Fly ash grading control system based on object identification resolution

Abstract

The present application belongs to the technical field of industrial control system, and relates to a fly ash grading control system based on object identification analysis, comprising: an identifier construction module for generating a unique object identifier; a characteristic decoding module for decoding the identifier to decode historical parameters and obtain a multi-dimensional feature vector; a strategy matching module for matching a target operation primitive set based on the multi-dimensional feature vector and a load parameter; a sequence packaging module for generating a batch control sequence containing a digital signature; an authorization verification module for issuing sequence verification signatures and generating an authorized operation signal; a topology reorganization module for loading sequences to construct an executable logic control topology based on the signal; and an instruction execution module for driving the topology to calculate deviations and generate grading equipment control instructions. The present application solves the problem that the prior art lacks pre-perception of batch characteristic fluctuations, resulting in fixed control strategies and difficulty in adaptive adjustment.

Description

Technical Field

[0001] This invention belongs to the technical field of industrial control systems and relates to a fly ash grading control system based on object identifier parsing. Background Technology

[0002] Fly ash produced by thermal power plants has wide applications in building materials and fine chemicals, and its classification is fundamental to achieving high-value utilization. The physicochemical properties of fly ash, such as the content of vitreous microspheres and its flowability, are dynamically affected by various factors, including changes in upstream boiler combustion conditions and coal type, leading to significant non-steady-state fluctuations in the properties of different batches of material. This uncertainty in the characteristics of the source material poses a challenge to the stable operation of the downstream classification control system, and the existing system's ability to perceive and adapt to this inherent property drift in real time needs improvement.

[0003] Currently, the control of the fly ash grading process mainly relies on two conventional technical methods. The first is open-loop regulation based on offline analysis and manual experience. This method requires operators to periodically take samples and send them to the laboratory for physicochemical testing. The operator then manually adjusts key process parameters such as the classifier speed and the main induced draft fan airflow based on the test results over a long period. The second method is a simple closed-loop control that incorporates feedback on the quality of the final product. This scheme monitors the particle size of the finished product by installing sensors at the discharge port, compares the measured data with the target value, and uses a controller to automatically adjust key actuators such as the frequency converter output of the classifier based on the magnitude of the deviation.

[0004] The two conventional solutions mentioned above have certain limitations when dealing with complex scenarios involving drastic fluctuations in material properties. The open-loop control method based on offline analysis has a long feedback cycle, causing control behavior to lag behind process disturbances. Manual adjustment strategies derived for older batches are often inapplicable to new batches, making real-time optimization difficult. While simple closed-loop control incorporating feedback achieves automated adjustment, its control model is relatively fixed. The controller tends to treat all unmodeled characteristic changes as general external disturbances. When facing materials with poor flowability or agglomeration, the fixed control gain parameters are insufficient to effectively suppress process oscillations, easily leading to controller output saturation or system instability. Existing control systems lack a pre-sensing mechanism for the physical characteristics of each batch of materials, making it difficult to achieve adaptive matching of control strategies from the root cause. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a fly ash grading control system based on object identifier parsing.

[0006] A fly ash grading control system based on object identifier resolution includes:

[0007] The identifier construction module is used to obtain displacement signals of fly ash storage and conveying nodes, extract process condition snapshot data of associated equipment, and generate unique object identifiers.

[0008] The feature decoding module is used to parse unique object identifiers, extract historical operating parameters, and decode historical operating parameters through a physical data hybrid model to obtain multi-dimensional feature vectors.

[0009] The strategy matching module is used to obtain the current mechanical load parameters of the staged fan equipment. Based on the multi-dimensional feature vector and the current mechanical load parameters, it performs matching in the built-in control primitive library to extract the target operation primitive set.

[0010] The sequence encapsulation module is used to perform topological recombination on the target operation primitive set and, in combination with preset process parameters, generate a batch instruction control sequence containing a digital signature.

[0011] The authorization verification module is used to send the batch instruction control sequence to the logic execution interpretation engine, verify the digital signature, and generate an authorization running status signal;

[0012] The topology reconfiguration module is used to load batch instruction control sequences into the microcontroller unit based on authorized operating status signals, and construct an executable logic control topology.

[0013] The instruction execution module is used to acquire the real-time deviation acquisition feedback value of the environmental sensor probe, drive the executable logic control topology to calculate the real-time deviation acquisition feedback value, and generate hierarchical equipment control instructions.

[0014] In a further embodiment of the present invention, the identifier construction module is specifically used for:

[0015] Monitor the mechanical opening and closing action of the discharge valve of the upstream ash storage silo, and simultaneously accumulate the flow pulse value output by the bottom conveyor belt scale. When the flow pulse value exceeds the preset batch tonnage threshold, generate a physical batch start cutoff event.

[0016] Obtain the timestamp parameter and hardware device code of the moment when the start truncation event occurs, extract key frames of fire images within the same time window based on the timestamp parameter, and construct process condition snapshot data;

[0017] The timestamp parameter, hardware device code, and process condition snapshot data are serialized and concatenated, and then fused using a hash function to generate a unique object identifier.

[0018] In a further embodiment of the present invention, the feature decoding module is specifically used for:

[0019] Use the unique object identifier as the query index, and extract the bound timestamp parameter and hardware device code as historical running parameters;

[0020] Based on the timestamp parameter, the boiler combustion temperature time series and the current fluctuation curve of the electrostatic precipitator station are extracted from the associated production cycle.

[0021] A frequency domain transformation operation is performed on the boiler combustion temperature time series to extract the combustion main frequency fluctuation amplitude parameter, and the spark discharge variance index of the current fluctuation curve is calculated and combined to form an intermediate feature vector.

[0022] The intermediate feature vector is standardized and then input into the physical data hybrid model to infer the estimated content of fly ash glass microspheres. The combustion frequency fluctuation amplitude parameter, spark discharge variance index and the estimated content of fly ash glass microspheres are then concatenated to generate a multidimensional feature vector.

[0023] In a further aspect of this invention, the characteristic decoding module, when calculating the spark discharge variance index of the current fluctuation curve, is specifically used for:

[0024] Obtain the current sample value at each data point in the current fluctuation curve, and calculate the average current value of all current sample values;

[0025] Calculate the difference between each current sample value and the average current value, and square the difference to obtain the squared difference of each data point;

[0026] The squared differences of all data points are calculated to obtain the spark discharge variance index.

[0027] In a further embodiment of the present invention, the strategy matching module is specifically used for:

[0028] The estimated content of vitreous microspheres in fly ash was extracted from the multidimensional eigenvectors.

[0029] Obtain the preset reference glass microsphere content and reference operating current reference value, calculate the first ratio of the reference glass microsphere content to the estimated value of fly ash glass microsphere content, and the second ratio of the current mechanical load parameter to the reference operating current reference value;

[0030] Based on preset weighting coefficients, the first ratio and the second ratio are weighted and summed to obtain the flow state early warning coefficient.

