From this time point, it continuously extracts dimensionless normalized temperature and pressure measurement values ​​within subsequent natural time periods at a set sampling frequency of 1Hz, and appends them sequentially to the corresponding sensor physical channel array storage fields. As a specific application example, the system extracts a single associated sample from a continuous data array within 300 seconds following a single cooling command. This sample can completely cover the entire cycle from the start-up of the water pump inverter to the thermal inertia decay of the internal fluid's actual thermal response. Based on this time-series trajectory comparison, the system extracts the time lag characteristic between the moment when the internal temperature reaches its peak and the moment when the pressure begins to fluctuate, denoted as... It represents the time difference between the temperature control device and the pressure relief device in the field actuators, in terms of the actual physical influence they exert on the fluid state. This value fluctuates between 45 seconds and 180 seconds.

[0074] Furthermore, in this embodiment, during the water tank commissioning (safety test) phase, step test signals are injected into the PLC controller to record the response time differences between the valve and the pump under different flow rates and opening degrees. These differences are then cleaned, aligned, and solidified in a pre-defined hardware response delay mapping table at the bottom layer (denoted as...). Specifically, during the water transport commissioning phase, under three baseline flow conditions (50%, 75%, and 100% rated circulation flow) and three baseline valve opening conditions (30%, 60%, and 90% full stroke opening), step test signals are injected in a cross-combination. The absolute time difference between the temperature-controlled water pump reaching the target speed and the pressure relief valve reaching the target stroke is captured as the basic anchor point. Based on this, a two-dimensional static grid matrix is ​​constructed with the cooling flow rate as the horizontal axis, the pressure relief opening as the vertical axis, and the time delay difference as the value of the cross cell. This matrix is ​​then solidified into a hardware response delay mapping table. .

[0075] During the continuous online operation phase, when the flow and opening commands of the real-time operating conditions do not directly hit the integer intersection nodes of the two-dimensional static grid matrix, the input parameters are extracted and wrapped around four adjacent basic discrete calibration points on the horizontal and vertical coordinate axes. A smooth mapping is then performed using the bilinear interpolation algorithm.

[0076] Specifically, on the horizontal coordinate axis corresponding to the cooling flow rate, the calibration delay values ​​of the adjacent flow nodes in the left and right horizontal envelopes are linearly weighted and summed for the first time using the horizontal deviation ratio. Subsequently, on the vertical coordinate axis corresponding to the pressure relief opening, the two weighted results obtained in the previous step are used as a benchmark, and a second linear weighted summation is performed using the vertical deviation ratio to calculate the precise initial value of dynamic damping at the intersection of the current non-integer flow rate and opening. The two-dimensional static mesh matrix represents the original mechanical motion scissor difference profile of the system under typical boundary conditions.

[0077] During the online operation phase, the time lag features extracted in real time will be... Passed to registers, then based on the hardware response delay mapping table The embedded motion slope annealing function performs nonlinear calculations to generate dynamic damping compensation coefficients. The operational formula is as follows: It extracts the system's nominal synchronization action base from the configuration. And introduce the maximum mechanical asynchronous span that the hardware can tolerate. The extracted actual time hysteresis features A normalized comparison is performed with this span, followed by the application of a hardware annealing sensitivity factor. After logarithmic shaping, the resulting attenuation constraint ratio is finally composited onto the system's nominal synchronous action base, outputting a dynamic damping compensation coefficient used to smooth the action slope of fast-acting hardware (such as a pneumatic pressure relief valve) in the time domain. The system's nominal synchronous action base number Take a constant of 1 (representing 100% full-speed response action), the maximum mechanical asynchronous span that the hardware can tolerate. Based on the factory parameters on the water pump nameplate, the optimal calibration time is 180 seconds, and the hardware annealing sensitivity factor is... The value is 1.5. The output range of the dynamic damping compensation coefficient is limited to the interval [0.15, 0.85].

[0078] When the temperature control device lags (time hysteresis characteristic) Approaching the limit value), the coefficient smoothly sinks to the bottom of 0.15, and the large opening degree of the pressure relief valve is suppressed to passively wait for the response of the temperature water cooling system, thereby effectively eliminating the violent pressure oscillation of fluid caused by the concurrent extreme value opening action of the traditional system with undamped restraint.

[0079] This embodiment further obtains the preceding features (multidimensional measured state parameter set). External thermal disturbance feedforward compensation predicted value Dynamic damping compensation coefficient The adaptive feature solving module performs dynamic boundary reset and multi-parameter nonlinear spatial distance calculation.

[0080] Table 1: Measured Table of Linkage Between Feedforward Thermal Disturbance and Boundary Dynamic Contraction

[0081]

[0082] Table 1 above is based on the high-temperature exposure monitoring records of the open-air polyurethane insulated storage tanks in the factory area. The data collaborative sensing module fully captures the entire dynamic transmission process of external heat waves penetrating to the hardware control baseline. The temperature readings continuously pushed by the outer wall-mounted thermocouples to the Ethernet interface reveal the nonlinear increase in the temperature rise rate. The initial gentle climb of 0.2℃ / min is filtered by the system's access dead zone, maintaining the internal zero-bias output. As the external thermal radiation power surges and breaks through the hardware access limit, the cumulative iterative calculation of the feature matrix inside the control unit is activated, rapidly incorporating the system exposure time node shortened to 90 minutes into the exponential decay weighted mapping. At this moment, the system thermal response sensitivity weight and the structural thermal inertia constant are cross-coupled, directly driving the feedforward compensation prediction value to jump from zero. The adaptive feature calculation module then captures this compensation bias and, based on the set shrinkage sensitivity factor, issues a write / overwrite instruction to the PLC control register for the pressure alarm red line.

[0083] Furthermore, by introducing the predicted value of external thermal disturbance feedforward compensation as a bias, the preset operating condition safety threshold range (denoted as...) is... Perform dynamic downward compensation calculation. This generates the strictly controlled safety threshold after dynamic downward compensation, denoted as... Its characterization is influenced by the threat of current external meteorological heat waves, resulting in a dynamic safety envelope boundary that is adaptively compressed inward by the system.

[0084] The adaptive feature calculation module has a built-in dynamic threshold compensation formula as follows: By retrieving a preset operating condition safety threshold range (the absolute static red line percentage of that dimension) for any dimension i in the multidimensional feature set. (and simultaneously incorporate the predicted values ​​of external thermal disturbance feedforward compensation obtained from previous deductions) The preset operating condition safety threshold range in this embodiment is written offline statically during the system initialization phase. The specific value of this red line percentage is independently calibrated based on the nonlinear destructive stress characteristics of different physical dimensions under extreme operating conditions. As a specific engineering embodiment, for the pressure dimension, combined with the full-scale safety process benchmark value of 1.6 MPa calibrated by the previous mapping, the corresponding absolute static red line percentage is preferably configured as 80%.

[0085] External thermal disturbance feedforward compensation prediction value Equivalent to the nominal maximum cooling power of the cooling water pump Perform ratio calculations to assess the current percentage of the system's cooling limit that is currently affected by thermal intrusion; introduce a contraction sensitivity factor. This load percentage is converted into a penalty reduction ratio and combined with the original static threshold to output the strict control safety threshold for dimension i of the current operating condition. The nominal maximum cooling power equivalent of the cooling water pump is... Shrinkage sensitivity factor is based on the rated constant fixed on the actual hardware nameplate of the factory. The preferred configuration is 0.15.

[0086] To further construct a unified evaluation base capable of accommodating the aforementioned multidimensional heterogeneous data, a set of continuous operation logs for temperature, pressure, and concentration of similar storage tanks, covering at least one complete calendar year and marked as being in a normal and safe state, is extracted from a historical database. When performing stripping and removal of abnormal discrete noise points, a forest base is constructed using multiple isolation trees. Historical sampling points for temperature, pressure, and concentration are recursively divided into independent leaf nodes within each subspace, and the average path depth of each sampling point in the entire forest is calculated. A preset isolation depth evaluation threshold is extracted. When the average path depth required for a sampling point to be segmented is lower than the isolation depth evaluation threshold, the point is determined to be an easily separable sensor-related sporadic disconnection and mutation noise point, and it is physically erased from the log set. The preset isolation depth evaluation threshold is the geometric distance measure used by the system to determine whether a time-series coordinate point is outside the distribution envelope of the mainstream steady-state data cluster. The theoretical average path length constant is calculated based on the total number of steady-state samples n in the current input log set, and this theoretical average path length constant is combined with a preset anti-false deletion penalty multiplier to establish the isolation depth evaluation threshold. In this embodiment, the isolation depth evaluation threshold is preferably configured to be 0.6 times the theoretical average path length constant.

[0087] The Gaussian envelope distribution algorithm is applied to the cleaned and aligned steady-state data stream. The system traverses the data stream, calculating the mean vectors for the three orthogonal dimensions of temperature, pressure, and concentration, and simultaneously solves for the covariance matrix representing the correlation between fluctuations across dimensions. Based on the calculated mean vectors and covariance matrix, the system calculates the Mahalanobis distance for any sampled coordinate point in the feature space. The deviation vector is obtained by subtracting the real-time input multidimensional measured state parameters from the historical steady-state mean vector. This deviation column vector is extracted and transposed to generate a deviation row vector. The deviation row vector is then multiplied left by the inverse of the covariance matrix, and the resulting operation matrix is ​​multiplied right by the original deviation column vector. The square root of the product is then taken to obtain the final distance value. During the topology fitting convergence process, isosurfaces with a Mahalanobis distance of 3 are directly mapped and anchored to the operating condition feature distribution model. The three-dimensional outer envelope boundary. The mean vector represents the baseline process operation anchor point when the reaction tank is in an absolutely healthy steady state; the covariance matrix represents the variance measure of the mutual coupling fluctuation of temperature, pressure and concentration parameters under historical operating conditions; the calculation threshold of Mahalanobis distance is 3, which represents the 3-sigma statistical confidence boundary covering 99.7% of historical data under a multidimensional normal distribution.

[0088] This model represents a principal component fluid thermodynamic and chemical steady-state logical coordinate system comprising three orthogonal dimensions: temperature, pressure, and concentration. During online operation, the three-dimensional outer envelope boundary of this coordinate system is directly configured and anchored to the aforementioned strictly controlled safety threshold. The adaptive feature calculation module obtains the multidimensional measured state parameter set generated at the current time. This is used as a three-dimensional dynamic coordinate point, and a vector space mapping is performed to the working condition feature distribution model. Within this topological structure, the absolute physical distance coordinates of the current real-time operating condition relative to the reference envelope are established. Furthermore, the adaptive feature calculation module relies on this model to process the multi-dimensional measured state parameter set. The temperature, pressure, and concentration dimensions included are subjected to differential stripping operations.