[0031] Based on the numerical range mapped by the flow state early warning coefficient, the built-in tags of the built-in control primitive library are retrieved, the matching processing node sequence is extracted, and the target operation primitive set is generated.

[0032] In a further embodiment of the present invention, the sequence encapsulation module is specifically used for:

[0033] Analyze the data input / output interface configuration of each independent processing node within the target operation primitive set, sort the nodes according to the control causal transmission logic, and establish a control logic chain.

[0034] Configure the deviation target setpoint and feedforward reference offset at the injection end of the control logic chain, and write the action dead zone boundary limit threshold at the output end of the tail to form batch control logic script code.

[0035] Extract the unique object identifier and the device factory verification code of the target controller, fuse them to generate a symmetric encryption derivative key, combine it with the obtained current clock random number factor, perform encryption operation on the batch control logic script code, generate a digital signature, and encapsulate it to obtain the batch instruction control sequence.

[0036] In a further embodiment of the present invention, the authorization verification module is specifically used for:

[0037] Write the batch instruction control sequence into the high-speed receive buffer hardware area of ​​the logic execution interpretation engine, and lock the current control output to pause logic switching;

[0038] Read the device factory verification code of the target controller, and use the device factory verification code to perform reverse comparison and verification of the digital signature carried by the batch instruction control sequence;

[0039] When the verification is successful, the memory addressing lock state of the protected area is released and an authorized running status signal is output; when the verification fails, a mechanical lock is triggered and an out-of-bounds alarm pulse is output.

[0040] In a further embodiment of the present invention, the topology reconfiguration module is specifically used for:

[0041] Extract the level state transition event of the authorized running status signal by listening to the pins, and clear the preset proportional-integral-differential parameter table in memory;

[0042] Write the batch instruction control sequence into the instruction temporary register, and scan the control node step-by-step execution pointer of the batch instruction control sequence line by line;

[0043] In the memory stack area, an independent local simulation storage block is allocated for each atomic-level algorithm subroutine. A data assignment link is established from the output of the preceding process algorithm subroutine to the input of the following process algorithm subroutine, generating an executable logic control topology.

[0044] In a further embodiment of the present invention, the instruction execution module is specifically used for:

[0045] The graded particle size error electrical signal and the ash flow concentration oscillation amplitude electrical signal of the physical entity are extracted according to a fixed scanning frequency and used as real-time deviation acquisition feedback values.

[0046] The graded granularity error electrical signal is converted into a digital quantity and injected into the data input node of the executable logic control topology.

[0047] Along the preset data transmission path, data smoothing filtering and proportional amplification deviation numerical simulation operations are performed sequentially to obtain the control solution value;

[0048] The calculated control values ​​are pushed to the digital-to-analog converter chip for electrically isolated output, and then converted into hierarchical equipment control commands.

[0049] A further aspect of this invention involves pushing the control calculation value to a digital-to-analog converter chip for electrically isolated output operation. When converted into hierarchical equipment control commands, this is specifically used for:

[0050] Convert the calculated control values ​​into analog control signals;

[0051] The analog control signal is used as the control command for the classifying equipment and output to the main air valve servo motor to drive the rotation step, or output to the classifier variable frequency traction motor to adjust the power supply frequency.

[0052] In summary, the present invention has the following beneficial technical effects:

[0053] 1. By collecting source process data from fly ash generation and conveying nodes, a unique physical batch object identifier is generated, and a multi-dimensional feature vector containing parameters such as vitreous microsphere content is decoded based on this identifier. This technical feature enables the system to perform feedforward quantitative assessment of key physical properties of materials before they enter the grading process. This mechanism transforms the implicit attributes of materials into digital features through data mapping, realizing data traceability and feature construction for each physical batch of fly ash. This provides a predictive decision-making basis for the generation of control strategies, thereby improving the control lag and adjustment deviation problems caused by unknown material characteristics in existing technologies.

[0054] 2. By importing multi-dimensional feature vectors into the control strategy compilation environment and combining them with the target operation primitive set matched to the safety boundary of process fluctuations, batch instruction control sequences are generated. This technical feature enables the system to dynamically combine and generate matching control algorithm logic based on the specific characteristics of different batches of fly ash. By selecting suitable algorithm components from the underlying control primitive library based on the material flowability warning coefficient, the control system has the ability to adaptively configure for specific batches of materials. This mechanism of adjusting the logic structure based on the characteristics of the controlled object helps improve the stability and robustness of the system when dealing with complex operating conditions, overcoming the limitations of traditional methods that rely solely on fixed models and single parameter adjustments.

[0055] 3. By sending batch instruction control sequences with attached digital signatures to the underlying controller, the controller performs a hardware-level identity verification process and dynamically reconstructs the executable logic control topology upon successful verification. This technical feature utilizes object identifiers and device factory serial numbers to establish an access control mechanism for instruction issuance, which helps reduce the risk of erroneous instruction loading or abnormal code injection, ensuring control security in industrial settings. Simultaneously, after successful verification, the data flow link is directly reconstructed within the microprocessor, enabling online switching of control logic. This improves the operational reliability and information protection capabilities of the control system while maintaining the continuity of hierarchical system operation. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The drawings are used to provide a further understanding of the present invention.

[0057] Figure 1 This is a schematic diagram of the framework in the embodiments of this application.

[0058] Figure 2 This is a flowchart illustrating an embodiment of this application.

[0059] Figure 3 This is a time-domain waveform diagram of the boiler combustion temperature time series in an embodiment of this application.

[0060] Figure 4 This is a schematic diagram of the frequency domain conversion of boiler combustion temperature and the frequency fluctuation of the main frequency in the embodiments of this application.

[0061] Figure 5 This is a schematic diagram of the current fluctuation and transient spark discharge curve of the electrostatic precipitator station in the embodiments of this application.

[0062] Figure 6 This is a schematic diagram of the curve representing the mapping law of the flow state early warning coefficient in the embodiments of this application.

[0063] Figure 7 This is a timing diagram of the authorized running status signal when the verification is successful in an embodiment of this application.

[0064] Figure 8 This is a timing diagram of the out-of-bounds alarm pulse when verification fails in an embodiment of this application.

[0065] Figure 9 This is a bar chart illustrating the evolution of the underlying topology in processing real-time deviation data in an embodiment of this application. Detailed Implementation

[0066] The following is in conjunction with the appendix Figures 1-9 A preferred description of the present invention is provided below.

[0067] See attached document Figures 1-9 This invention proposes a fly ash grading control system based on object identifier parsing, comprising the following modules:

[0068] The identifier construction module is used to obtain displacement signals of fly ash storage and conveying nodes, extract process condition snapshot data of associated equipment, and generate unique object identifiers.

[0069] The feature decoding module is used to parse unique object identifiers, extract historical operating parameters, and decode historical operating parameters through a physical data hybrid model to obtain multi-dimensional feature vectors.