[0089] The real-time dimensionless current value of each dimension is subtracted from the aforementioned strictly controlled safety threshold and compared to extract the single-dimensional safety critical difference, denoted as . Its one-dimensional stripping feature formula is expressed as: The adaptive feature calculation module extracts the strictly controlled safety threshold that has just completed dynamic contraction when iterating through the dimensions i (including temperature, pressure, and concentration dimensions) of each orthogonal feature channel in the working condition feature distribution model. And subtract the corresponding multidimensional measured state parameters imported from the sensor at the current moment. Output the single-dimensional safety threshold difference for that specific dimension. It represents the remaining engineering safety redundancy in a single physical dimension from the system's dynamic collapse threshold. The output is the single-dimensional safety criticality difference. It is a pure numerical value between 0 and 1. If the single-dimensional safety threshold difference of dimension i approaches 1, it indicates that the current physical state is far below the alarm red line and the system is safe; if the value crosses 0.1 and approaches 0 sharply, it directly indicates that the single dimension (such as the internal pressure of the reactor) is deteriorating drastically, and the difference data will be pushed into the subsequent nonlinear risk amplification network as a sensitive trigger source.

[0090] This embodiment is based on a preset nonlinear risk-gain mapping benchmark (denoted as...). This is represented by an exponentially growing search curve with the natural constant e as its base, which maps the extracted one-dimensional safety threshold differences. The system assigns an exponentially amplified over-limit approximation gain coefficient to the dimension whose numerical value rapidly shrinks and approaches the one-dimensional limit safety boundary, denoted as... It represents the amplified threat weight posed by the deterioration trend of a single parameter to the overall system safety under current chemical fluid storage conditions. Its value range jumps from 1 under normal conditions to over 15 in an instant.

[0091] If a minor, localized impurity reaction occurs inside the storage tank, causing a rapid increase in internal pressure within a short period (the single-dimensional safety threshold difference in pressure dimension)... The pressure value approaches 0, and at this point, the large overall solution volume causes the temperature and concentration parameters to remain stable. If a traditional linear weighted algorithm is used, the dangerous signal of pressure will be "diluted" by the stable signals of temperature and concentration, leading to missed detections in the system. However, in this system, the pressure dimension will instantly capture a high gain coefficient.

[0092] Specifically, the over-limit approximation gain coefficients are composited into the corresponding dimensional feature channels and weighted aggregation is performed to generate spatial topology evaluation parameters containing nonlinear hazard weights for each dimension. This multi-parameter coupled risk fusion formula and nonlinear gain function are expressed as follows: By extracting the single-dimensional safety threshold difference of each dimension. A zero-value lower bound clamping operator (max function) is introduced to intercept negative value reversal anomalies caused by the complete breach of the single-dimensional safety red line under deteriorating operating conditions; a minimum constant to prevent computational overflow is introduced. and the sensitivity coefficient to approaching punishment By applying the exponential amplification rule of the natural logarithm base, a gain coefficient specifically designed for approaching the redline dimension is generated to achieve the maximum approximation gain. ;

[0093] Extract the basic weight configurations pre-assigned to each physical dimension during the offline initialization phase. The steps for determining the basic weight configuration are as follows: obtain the historical equipment failure ledger database, and extract the frequency of system-level physical damage accidents caused by unilateral abnormal exceedances in each physical dimension (temperature, pressure, concentration);

[0094] The division logic is refactored to be more fault-tolerant based on the Laplace-Smoothing defense principle. Specifically, the frequency of a single accident corresponding to a specific physical dimension i is extracted from the historical equipment failure ledger, denoted as . The historical equipment failure ledger is a static mapping dictionary container used to quantify the destructive power of various physical dimensions at the control strategy level. This dictionary container contains anomaly event index fields and strictly corresponding single root cause feature fields. During the offline phase before system deployment, field process engineers, based on past accident repair reports, write the classification feature code of the first direct physical cause (determined to be the sole dominant factor among temperature, pressure, or concentration exceeding limits) that led to equipment downtime or physical damage into the corresponding single root cause feature field. During data extraction, the system traverses this dictionary container and performs classification and cumulative counting on the root cause feature codes belonging to specific physical dimensions to obtain the frequency of a single accident. As a specific mapping execution example, this is combined with the system's preset failure prevention smoothing constant. If the system finds that out of a total of 50 historical accidents, 40 were directly caused by pressure pipeline tearing, 10 were caused by temperature exceeding limits leading to agglomeration, and 0 were caused by concentration leading to precipitation, the system uses the Laplace anti-failure formula to perform a composite dimensionality reduction calculation on the above-mentioned frequency categories.

[0095] And retrieve the preset anti-failure smoothing constant. ;Accumulate the historical event frequencies across all orthogonal dimensions to obtain the unmodified global total event frequency;When performing the final ratio dimensionality reduction calculation, use As the numerator, add the total frequency of global accidents. (Where n1 represents the total number of orthogonal dimensions monitored, including temperature, pressure, and concentration, i.e.) The calculation is performed using the principal component weights, which represent the bias in the allocation of safety resources. This proportion of the probability of disaster is then directly mapped to and fixed as the basic weight configuration. Among them, the failure prevention smoothing constant This refers to the minimum basic attention unit allocated by the system and which cannot be preempted; in this embodiment, its value is preferably set to 1. The basic weight configuration is then combined with this over-limit approximation gain coefficient and the real-time state parameter. Perform compound multiplication;

[0096] Through the multiplicative composition isolated by this channel, multi-parameter coupled risk feature values ​​containing weighted risk assessment indicators of various dimensions are generated in the logic solution unit. Specifically, the generated multi-parameter coupled risk feature values ​​are extracted, and based on preset feature space anchoring rules, their meaning is represented as the magnitude of the absolute directed tangential vector that approximates the envelope surface of the strictly controlled safety threshold along the orthogonal axes of temperature, pressure, and concentration of the current three-dimensional real-time operating condition coordinate point. The adaptive feature solution module performs data encapsulation and type conversion overload operations at the register level, directly outputting the multi-parameter coupled risk feature values ​​as spatial feature distance vectors.

[0097] Finally, a scalar summation is performed on all orthogonal dimension channels to output a comprehensive limit-crossing risk assessment index. Among them, the minimum constant for preventing overflow calculations. Setting it to 0.01 approximates the penalty sensitivity coefficient. The preferred calibration value is 0.15. Basic weight. Must meet .

[0098] Table 2: Record of Single-Dimensional Deterioration Nonlinear Risk Multiplication and Hardware Interlocking Triggering

[0099]

[0100] The fluid phase change expansion caused by the abnormal local impurities in Table 2 above reproduces the deterioration trajectory of a single-dimensional variable instantly approaching the container's yield limit within a physically isolated, sealed pressure test zone. Faced with the surge in dimensionless pressure values ​​transmitted from the PLC analog input channel at high frequency, although the temperature and concentration probe parameters of the storage tank are still within the absolutely safe range, the critical difference in the pressure dimension is drastically compressed to 0.01. The basic weight configuration matrix is ​​detonated by the underlying logarithmic formula within this redundant interval, causing the penalty gain coefficient specific to the pressure channel to explode from the normal level to over 1800 times.

[0101] Furthermore, this cutoff signaling is specifically routed and broken down into three concurrent actions at the physical execution layer:

[0102] The first signal is instantly reset to zero and the frequency converter enable terminal of the upstream feed pump of acrylamide is turned off, cutting off the subsequent flow of polymerization heat from the physical topology source.

[0103] The second signaling bypasses all the aforementioned anti-vibration damping smoothing algorithms, directly boosting and locking the analog output voltage of the chilled water pump to full scale. Furthermore, when the system receives this forced full-scale signal, the inverter does not perform transient jumps, driving the chilled water pump motor to approach 100% of its rated speed within 3 to 5 seconds at the maximum acceleration allowed by the hardware. This asymmetric compensation strategy ensures that when the system determines that fluid thermodynamic runaway has become the primary problem, it decisively abandons protection against conventional mechanical fatigue life, instead releasing full relief cooling capacity while ensuring that the pump hardware does not suffer transient physical burnout.

[0104] The third signaling path sends a long high-level pulse to the backup explosion-proof emergency relief solenoid valve, triggering it to perform a full-open pressure relief action without being restricted by the dead zone. The meaning of the hardware interlock cutoff signaling is: when the system determines that the thermodynamic conditions of the storage tank are approaching the limit of physical runaway, the highest priority forced safety intervention and over-level logic scheduling flag is triggered.

[0105] In this step, the adaptive feature calculation module will generate control parameters that can directly drive the underlying hardware without causing severe overshoot, based on the various comprehensive indicators and damping features extracted above.

[0106] When determining the comprehensive risk assessment indicators for exceeding limits When the upper limit of the safety threshold is exceeded, the system automatically extracts the dominant weight dimension feature that caused the increased risk from the multi-parameter coupled risk feature value calculated above. For example, if the exceedance is mainly caused by a surge in the approximation gain coefficient of the pressure dimension, the system locks the pressure dimension as the dominant dimension; if it is caused by an external feedforward temperature bias, the system locks the temperature dimension as the dominant dimension.

[0107] Based on the inherent hardware regulation characteristics of the execution components corresponding to the dominant weight dimension features (e.g., pressure corresponds to a pressure relief valve, temperature corresponds to a cooling water pump), a preset basic control step size is extracted from the local control strategy library, denoted as . The local control strategy library is a static data container used to establish software and hardware control mapping relationships and output safety control spans. During the initialization phase before system deployment, each physical monitoring dimension of the storage tank is paired one-to-one with the serial numbers of the physically bound hardware devices on-site, and this information is entered into the container's associated configuration area. Based on the maximum single safe action span allowed by the nameplate of each hardware device, specific quantitative control step sizes are configured in the container's strategy output area. The container is traversed using the dominant weight dimension feature as the addressing condition, the corresponding device is retrieved, and its associated basic control step size is output. If the dominant weight dimension feature is a pressure dimension, the system matches the on-site pressure relief valve component inside the container and extracts its corresponding preset basic control step size, which is preferably configured as a 15% increment of the valve's absolute stroke opening. This step size is the initial correction amount.

[0108] If the system directly uses this idealized initial correction amount (preset basic control step size) The command is sent down to the lower level for execution. Due to the objective existence of the aforementioned time lag characteristics (fast pressure release, slow cooling), the module will use the previously generated dynamic damping compensation coefficient. Introduced into the current solution stream and converted into an instruction smoothing suppression ratio, denoted as... It represents a numerical restraint factor used to actively attenuate or delay the intensity of some hardware actions.