[0070] The strategy matching module is used to obtain the current mechanical load parameters of the staged fan equipment. Based on the multi-dimensional feature vector and the current mechanical load parameters, it performs matching in the built-in control primitive library to extract the target operation primitive set.

[0071] The sequence encapsulation module is used to perform topological recombination on the target operation primitive set and, in combination with preset process parameters, generate a batch instruction control sequence containing a digital signature.

[0072] The authorization verification module is used to send the batch instruction control sequence to the logic execution interpretation engine, verify the digital signature, and generate an authorization running status signal;

[0073] The topology reconfiguration module is used to load batch instruction control sequences into the microcontroller unit based on authorized operating status signals, and construct an executable logic control topology.

[0074] The instruction execution module is used to acquire the real-time deviation acquisition feedback value of the environmental sensor probe, drive the executable logic control topology to calculate the real-time deviation acquisition feedback value, and generate hierarchical equipment control instructions.

[0075] In one embodiment of the present invention, the identifier construction module is configured to perform the following steps:

[0076] Monitor the mechanical opening and closing action of the discharge valve of the upstream ash storage silo, and simultaneously accumulate the flow pulse value output by the bottom conveyor belt scale. When the flow pulse value exceeds the preset batch tonnage threshold, generate a physical batch start cutoff event.

[0077] Obtain the timestamp parameter and hardware device code of the moment when the start truncation event occurs, extract key frames of fire images within the same time window based on the timestamp parameter, and construct process condition snapshot data;

[0078] The timestamp parameter, hardware device code, and process condition snapshot data are serialized and concatenated, and then fused using a hash function to generate a unique object identifier.

[0079] Specifically, the steps of collecting displacement signals of fly ash storage and conveying nodes, extracting associated equipment process states to form working condition snapshot data, and constructing a globally unique object identifier for a physical batch are executed by an identifier construction module integrated in the industrial control system. This module first continuously monitors the position sensor of the discharging valve of the upstream ash storage silo through hard wiring or an industrial field bus, and the position sensor outputs a binary level signal representing the fully open or fully closed state of the valve.

[0080] At the same time, the module synchronously collects the flow pulse signal output by the weighing sensor of the bottom conveying belt scale linked with the ash storage silo through a high-speed counter interface, where each pulse corresponds to a calibrated material mass. The accumulator logic unit inside the module continuously accumulates the flow pulse signals to form a real-time accumulated flow value.

[0081] When the module detects that the discharging valve completes one complete mechanical opening and closing action from open to closed, and the accumulated flow value exceeds the preset batch tonnage threshold for the first time, the system determines that the conveying of one complete physical batch is finished, and immediately generates an internal logic signal, which is defined as the start truncation event of the physical batch.

[0082] It should be noted that the preset batch tonnage threshold is an engineering parameter set according to the processing capacity of the fly ash conveying system and the target processing batch of the downstream classification equipment, and its value range is usually between 50 and 200 metric tons. This setting aims to ensure that the material quantity of each batch is relatively constant, which facilitates the standardization of control strategies. Assuming the application scenario is a classification system matched with a single large cement mill, whose single processing capacity is 100 metric tons, the threshold is set to 100 metric tons.

[0083] After receiving the start truncation event of the physical batch, the central processing unit immediately captures the millisecond-accurate timestamp parameter provided by the current operating system kernel, and retrieves the hardware device codes associated with the ash storage silo and the belt scale that trigger the event from the equipment asset management database. Among them, the timestamp parameter adopts the internationally standardized Coordinated Universal Time and is recorded with millisecond-level precision to ensure the uniqueness and traceability of the event; the hardware device code is a set of alphanumeric strings pre-allocated during system deployment and conforming to the enterprise asset coding specification. For example, the number of the ash storage silo can be set as SHC01, and the number of the belt scale can be set as BTW01.

[0084] Based on the timestamp parameter, the processor unit searches the pre-configured system association table for the corresponding network address of the boiler control system and the data tag of the combustion diagnosis node, and initiates a data query request. This request specifies to extract key frames of thermal power images within the same time window with a specific time width centered on the timestamp parameter.

[0085] The same time window refers to a time interval extending 1 to 5 seconds forward and backward from the timestamp parameter as the center point. The setting of this interval is based on the typical time constant of the boiler combustion condition change, to ensure that the captured image can effectively reflect the core combustion state when the batch of fly ash is generated.

[0086] The boiler control system responds to the request by returning one or more frames of specified two-dimensional matrix data from its historical database or image cache. The processor unit receives this data and encapsulates it to construct a structured process condition snapshot data. Specifically, the process condition snapshot data is a single frame of 8-bit grayscale image with a resolution of 640 pixels by 480 pixels, acquired from the boiler furnace flame monitoring camera, and its data structure is a two-dimensional integer array.

[0087] Finally, the system activates its built-in irreversible hash calculation core. This core executes a deterministic cryptographic hash function, merging the timestamp parameters, hardware device codes, and process snapshot data obtained in the preceding steps after unified serialization processing. The result is a fixed-length hexadecimal string with a high avalanche effect, which serves as the unique object identifier binding the source characteristics of the current batch. The irreversible hash calculation core is a software-implemented computation module that consistently uses a 256-bit version of the secure hash algorithm to ensure the uniqueness and tamper-proof nature of the identifier.

[0088] The process of generating a unique object identifier can be defined by the following formula:

[0089]

[0090] In the formula, The final unique object identifier is a fixed-length hash value string. A standard irreversible hash function, such as the 256-bit version of the Secure Hash Algorithm, takes an input of byte sequence of arbitrary length and outputs a fixed-length digest. This represents a serialization function, which converts inputs of different data types, such as numbers, strings, and matrices, into a deterministic and unambiguous sequence of bytes to ensure the consistency of hash operations. The timestamp parameter records the moment when the physical batch's initial truncation event occurred. A collection of hardware device codes that uniquely identifies the physical device from which the material originated. This represents snapshot data of the process conditions, typically a two-dimensional or three-dimensional matrix containing pixel intensity information. The symbol represents the concatenation operation of byte sequences, that is, the concatenation of the serialized byte streams of each variable in a predetermined order to form a single input data stream.

[0091] In one specific embodiment, the generation process of a physical batch is as follows: a preset batch tonnage threshold is set to 50 metric tons. At a certain moment, the identifier construction module detects that the discharge valve of the ash storage silo numbered SHC02 has completed one opening and closing action, and the cumulative conveying volume of the conveyor belt scale numbered BTW02 associated with it has reached 50.1 metric tons, thereby triggering the start-stop event of the physical batch. The system captures the timestamp parameter at this time. The corresponding hardware device code is 1672531200000. The string is SHC02&BTW02. The system immediately uses the timestamp parameter and sets the time window to 2 seconds before and after, requesting data from the boiler control system. The boiler system returns a frame of grayscale image simplified to 3 pixels by 3 pixels as a snapshot of the process conditions. Its pixel value matrix is ​​[[180,195,210], [185,220,200], [190,215,198]]. Next, the irreversible hash calculation core performs the operation. First, the serialization function... The input data is converted into a uniform string. For example, all input data is concatenated using "|" as separators to form the input string: "1672531200000|SHC02&BTW02|180,195,210;185,220,200;190,215,198". This string is then input as a whole into the 256-bit version of the secure hash algorithm function. After calculation, a hexadecimal string of length 64 characters is generated, such as "e3b0...b855". This string is the final generated unique object identifier that binds all source characteristics of the 50.1 metric tons of fly ash physical batch.