[0109] The system will smooth out the proportion of this instruction. Composite to preset basic control step size The above is used to achieve the dimensionality reduction slope modification of the actuator, and its determination logic is as follows:

[0110]

[0111]

[0112] By receiving the dynamic damping compensation coefficient generated from the previous calculation. (The lower this value, the more severe the system delay misalignment), and an instruction smoothness attenuation gain is introduced. The original damping coefficient is reverse-mapped to generate a command smoothing suppression ratio that is directly used to suppress the amplitude of the electromagnetic signal. Extract the preset basic control step size output from the conventional PID core scheduling loop. (Ideal instruction load abrupt change span), multiplicatively weaken and fuse it using the instruction smoothing suppression ratio, outputting the restricted target action slope parameter. Its characterization reflects the flexible control response speed and output voltage / frequency ramp-up gradient per second sent to the underlying PLC analog output channel after the system's anti-vibration active correction. This includes the command smoothness attenuation gain. The preferred range is set to [0.8, 1.2], and this coefficient is fine-tuned based on the mechanical fatigue tolerance of the actuator. This is achieved by introducing a delay-constrained instruction smoothing suppression ratio. If the slope is reduced to a lower limit than the set safety action threshold, the full-speed response permission of the underlying layer will be deprived, and the execution hardware will be gradually activated according to the calculated gentle slope, effectively avoiding control overshoot and oscillation feedback caused by the mismatch between the response speed of the cooling water pump and the pressure relief valve.

[0113] The control parameters, including the target motion slope parameter, are packaged and fused at the data structure level to ultimately output smoothly decoupled control command parameters, denoted as... .

[0114] Table 3: Mapping Table of Time Hysteresis Damping and Command Slope Flexible Suppression

[0115]

[0116] Table 3 above addresses the ideal control step size of 15% full opening of the fast-acting valve required by the PID feedback loop. The communication isolation layer refuses to blindly transmit this control. The computational center converts the normalized hysteresis characteristics into a command smoothing suppression ratio, applying a multiplicative stripping weakening effect to the ramp voltage of the pneumatic valve control analog channel. The execution results of the field control network clearly show that when the pump response hysteresis is severely prolonged to 150 seconds, the system automatically determines that synchronous pressure relief poses a risk of pipeline collapse. At this point, the smoothing suppression ratio drops to a low safety zone of 0.28. The original 15% valve opening abrupt change signal is stripped of its full-speed response authority. The decoupling strategy execution module only outputs a gentle 4.2% ramp signal to the field execution components, using the slowing of physical action to offset the system oscillation damage caused by thermal inertia misalignment.

[0117] Furthermore, in this embodiment, the decoupling strategy execution module processes the received smooth decoupling control command parameters. Protocol-level packet unpacking and parsing are performed. Specifically, based on the channel isolation and decoupling rules, the smooth decoupling control command parameters are separated and decomposed at the multi-dimensional control feature level. The channel isolation and decoupling rules are protocol parsing mapping relationships used to prevent overlapping and conflicting simulated commands from multiple actuators. At the control protocol message level, the smooth decoupling control command parameters are uniformly encapsulated by the system into structured data frames containing fixed-byte lengths. Based on the masked byte addressing mechanism preset in the decoupling rules, the system extracts the first data segment in the high-order address segment of the data frame, which carries the cooling regulation feature, defines it as the cooling flow configuration word, and maps it to the cooling target flow rate index for the cooling equipment. Simultaneously, the second data segment in the low-order address segment, which carries the pressure control feature, is extracted, defined as the pressure relief configuration word, and mapped to the pressure relief target opening index for the pressure relief equipment.

[0118] In this embodiment, to stagger two physical actions with different response hysteresis rates on the time axis, the module generates a first control signal (an analog signal driving the water pump inverter) for the field cooling component included in the field actuator, and a second control signal (a pulse width signal driving the pneumatic head of the pressure relief valve) for the field pressure relief component included in the field actuator, based on the configured channel isolation and decoupling rules. After generating the above signals, the system combines the first and second control signals according to the aforementioned time hysteresis characteristics. The system synthesizes drive instructions with timing buffer isolation features and sends them to the underlying PLC interface. This feature requires: prioritizing the activation of cooling flow rate regulation actions with longer response hysteresis, and ensuring that the hysteresis occurs only when crossing the time hysteresis feature extracted from the preceding sequence. After a matching execution dead zone span (a waiting time to tolerate the thermal inertia response of the water pump, during which no new commands are output to the quick-acting pressure relief valve channel), the opening of the quick-acting pressure relief valve is delayed. Through the above-mentioned micro-command interleaving mechanism, the field actuators ultimately perform quantitative adjustments to the output power and start / stop frequency according to the drive commands within the interleaved execution cycle with time delay compensation (for example, quantitatively adjusting the output power of the water pump inverter according to the first control signal, and quantitatively adjusting the start / stop frequency of the pressure relief valve according to the second control signal).

[0119] This embodiment further addresses the potential failure of conventional PID adaptive control due to hidden wear and tear of equipment in chemical plants after long-term continuous operation.

[0120] Under normal operating conditions, the decoupled control commands can operate stably. However, under severe edge asymmetric conditions (e.g., the impeller of the cooling water pump at the bottom of the storage tank experiences hidden mechanical wear due to long-term transport of fluid containing trace amounts of polymer particles), the actual cooling response time of the water pump will continuously increase. In this case, the equipment fatigue fault-tolerant control module is configured to continuously monitor and intercept the smooth decoupled control command parameters sent to the hardware at the communication bus level via a bypass protocol. Because the command parameters include a dynamic damping compensation coefficient for oscillation prevention. The takeover module performs a parsing operation on the command message to extract the damping values ​​for each cycle, thereby generating and continuously monitoring the long-term decay trend sequence of the dynamic damping compensation coefficient.

[0121] Specifically, it captures the action response logs of field-actuated components within multiple independent and discrete cooling cycle control cycles. During this process, a high-frequency noise filtering algorithm is executed to remove abnormal discrete points, characterized by transient operating condition fluctuations, whose duration is shorter than a set time-domain detection period. These are denoted as... When performing time-domain determination, a sliding time observation window is activated to continuously count the number of control cycles in which the current damping attenuation data exceeds the upper limit of the mean variance. If this number does not reach the set time-domain detection cycle threshold, the data set is classified as a high-frequency glitch and physically deleted; this represents a single instance of non-real hardware response sluggishness caused by a sudden drop in grid voltage or local pipeline air resistance. In this preferred embodiment, the time-domain detection cycle threshold is preferably set to 3 consecutive control cycles. This filtering mechanism will strip away damping data with a duration of less than 3 consecutive control cycles. After stripping away the aforementioned transient interference noise, to further prevent the rise in ambient water temperature caused by seasonal changes from being misjudged as mechanical equipment failure, environmental benchmark compensation is introduced to perform common-mode interference stripping.

[0122] The system synchronously calls the ambient temperature calibration module deployed at the main inlet of the plant's public cooling pipe network to obtain real water temperature reference data. As a specific implementation example, this ambient temperature calibration module uses a conventional insertion-type armored resistance temperature detector or a surface-mount temperature transmitter to measure and output the slope of the reference water supply temperature change of the plant's public cooling pipe network in real time within the same time period.

[0123] The system retrieves a preset water-cooling damping temperature sensitivity mapping coefficient (which is configured as an engineering conversion reference to convert the physical dimension of temperature slope into a dimensionless hysteresis offset scale). As a specific configuration embodiment, for the conventional carbon steel cooling water pipe network in the plant area, this sensitivity mapping coefficient is preferably calibrated to 0.025 [dimensionless / (℃ / day)], which means that for every 1℃ / day increase in the slope of the water supply temperature rise, the fluid heat exchange efficiency will decrease, which will manifest as a thermal inertia offset of 0.025 in the equipment damping coefficient.

[0124] The extracted slope of the baseline water supply temperature change is multiplied and mapped with the sensitivity mapping coefficient to generate a dimensionless seasonal thermal inertia offset factor; this offset factor is then applied to the original daily average dynamic damping compensation coefficient. The difference stripping correction is performed to obtain the net damping attenuation sequence after removing external meteorological disturbances. .

[0125] A linear feature stripping operation is performed on the net damping decay sequence after cleaning and alignment in the time dimension to extract the mechanical wear-induced lag rate of change, denoted as . The slope derivation was performed using the following modified least squares fitting formula to set the long-term aging observation window: By obtaining the valid historical statistical days within the currently set observation time window. And correspondingly retrieve the net damping attenuation sequence generated by common-mode interference stripping in the preceding sequence. Then calculate the time mean respectively. Mean of net damping decay sequence The process involves covariance accumulation and variance division to convert the variance into a single value, ultimately outputting the mechanical wear-induced hysteresis rate, which characterizes the hardware response hysteresis degradation gradient. This parameter converts the discrete damping value into a characteristic of the degradation rate of cooling hardware (such as a water pump impeller) as its response delay increases month by month due to irreversible physical deterioration. The total number of statistical samples N within the observation time window is preferably set to 30 (data from the past month is extracted for long-term fitting).

[0126] The extracted mechanical wear-induced rate of change The data is sent to the aging assessment core stream and compared with the preset baseline hardware aging hysteresis rate (denoted as λ) in the database. The system performs a time-series gradient comparison. Based on the factory-calibrated big data, it checks whether the mechanical wear-induced hysteresis rate exceeds the normal wear safety tolerance range defined by the baseline hardware aging hysteresis rate, and records this as... Its characterization allows for reasonable attenuation of the envelope boundary of equipment such as water pumps within their normal lifespan.

[0127] In this embodiment, based on the lifecycle test logs of multiple batches of the same model of equipment stored in the database, this safety tolerance range... The baseline upper limit is set to 120% of the baseline hardware aging hysteresis rate. The system directly maps the absolute upper limit of this tolerance range to the hardware mechanical aging sluggishness threshold used to characterize the risk of the equipment entering the scrap period.

[0128] The equipment fatigue fault-tolerant takeover module continuously compares and verifies the previously calculated mechanical wear-induced delayed change rate with the static hardware mechanical aging sluggishness threshold. When the system determines that the mechanical wear-induced delayed change rate is within the lower limit range defined by the threshold, it determines that the current system is in a normal wear safety state and maintains the existing local closed-loop control strategy. If the comparison determines that the mechanical wear-induced delayed change rate crosses the upper limit boundary defined by the corresponding threshold (the equipment attenuation rate exceeds the 120% red line limit), the parameter status immediately transitions from silent to extreme aging danger state.

[0129] When the rate of change due to mechanical wear exceeds the upper limit of the aforementioned safety tolerance range, i.e., when the long-term decay trend sequence continuously exceeds the hardware mechanical aging sluggishness threshold, it is immediately logically confirmed that the current on-site cooling actuators have undergone irreversible mechanical aging, and the conventional local closed-loop circuit has completely lost its disaster relief cooling capability. Faced with this condition, the equipment fatigue fault-tolerant takeover module cuts off the conventional control circuit, automatically triggers and generates a cascading backup resource scheduling signal parallel to the conventional drive command, denoted as... Its characterization represents the highest-level emergency drive parameter that breaks through the level of individual storage tank equipment and directly requests energy from the plant's topology network.