[0092] In one embodiment of the present invention, the feature decoding module is configured to perform the following steps:

[0093] Use the unique object identifier as the query index, and extract the bound timestamp parameter and hardware device code as historical running parameters;

[0094] Based on the timestamp parameter, the boiler combustion temperature time series and the current fluctuation curve of the electrostatic precipitator station are extracted from the associated production cycle.

[0095] A frequency domain transformation operation is performed on the boiler combustion temperature time series to extract the combustion main frequency fluctuation amplitude parameter, and the spark discharge variance index of the current fluctuation curve is calculated and combined to form an intermediate feature vector.

[0096] The intermediate feature vector is standardized and then input into the physical data hybrid model to infer the estimated content of fly ash glass microspheres. The combustion frequency fluctuation amplitude parameter, spark discharge variance index and the estimated content of fly ash glass microspheres are then concatenated to generate a multidimensional feature vector.

[0097] Specifically, the steps of parsing the unique object identifier, extracting hidden historical operating parameters, and retrieving the physical data hybrid model to decode and generate the corresponding batch's multi-dimensional feature vector are executed by the feature decoding module deployed on the hierarchical control system. This module first receives the unique object identifier transmitted across network segments by the central computing cluster via the industrial Ethernet interface.

[0098] Upon receiving the data, the module does not directly perform mathematical decomposition on the hash string of the unique object identifier itself. Instead, it uses it as a unique query index to match the temporary data mapping table that was created synchronously when the identifier was generated.

[0099] Through this matching operation, the system probes and extracts the original historical operating parameters bound to the identifier, which serve as its generation source; these parameters are the timestamp parameter and the hardware device code. After obtaining these original parameters, the feature decoding module uses the timestamp parameter as a time reference to initiate two parallel data extraction requests to the factory-level real-time database.

[0100] It should be noted that a real-time database is a high-performance time-series database used for storing and retrieving data in industrial processes, such as the PI system or InfluxDB, which can record and query historical data with microsecond-level timestamp accuracy.

[0101] The first request is to extract the combustion temperature time series of a specified boiler within a specific production cycle associated with the timestamp parameter;

[0102] The second request is to synchronously extract the secondary current fluctuation curve of the corresponding electrostatic precipitator station during this time period.

[0103] Among them, the boiler combustion temperature time series is obtained from thermocouples installed in a designated area inside the boiler, which are continuously sampled at a frequency of 1 Hz to record the temperature changes in the furnace; at the same time, the current fluctuation curve of the electrostatic precipitator station is obtained from the current sensor on the secondary side of the high-voltage rectifier transformer of the electrostatic precipitator, which is sampled at a frequency of 10 Hz or higher to capture transient spark discharge events.

[0104] After data extraction is complete, the system starts the pre-feature engineering processing unit, which contains two parallel processing threads:

[0105] like Figure 3 and Figure 4As shown, a fast Fourier transform operation is performed on the obtained boiler combustion temperature time series to transform it from the time domain to the frequency domain. The non-DC component with the highest amplitude is located from the generated spectrum and its amplitude is extracted as the key combustion main frequency fluctuation amplitude parameter.

[0106] Parallel calculation of the statistical variance of the current fluctuation curve over the entire time series, combined with Figure 5 It can be seen that the curve contains transient spark discharge events with abrupt changes, and this variance value is defined as the spark discharge variance index.

[0107] These two calculated values—the combustion frequency fluctuation amplitude parameter and the spark discharge variance index—are combined to form a two-dimensional intermediate feature vector. The combustion frequency fluctuation amplitude parameter quantifies the stability of the combustion process; a larger value indicates more intense combustion fluctuations, with a typical range between 0.01 and 0.5 Hz. Furthermore, the spark discharge variance index is strongly correlated with the resistivity of fly ash; excessively high or low resistivity leads to changes in spark discharge frequency and intensity, thus affecting the variance value. Its unit is the square of amperes.

[0108] The system invokes a preset feature scaler to perform standardization on the intermediate feature vectors, eliminating numerical scale differences caused by different physical units. Finally, the system imports a pre-trained offline physical data hybrid model that includes Gaussian process regression computational power.

[0109] The physical data hybrid model refers to a hybrid inference model constructed by pre-extracting the physical characteristics of the underlying mechanisms and mapping them into a data-driven regression algorithm. It is pre-trained offline using a large amount of historical production data, including combustion and electrostatic precipitator parameters for each batch, and the corresponding actual content of fly ash vitreous microspheres measured through laboratory physicochemical analysis. The model uses the standardized intermediate feature vector as the mapping independent variable in its multidimensional input space. Through inference calculations using the kernel function and covariance matrix within the model, a scalar output is obtained. This output is an estimate of the fly ash vitreous microsphere content, which is physically difficult to measure directly online.

[0110] The estimated content of vitreous microspheres in fly ash is a percentage value used to characterize the proportion of spherical smooth particles in fly ash. This value is a key intrinsic attribute affecting the flowability and subsequent utilization value of fly ash, and its range is usually between 30% and 80%.

[0111] Finally, the characteristic decoding module sequentially splices the combustion frequency fluctuation amplitude parameter, spark discharge variance index, and estimated fly ash vitreous microsphere content in a predetermined order to form a multidimensional feature vector, which serves as a quantitative representation of the intrinsic physical characteristics of the current fly ash batch. It should be noted that the multidimensional feature vector is a fixed-dimensional numerical vector containing all the aforementioned key features, and its structure is defined as: [combustion frequency fluctuation amplitude parameter, spark discharge variance index, estimated fly ash vitreous microsphere content].

[0112] The key calculations involved in this step include the calculation of the spark discharge variance index and the inference of the Gaussian process regression model.

[0113] Spark discharge variance index The calculation formula is:

[0114]

[0115] In the formula, This represents the total number of data points in the current fluctuation curve. Represents the first in the time series Each current sample value is in amperes. This represents the average current over the entire time series, and its unit is amperes. This represents the final calculated variance index, and its unit is the square of amperes.