[0130] The system directly sends the signal through the plant's industrial Ethernet backbone. The data frame of the tiered backup resource scheduling signal encapsulates a unique node identifier (e.g., a pre-assigned tank IP address or station coordinate) representing the spatial topology location of the currently abnormal tank node, as well as a demand characteristic value representing the maximum required emergency flow. Upon receiving the signal, the core control and scheduling node of the plant's public power pipeline network extracts the unique node identifier and compares it with the fixed pipeline topology addressing routing table. The pipeline topology addressing routing table is a static mapping data container used to achieve cross-level energy spatial directional scheduling. This data container is pre-configured with unique network communication identifiers for each tank node in the entire plant, as well as the controller hardware addresses of the common main pneumatic distribution valves that are directly connected to the cooling system in physical space; the engineering configuration of this mapping relationship originates from the physical routing data of the plant's pipeline network construction drawings. When performing cross-level address matching, the core control scheduling node uses the received unique network communication identifier as the retrieval basis, traverses the routing table to lock the corresponding main pneumatic distribution valve hardware address, and then directly outputs control voltage through the bus to drive the main pneumatic distribution valve physically connected to the branch where the abnormal storage tank node is located to open to the maximum degree. This bypasses the conventional local pipeline cascade distribution process and prioritizes the forced orientation of the backup refrigeration flow in the public cooling network system to the currently abnormal acrylamide aqueous solution storage tank node.

[0131] like Figure 1As shown, the online control system of this invention relies on the physical entity of an open-air storage tank in a chemical plant area, constructing a closed-loop control system that extends from front-end multi-source heterogeneous data acquisition and clustering, central logic feature calculation, to terminal decoupling and isolation. The external detectors and network antennas attached to the outer wall of the large acrylamide aqueous solution storage tank in the center of the diagram, as well as the three internal monitoring probes, directly correspond to and are controlled by the "data collaborative sensing module." This module is responsible for synchronously acquiring a multi-dimensional set of measured state parameters of the internal fluid state and extracting feedforward compensation prediction values ​​representing the trend of external heat intrusion. The industrial control cabinet on the side serves as the control center, corresponding to the "adaptive feature calculation module." Its core logic is responsible for performing feature comparison and compensation fusion calculations between the aforementioned heterogeneous data and the dynamically shrinking operating condition safety threshold range. The line graph on the control cabinet panel physically represents the calculation process of the comprehensive over-limit risk assessment index, as well as the smooth dimensionality reduction calculation triggered by the dynamic damping coefficient when exceeding the safety upper limit boundary. Two dashed lines with physical time delay isolation marks, output from the control cabinet and extending to the cooling water pump at the bottom of the storage tank and the pneumatic pressure relief valve at the top, map the workflow of the "decoupling strategy execution module": Based on the smooth decoupling control command parameters calculated from the system output, the system separates and decomposes the cooling target flow rate index and the pressure relief target opening index at the timing command level, so that the cooling components and pressure relief components on site perform actions within the staggered execution cycle with time delay compensation, thus preventing pipeline water flow oscillation and mechanical fatigue caused by concurrent opening commands.

[0132] The above description of the embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. It should be noted that those skilled in the art can make several improvements and modifications to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. An online control system for storing acrylamide aqueous solution based on multi-parameter coupling, characterized in that, Specifically, it includes: The data collaborative sensing module is configured to: acquire the internal temperature, pressure and concentration data collected in real time by the on-site state sensing equipment, and generate a multi-dimensional measured state parameter set characterizing the current internal fluid state of the storage tank after dimensional unification processing; Simultaneously acquire environmental meteorological forecast data and outer wall temperature from external sensors and network interfaces, and extract external thermal disturbance feedforward compensation forecast values ​​to characterize the trend of external heat intrusion. Based on the acquisition of response delay characteristics of historical control fluctuations of pre-configured field actuators, dynamic damping compensation coefficients are extracted to characterize the differences in different physical response rates within the field actuators. The adaptive feature calculation module is configured to: compare the multi-dimensional measured state parameter set with the preset working condition safety threshold range, and simultaneously introduce the external thermal disturbance feedforward compensation prediction value as the bias to perform compensation fusion calculation, thereby outputting a comprehensive over-limit risk assessment index that characterizes the overall system operation over-limit probability. When it is determined that the comprehensive over-limit risk assessment index exceeds the upper limit of the working condition safety threshold range, the dominant weight dimension feature that leads to the increase in risk is automatically extracted. At the same time, the dynamic damping compensation coefficient is used to perform a smooth dimensionality reduction operation on the initial correction amount for the dominant weight dimension feature, and finally outputs the smooth decoupling control command parameters. The decoupling strategy execution module is configured to generate drive instructions for the field execution component based on the smooth decoupling control instruction parameters. The drive instructions are configured to drive the field execution component to perform quantitative adjustments to the output power and start / stop frequency.

2. The online control system for storing acrylamide aqueous solution based on multi-parameter coupling according to claim 1, characterized in that: The data collaborative sensing module is specifically configured as follows when extracting the dynamic damping compensation coefficient: Extract the continuous numerical fluctuation curves of the internal temperature and pressure within a preset historical time window; Extract the time lag characteristics between the moment when the internal temperature reaches its peak and the moment when the pressure begins to fluctuate accordingly; The time hysteresis feature is mapped to a preset hardware response delay mapping table to generate the dynamic damping compensation coefficient, which characterizes the degree of misalignment between the action delay of the temperature control device and the pressure relief device in the field execution component.

3. The online control system for storing acrylamide aqueous solution based on multi-parameter coupling according to claim 1, characterized in that: When extracting the feedforward compensation prediction value of the external thermal disturbance, the data collaborative sensing module is specifically configured as follows: Calculate the rate of temperature rise of the outer wall temperature within a predetermined time period; By performing a heat conduction delay mapping between the temperature rise rate and the high temperature label in the environmental meteorological prediction data, the system exposure time node and the expected temperature rise limit of the external heat penetrating into the storage tank are determined; Based on the system exposure time point and the expected temperature rise limit, a feedforward compensation prediction value for external thermal disturbance is generated to compensate for the lag in internal fluid temperature perception.