[0116] Gaussian process regression model for new inputs Predicted output The posterior distribution is:

[0117]

[0118] Among them, the predicted mean and prediction variance It is given by the following formula:

[0119]

[0120]

[0121] In the formula, This represents the normal distribution function. This represents the model's predicted output for a new input. It represents the intermediate feature vector, which is composed of the combustion frequency fluctuation amplitude parameter and the spark discharge variance index and has been standardized. and These represent the historical input feature matrix and the corresponding output used for offline training of the model, respectively, which is the experimentally measured vector of vitreous microsphere content. This represents the covariance kernel function, such as the commonly used radial basis function kernel. It is a hyperparameter representing the variance of the measurement noise. It is an identity matrix. It is the predicted mean of the estimated values ​​of fly ash vitreous microsphere content output by the model, which is the final estimated value used. This refers to the uncertainty of the prediction, i.e., the prediction variance.

[0122] Furthermore, the offline training process of the physical data hybrid model specifically includes:

[0123] Acquire a large amount of historical production batch data, and extract the standardized intermediate feature vectors of each batch as the input feature array for the training set. The actual content of fly ash vitreous microspheres in the corresponding batches, measured by laboratory physicochemical analysis, was obtained as the training set label vector. ;

[0124] The negative log-marginal likelihood function of the Gaussian process regression model is constructed as the loss function;

[0125] Using iterative optimization algorithms, such as the L-BFGS-B algorithm or the Adam optimizer, the hyperparameters in the model, including the covariance kernel function, are iteratively updated with the goal of minimizing the loss function. The characteristic length scale parameter and measurement noise variance in ;

[0126] When the loss function value reaches the convergence condition or the preset maximum number of iterations, training stops, the converged hyperparameters are output and fixed into the model, thus obtaining the trained physical data hybrid model.

[0127] In one specific embodiment, the hierarchical control system receives a unique object identifier "e3b0...b855". The system queries a data mapping table and parses the associated timestamp parameter as 1672531200000. The feature decoding module then uses this timestamp as a reference, looking back 30 minutes, and requests the boiler combustion temperature time series and the current fluctuation curve of the electrostatic precipitator station from the real-time database for this time period. Assume that the acquired temperature series has 1800 data points and the current series has 18000 data points. The pre-processing feature engineering unit initiates parallel computation. In the first thread, a Fast Fourier Transform is performed on the 1800 temperature data points. The analysis results show a significant peak at 0.08 Hz, with a corresponding amplitude of 3.5 degrees Celsius. Therefore, the combustion frequency fluctuation amplitude parameter is determined to be 3.5. In the second thread, the statistical variance of the 18000 current data points is calculated. The result is, for example, 0.045 amperes squared, which is the spark discharge variance index. At this point, the resulting intermediate feature vector is [3.5, 0.045]. Next, the system calls a feature scaler to standardize it, for example, converting it into a dimensionless vector [0.85, -0.21]. The system then uses this standardized two-dimensional vector [0.85, -0.21] as input to a pre-trained physical data mixture model. The model performs inference calculations based on its internal Gaussian process regression logic, outputting the predicted mean. The value is 62.5. This value is the estimated content of vitreous microspheres in fly ash, expressed as a percentage. Finally, to preserve the physical interpretability of the data, the feature decoding module extracts the unstandardized raw feature parameters and concatenates them with the estimated value to generate a three-dimensional multidimensional feature vector with specific values ​​of [3.5, 0.045, 62.5]. This vector will be passed to subsequent steps to generate a refined control strategy.

[0128] In one embodiment of the present invention, the strategy matching module is configured to perform the following steps:

[0129] The estimated content of vitreous microspheres in fly ash was extracted from the multidimensional eigenvectors.

[0130] Obtain the preset reference glass microsphere content and reference operating current reference value, calculate the first ratio of the reference glass microsphere content to the estimated value of fly ash glass microsphere content, and the second ratio of the current mechanical load parameter to the reference operating current reference value;

[0131] Based on preset weighting coefficients, the first ratio and the second ratio are weighted and summed to obtain the flow state early warning coefficient.

[0132] Based on the numerical range mapped by the flow state early warning coefficient, the built-in tags of the built-in control primitive library are retrieved, the matching processing node sequence is extracted, and the target operation primitive set is generated.

[0133] Specifically, the steps of importing the multidimensional feature vector into the control strategy compilation environment, evaluating the safe extreme boundary of fly ash material process fluctuations, and matching and extracting the target operation primitive set from the built-in control primitive library are executed by the strategy matching module. This module first parses and extracts the third dimension value from the multidimensional feature vector, namely the estimated value of fly ash vitreous microsphere content.

[0134] Meanwhile, this module queries and obtains the current mechanical load parameters of the grading fan equipment, which are strongly related to the grading process, in real time through the data bus interface. It should be noted that the current mechanical load parameters of the grading fan equipment specifically refer to the three-phase average operating current of the AC asynchronous motor of the main fan driving the airflow circulation within the grading system. This parameter is measured in real time by a current transformer installed in the motor control cabinet, and its value directly reflects the aerodynamic resistance of the gas-solid two-phase flow within the pipeline, serving as a key indirect indicator for assessing material flowability.

[0135] Based on the estimated content of vitreous microspheres in fly ash and the current mechanical load parameters, the system calls a preset risk assessment function to calculate a comprehensive flow state early warning coefficient. This coefficient is used to quantify the potential risk of flow deterioration or blockage of the current batch of fly ash within the graded pipeline system.

[0136] The flow condition warning coefficient is a normalized risk index with a baseline value of 1.0, representing that both material characteristics and equipment load are at their optimal reference state. When the coefficient value is greater than 1.0, it indicates a deteriorating flow trend or potential blockage risk; the higher the value, the higher the risk. Figure 6 The warning coefficient shown is influenced by the mapping curve between the front-end prediction attribute and the real-time load. For example, when... Within the interval [0.8, 1.1], it is mapped to a low-risk interval; within (1.1, 1.4], it is mapped to a medium-risk interval; and when it is greater than 1.4, it is mapped to a high-risk interval.

[0137] After the calculation is complete, the system activates its underlying load condition control strategy compiler, which is the core software decision-making module. This compiler is a rule-based software module whose core function is to map continuous risk coefficient values ​​to discrete sets of control primitives. The compiler then loads a structured, low-level atomic control primitive library from non-volatile memory into the active memory area. This low-level atomic control primitive library predefines and encapsulates various modular control algorithm components, including but not limited to independent filtering extraction operators, integral adjustment functions with anti-integral saturation capabilities, and output rate limiting modules that can smooth the output.

[0138] The underlying atomic-level control primitive library is a software library stored in key-value pair format. The key is a unique identifier for the primitive, such as `Filter_Smooth_V1`, and the value is a data structure containing an execution code pointer, resource requirements, and multiple descriptive built-in tags. Finally, the operating condition control strategy compiler maps the numerical range of the flow state warning coefficient calculated in the previous step to one or more search tags. It should be noted that the built-in tags are pre-defined classification identifiers for each control primitive. For example, the basic tag represents the core primitives that all strategies must include, the flow improvement tag is associated with regulatory primitives applicable to medium-risk ranges, and the anti-blocking tag is associated with strongly interventionist primitives applicable to high-risk ranges.

[0139] Using these tags, the compiler performs a fast search and match in the underlying atomic-level control primitive library, selecting all built-in tags that match the current search tag and the sequence of matching processing nodes. These selected processing nodes are combined to form a set of target operation primitives whose internal execution timing is not yet determined, and this is used as the final output of this step.

[0140] Flow state early warning coefficient The calculation is defined by the following formula:

[0141]

[0142] in, The final calculated flow state warning coefficient is a dimensionless scalar value used to measure the degree to which the system deviates from the optimal flow state. The value is an estimated value of the vitreous microsphere content of fly ash extracted from the multidimensional eigenvector, expressed as a percentage. Due to the physical formation mechanism of fly ash, its vitreous microsphere content has a natural physical lower bound and is always greater than zero. Therefore, in actual industrial applications, there is no computational singularity anomaly with a denominator of zero. It is a baseline vitreous microsphere content calibrated based on historical data, and the unit is also a percentage, for example, set to 70%. The current mechanical load parameters of the staged fan equipment are collected in real time, specifically the operating current of the fan drive motor, in amperes. It is the reference value of the operating current of the blower when processing standard flowable materials, and the unit is amperes. For example, it is set to 140 amperes based on a large number of experimental statistics. and These are the weighting coefficients for the deviation of vitreous microsphere content and the deviation of mechanical load, respectively. Both are dimensionless constants and satisfy the following conditions: These two weighting coefficients are preferably determined through expert experience or offline optimization based on the process characteristics of a specific grading system. For example, they can be set... , .

[0143] In one specific embodiment, the strategy matching module receives a multidimensional feature vector with specific values ​​[3.5, 0.045, 62.5]. The module first parses out the estimated content of fly ash vitreous microspheres. The figure is 62.5%. Simultaneously, the system queries the current mechanical load parameters of the staged fan equipment via a real-time data interface. The current is 142 amperes. The system retrieves a preset reference value, in which the content of the reference vitreous microspheres is... 70%, reference operating current value 140 amperes, weighting factor and The values ​​are set to 0.6 and 0.4 respectively. The flow state early warning coefficient is calculated based on the risk assessment function. The calculated coefficient value of 1.0777 falls within the predefined low-risk range [0.8, 1.1]. At this point, the operating condition control strategy compiler maps this coefficient to the search label "low-risk." The compiler then uses the low-risk label to search the underlying atomic-level control primitive library. Assume there are two primitives in the library labeled "low-risk": a proportional-integral-derivative control function named PID_Standard and an input signal filtering operator named Filter_Input_Light. Therefore, the search operation ultimately returns these two control primitives. These two primitives together constitute the target operation primitive set and are passed to the next step for timing and topology reorganization.

[0144] In one embodiment of the present invention, the sequence encapsulation module is configured to perform the following steps:

[0145] Analyze the data input / output interface configuration of each independent processing node within the target operation primitive set, sort the nodes according to the control causal transmission logic, and establish a control logic chain.

[0146] Configure the deviation target setpoint and feedforward reference offset at the injection end of the control logic chain, and write the action dead zone boundary limit threshold at the output end of the tail to form batch control logic script code.

[0147] Extract the unique object identifier and the device factory verification code of the target controller, fuse them to generate a symmetric encryption derivative key, combine it with the obtained current clock random number factor, perform encryption operation on the batch control logic script code, generate a digital signature, and encapsulate it to obtain the batch instruction control sequence.

[0148] Specifically, the steps of reorganizing the spatial topology and temporal link dependencies of the target operation primitive set, attaching basic closed-loop operation boundaries, and packaging the computation to generate a batch instruction control sequence containing a verification key are executed by the sequence encapsulation module. This module first scans the target operation primitive set output from the previous step, identifies and parses the data interface definition of each independent processing node, i.e., its data input / output interface configuration. Here, data input / output interface configuration refers to explicitly specifying the source type of data input and the target type of data output when defining each atomic-level control primitive. For example, the deviation input interface of the PID_Standard primitive must be connected to the output of the subtractor, and its control output interface must be connected to the actuator driver module.

[0149] Based on preset control causal transmission logic, such as stipulating that filtering operations must precede deviation calculations and integral accumulations must follow proportional calculations, the module performs topological sorting on these nodes to establish a loop-free action execution control logic chain. It should be noted that the control causal transmission logic is a set of rules pre-defined in the encapsulation module, which clarifies the sequence between different types of control operations, ensuring that the generated control logic chain conforms to the basic principles of control theory and process safety requirements. Simultaneously, the control logic chain is a logically directed acyclic graph, where nodes represent target operation primitives, and directed edges represent the execution order and data flow between nodes.

[0150] Subsequently, the module locates the starting point of the data flow in the control logic chain, namely the head injection end, and the final output point of the data flow, namely the tail output end. At the head injection end, the system reads and configures the deviation target setpoint (set to zero) from the preset process parameter library, and at the same time extracts the classifier speed reference value corresponding to the current production task as the control feedforward reference bias.

[0151] The feedforward reference bias is a pre-set engineering control baseline based on the required fly ash product grade. For example, when producing grade 1 fly ash, this baseline corresponds to the basic drive signal of the classifier's frequency converter, such as a 12 mA reference output current. Simultaneously, at the tail output, the module is programmed with an action dead zone boundary limit threshold. This threshold defines the range of changes in the controller output that will be ignored, preventing unnecessary frequent adjustments to minor disturbances.

[0152] The dead zone limit threshold is a parameter used to increase the stability of the controller. Its typical value is usually set to 0.5% to 1.0% of the controller output range. For example, if the controller output range is 0 to 100%, the threshold can be set to 0.5%.

[0153] These configurations, combined with ordered chains of primitives, form a complete, interpretable batch control logic script. The batch control logic script is an intermediate, human-readable or machine-parseable text or structured data format, such as using Extensible Markup Language or a JavaScript-like object notation format to describe the control logic.

[0154] Finally, the module performs the encryption encapsulation operation. It extracts the unique object identifier generated in the first step and retrieves the target controller's factory verification code from the device asset database, then combines and hashes the two to generate a symmetric encryption derivative key.

[0155] It should be noted that the symmetric encryption derivative key combines proprietary information from both communicating parties, and its usage ensures the security of targeted instruction delivery.

[0156] Specifically, the system employs a symmetric encryption scheme. Only by simultaneously knowing the dynamically generated unique object identifier and the device's factory verification code can the derived key be successfully constructed for encryption, decryption, and signature operations.

[0157] Simultaneously, to increase randomness, the module calls the operating system interface to obtain the random number factor of the current clock, for example, taking the nanosecond portion of the current time. The module combines this symmetric encryption derived key with the random number factor to perform encryption operations on the aforementioned batch control logic script code. This process not only encrypts the script content but also appends a unique object identifier to the plaintext header of the sequence and embeds a digital signature generated based on the aforementioned derived key at the tail. The digital signature is a small piece of data generated using the unique object identifier and the script content itself through a hash function and encryption algorithm, appended to the tail of the sequence for the receiving end to verify the integrity and authenticity of the sequence's origin.