4. The online control system for storing acrylamide aqueous solution based on multi-parameter coupling according to claim 1, characterized in that: The adaptive feature calculation module is specifically configured as follows when it compares the multidimensional measured state parameter set with the preset operating condition safety threshold range and simultaneously introduces the external thermal disturbance feedforward compensation prediction value as a bias to perform compensation fusion calculation: Based on the predicted value of the external thermal disturbance feedforward compensation, a dynamic downward compensation calculation is performed on the preset operating condition safety threshold range to generate a strictly controlled safety threshold after dynamic downward compensation. The multidimensional measured state parameter set is mapped to the working condition feature distribution model, and the spatial feature distance vector of the multidimensional measured state parameter set relative to the strictly controlled safety threshold after dynamic down-compensation is calculated within the model. Based on the calculated spatial feature distance vector, the comprehensive over-limit risk assessment index is generated to configure the trigger system to perform a pre-warning interception action before the internal fluid temperature reaches the critical upper limit.

5. The online control system for storing acrylamide aqueous solution based on multi-parameter coupling according to claim 1, characterized in that: When the adaptive feature solving module performs smooth dimensionality reduction on the initial correction amount for the dominant weight dimension feature using the dynamic damping compensation coefficient, it is specifically configured as follows: Based on the hardware adjustment characteristics of the execution components corresponding to the dominant weight dimension features, a preset basic control step size is extracted as the initial correction amount; The dynamic damping compensation coefficient is converted into a command smoothing suppression ratio and combined with the preset basic control step size to generate a target motion slope parameter with physical response suppression characteristics. The control parameter execution data, which includes the target action slope parameter, is fused together to output the smooth decoupled control command parameters, so as to avoid control overshoot and frequent start-stop caused by inconsistent response speeds of different execution components.

6. The online control system for storing acrylamide aqueous solution based on multi-parameter coupling according to claim 1, characterized in that: When generating drive instructions for the field execution components, the decoupling strategy execution module is specifically configured as follows: The smooth decoupling control command parameters are analyzed, and the cooling target flow rate index and the pressure relief target opening index are extracted at the command level. Based on the channel isolation and decoupling rules, a first control signaling is generated for the field cooling component included in the field execution component and a second control signaling is generated for the field pressure relief component included in the field execution component. The first control signaling and the second control signaling are combined based on the dynamic damping compensation coefficient to form a drive command with a timing buffer isolation zone feature and issued, so that the field execution unit performs quantitative adjustment of the output power and start / stop frequency within the interleaved execution cycle with the time hysteresis feature compensation.

7. The online control system for storing acrylamide aqueous solution based on multi-parameter coupling according to claim 6, characterized in that: The system also includes a fatigue-tolerant fault-tolerant takeover module, which is configured as follows: Continuously monitor the long-term decay trend sequence of the dynamic damping compensation coefficient carried within the parameters of the smooth decoupling control command; If the assessment determines that the long-term decay trend sequence continuously breaks through the hardware mechanical aging and sluggishness threshold that characterizes mechanical decay, then a cross-level backup resource scheduling signaling parallel to the drive command is automatically triggered and generated. The bypass backup resource scheduling signaling is configured to trigger the plant's public power pipeline network to perform directional allocation of cooling capacity, so as to prioritize the allocation of backup cooling flow in the public cooling pipeline network system to the acrylamide aqueous solution storage tank node that is currently in an abnormal state.

8. The online control system for storing acrylamide aqueous solution based on multi-parameter coupling according to claim 4, characterized in that: The adaptive feature calculation module is further specifically configured to, when calculating the spatial feature distance vector of the multidimensional measured state parameter set relative to the dynamically down-compensated strict safety threshold within the model: Perform difference stripping operation on the temperature, pressure and concentration dimensions included in the multidimensional measured state parameter set to extract the single-dimensional safety critical difference that represents the deviation between the current value of each dimension and its corresponding strict safety threshold. Based on a preset nonlinear risk gain mapping benchmark, a numerical mapping is performed on the single-dimensional safety critical difference to allocate an exponentially amplified over-limit approximation gain coefficient to the dimension approaching the red line boundary. The over-limit approximation gain coefficient is combined with the corresponding dimensional state parameter in the multidimensional measured state parameter set and a weighted fusion operation is performed to generate a multi-parameter coupled risk feature value containing weighted risk assessment indicators of each dimension. The multi-parameter coupled risk feature value is then output as the spatial feature distance vector.

9. The online control system for storing acrylamide aqueous solution based on multi-parameter coupling according to claim 7, characterized in that: When the equipment fatigue tolerance takeover module assesses and determines that the long-term decay trend sequence continuously exceeds the hardware mechanical aging sluggishness threshold characterizing mechanical degradation, it is further specifically configured as follows: Capture the action response logs of the field actuators during multiple independent and discrete cooling cycle control cycles; In the action response log, abnormal discrete points representing transient operating condition fluctuation error values ​​are removed, and the mechanical wear delayed change rate in the time dimension of the long-term decay trend sequence is extracted. The mechanical wear-induced hysteresis rate is compared with the preset benchmark hardware aging hysteresis rate using a time-series variation gradient. If the comparison determines that the mechanical wear-induced lag rate exceeds the tolerance upper limit boundary corresponding to the preset benchmark hardware aging lag rate, then it is confirmed that the field execution component has undergone irreversible mechanical aging, thereby triggering the generation of the bypass backup resource scheduling signaling.

Images