[0158] The entire encrypted and signed data packet is encapsulated into a binary-formatted batch instruction control sequence embedded with a digital signature. This sequence, the final product of this step, awaits distribution to the physical device. It can be understood that the batch instruction control sequence is the final generated binary data stream that can be directly parsed by the underlying logic execution interpretation engine.

[0159] In one specific embodiment, the sequence encapsulation module receives a set of target operation primitives containing two primitives: PID_Standard and Filter_Input_Light. After scanning, the module identifies that the input interface of Filter_Input_Light receives the raw sensor signal, and its output interface is connected to the logic operation unit; the deviation input interface of PID_Standard must receive the difference between the setpoint and the measured value, and its output interface is connected to the actuator. Based on the control causal transfer logic, the module determines that the filtering operation should precede the proportional-integral-derivative (PID) operation. Therefore, the established control logic chain is: sensor signal → Filter_Input_Light → difference calculation → PID_Standard → actuator. The module locates the head injection end, i.e., the target setpoint required for difference calculation, and reads the configuration from the process library, writing the deviation target setpoint as 0, while configuring the feedforward reference bias as 12. At the tail output end, i.e., after the output of PID_Standard, the module writes the action dead zone boundary limit threshold as 0.5%. This forms a batch control logic script code. Finally, the module extracts the unique object identifier "e3b0...b855" and obtains the target controller's factory verification code. These two are then merged to generate a symmetric encryption derived key, and the nanosecond portion of the current operating system clock is used as a random number factor. The system calls its internal serialization engine to compress the aforementioned text-formatted script code into a compact binary machine code stream. A symmetric encryption algorithm is then used to ciphertext this binary machine code stream byte-by-byte. A unique object identifier is appended to the plaintext header of the data packet, and a 128-bit digital signature generated by hashing the unencrypted machine code stream and the derived key is appended to the end of the data packet. The entire binary data packet constitutes a batch instruction control sequence, ready to be sent to the underlying controller.

[0160] In one embodiment of the present invention, the authorization verification module is configured to perform the following steps:

[0161] Write the batch instruction control sequence into the high-speed receive buffer hardware area of ​​the logic execution interpretation engine, and lock the current control output to pause logic switching;

[0162] Read the device factory verification code of the target controller, and use the device factory verification code to perform reverse comparison and verification of the digital signature carried by the batch instruction control sequence;

[0163] When the verification is successful, the memory addressing lock state of the protected area is released and an authorized running status signal is output; when the verification fails, a mechanical lock is triggered and an out-of-bounds alarm pulse is output.

[0164] Specifically, the steps of issuing batch instruction control sequences to the logic execution interpretation engine of the physical device, triggering the digital certificate comparison process of the underlying control chip, and calculating and generating the authorized running status signal of the driver controller level are executed by the authorization verification module that relies on the communication gateway and the underlying controller to work together.

[0165] First, the central computing cluster establishes a secure point-to-point communication session with the solid-state logic controller of the target physical device through an industrial-isolated transmission channel equipped with electrical isolation and network firewall capabilities. It should be noted that an industrial-isolated transmission channel typically refers to physical layer isolation implemented using technologies such as optocouplers or isolation transformers, combined with application-layer firewall rules that only allow communication via specific protocols and ports, to prevent security threats from the upper-layer network from directly impacting the lower-level control equipment.

[0166] After establishing a session, the system takes the batch instruction control sequence generated in the previous step, encapsulates it, and pushes and writes it byte by byte into the high-speed receive buffer hardware area with a fixed physical address inside the solid-state logic controller. This high-speed receive buffer hardware area is a static random access memory region with an independent direct memory access controller, enabling high-speed data reception without consuming main processor resources.

[0167] Once the data reception is complete and the checksum is correct, the logic execution interpretation engine built into the solid-state logic controller is activated. It is understood that the logic execution interpretation engine is a miniature interpreter embedded in a field-programmable gate array (FPGA) or application-specific integrated circuit (ASIC). It does not run a general-purpose operating system but directly parses and executes batch instruction control sequences.

[0168] Once activated, the engine launches its built-in hardware decryption core module, which implements a specific decryption algorithm on the physical circuitry. This hardware decryption core module is a set of logic gates programmed into the chip using a hardware description language. Its decryption and verification operations are significantly faster than those implemented in software and cannot be modified by external malware.

[0169] Before the decryption and verification process begins, the controller enforces a security pre-operation, temporarily blocking any internal processes that could lead to a transfer of feedback control from the existing logic to the new logic, ensuring that the device maintains its current operating state until verification is complete. Next, the hardware decryption core module extracts a unique object identifier from the plaintext header of the received data block and reads the device factory verification code stored in the controller's communication motherboard's read-only memory via a dedicated bus. This device factory verification code is a long string bound to the device serial number, written during the manufacturing stage via a fuse or one-time programmable memory, serving as the underlying proof of the device's identity.

[0170] The module performs the same fusion hash operation on both, reconstructing a symmetric encryption derivative key. This derivative key is then used to reverse-decrypt the batch instruction control sequence and compare it with the digital signature carried at the end. The verification process involves performing the same hash operation on the received instruction sequence content itself and comparing it with the decrypted digital signature. If the verification result does not match, it is determined to be a verification failure instruction. The controller will immediately trigger the hardware circuit connected to the normally closed pin of the safety relay to mechanically lock the critical actuator and simultaneously send an out-of-bounds alarm pulse to the monitoring system through the digital output port. It should be noted that the safety relay is a special relay with a forced-guided contact mechanism, ensuring the safety circuit is broken even in the event of contact fusion; it is a key component in functional safety design.

[0171] Conversely, if the verification result matches perfectly, it is considered a successful verification instruction. The controller will then execute a specific internal operation, namely, sending an instruction to the memory management unit to release the memory addressing lock state on the protected area. It can be understood that the memory addressing lock state on the protected area is a security mode of the underlying processor. In this mode, write operations to a specific memory address range will be directly rejected by the hardware to prevent unauthorized programs from loading.

[0172] This protected area is used to store and execute new control topology logic. Simultaneously, specific state registers of the controller are set, resulting in a stable high-level signal output on the physical pin. This signal is the authorized operating status signal, allowing the reconfiguration of the control topology, indicating that the underlying hardware is ready to receive and execute new control logic. The authorized operating status signal is a binary hardware status signal, and its authorized state serves as a hardware-level enable condition for subsequent instruction loading and execution.

[0173] Reverse verification of digital signatures refers to the recipient using an agreed algorithm and key (in this case, the device's factory verification code) to recalculate the message digest, and then comparing it with the digest appended by the sender. If they match, it proves that the message has not been tampered with and its source is trustworthy.

[0174] In one specific embodiment, assume that a central computing cluster establishes communication with a Siemens S7 series controller via a fiber optic industrial Ethernet network. The cluster sends a packaged batch instruction control sequence, a 2048-byte binary data block, to the controller's designated receive buffer. After data transmission, the controller's internal logic execution interpretation engine is activated and immediately locks control, maintaining the classifier inverter at its previous frequency of 75 Hz. The engine's built-in hardware decryption core module starts, extracts a unique object identifier from the data block header, and reads a 128-bit device factory verification code stored on the target controller, such as A5B5C5D5E5F5AABBCCDDEEFF00112233. The module merges these two to reconstruct a symmetric encryption derivative key. The module then locates the end of the received 2048-byte data block and extracts the digital signature. Next, the module uses the reconstructed symmetric encryption derivative key to reverse-decrypt and hash all ciphertext content in the batch instruction control sequence except for the header identifier and the tail signature. Assume that after the operation, the module internally generates a checksum DEADBEEF… Then, it compares this internally generated digest bit by bit with the digital signature extracted and decrypted from the data block. In this example, the comparison results are identical. Therefore, the system determines that the verification instruction is successful. The controller's internal logic immediately sends an unlock signal to the memory management unit, granting write access to the program memory region from address 0x8000 to 0xFFFF. Simultaneously, the level of the Q0.0 output pin on the controller toggles from 0 volts to 24 volts; its timing status is detailed in [link to relevant documentation]. Figure 7 This stable 24-volt DC voltage signal is the authorized operating status signal, which will be connected to the interrupt input pin of the microcontroller to trigger subsequent program loading. If the comparison is inconsistent, for example, if the data is tampered with during transmission, the safety relay will activate, cutting off the main circuit power supply to the classifier inverter. Simultaneously, the Q0.1 output pin will generate a 10Hz pulse train; its timing is detailed in [link to timing information]. Figure 8 This triggers the local alarm.

[0175] In one embodiment of the present invention, the topology reconfiguration module is configured to perform the following steps:

[0176] Extract the level state transition event of the authorized running status signal by listening to the pins, and clear the preset proportional-integral-differential parameter table in memory;

[0177] Write the batch instruction control sequence into the instruction temporary register, and scan the control node step-by-step execution pointer of the batch instruction control sequence line by line;

[0178] In the memory stack area, an independent local simulation storage block is allocated for each atomic-level algorithm subroutine. A data assignment link is established from the output of the preceding process algorithm subroutine to the input of the following process algorithm subroutine, generating an executable logic control topology.

[0179] Specifically, the steps of extracting the authorized running status signal through pin monitoring, guiding the loading of the batch instruction control sequence into the microcontroller unit, and dynamically reconstructing the underlying executable logic control topology using the internal instruction stack are executed by the topology reassembly module deployed in the solid-state logic controller microcontroller unit. The microcontroller unit's interrupt input pin continuously monitors the level of the authorized running status signal. When a low-to-high transition event is detected in the signal, a preset interrupt service routine is automatically triggered by the hardware.

[0180] It's important to note that interrupt service routines are high-priority code that are executed by the processor in response to external hardware signals, capable of interrupting currently running regular programs. The interrupt service routine first performs a crucial initialization operation: it sends an instruction to the volatile memory area inside the control chip to clear the memory area used to store the preset proportional-integral-derivative parameter table. This operation ensures that old, no longer applicable control parameters are cleared.

[0181] The aforementioned pre-set proportional-integral-derivative (PI / DE) parameter table in memory, also known as the default PID PI / DE gain address allocation table, is a set of parameters preset in the controller firmware. It provides a general, conservative set of control performance when no specific instructions are available. Clearing this table is to make room for new, customized control logic and prevent conflicts between old and new parameters.

[0182] Next, the interrupt service routine uses a direct memory access channel to completely copy and write the batch instruction control sequence previously stored in the high-speed receive buffer hardware area into a set of dedicated instruction temporary registers in the microcontroller unit. These instruction temporary registers are a set of high-speed memory units near the microcontroller core, used to temporarily store instruction codes to be processed before execution.

[0183] The microcontroller's main program initiates the interpreter engine execution thread, which resides in memory. This interpreter engine execution thread is a continuously running software process whose responsibility is not to compile code, but rather to read instructions line by line, understand their meaning, and immediately invoke the corresponding underlying hardware functions for execution. Starting from the first memory address in the instruction temporary storage register, this thread scans and parses the control node step pointers contained in the batch instruction control sequence line by line. These control node step pointers are data contained within the batch instruction control sequence, specifying the execution order of various control modules, such as filtering, summation, and proportional-integral-differential operations.

[0184] For each control node parsed, the interpreter engine dynamically maps and allocates appropriately sized local simulated storage blocks in the microprocessor's memory stack area for the atomic-level algorithm subroutine represented by that node. These blocks store intermediate variables and states during the computation process. It's important to note that these local simulated storage blocks are private memory spaces temporarily allocated in dynamic random access memory for each algorithm function block. Their function is similar to function scope in high-level languages, used to isolate variables from different function blocks and prevent mutual interference.

[0185] Finally, the interpreter engine establishes direct data paths between different subroutine memory blocks based on the connection relationships defined in the batch instruction control sequence. Specifically, it hardcodes the memory address of the output of the preceding process algorithm subroutine as the forced assignment source address of the input of the following process algorithm subroutine.

[0186] In this way, the unidirectional transmission path of data bytes between different algorithm modules is solidified at the hardware level, thereby dynamically and non-compiling a new executable logic control topology that matches the characteristics of the current batch of fly ash within the microprocessor.

[0187] It is understandable that the forced assignment link corresponding to the above unidirectional transmission transfer path is implemented by modifying the memory address pointer, so that any data update in the memory location will be automatically and in real time copied to another predetermined memory location, thereby simulating the data flow effect of the hardware connection; and the executable logic control topology is a logical structure built in memory through pointers and data links, which completely reproduces the control logic chain designed in the fourth step, but in the form of memory objects and data flow paths that can be directly executed by the microprocessor.

[0188] In one specific embodiment, the INT1 pin of the microcontroller detects a jump in the authorized running status signal connected to it from 0 volts to 24 volts. This jump triggers the corresponding interrupt service routine. The routine executes immediately, first issuing a clear instruction to the memory address range 0xF000 to 0xF0FF. This range originally stored a parameter table with a default proportional gain of 2.0, an integral time of 10 seconds, and a derivative time of 0.5 seconds; after execution, this range is completely overwritten with zeros. Next, the program activates the direct memory access controller, copying the 2048-byte batch instruction control sequence stored in the high-speed receive buffer hardware area at addresses 0x1000 to 0x17FF intact to the instruction temporary register starting at address 0x2000. Subsequently, the resident interpreter engine running thread is awakened and begins scanning the data at address 0x2000. The first instruction it reads is `Filter_Input_Light`, so it allocates 32 bytes as a local analog storage block for this filtering function in the memory stack area, for example, starting from address 0xA000. The second instruction is `Subtract`, which allocates 16 bytes at address 0xA020. The third instruction is `PID_Standard`, which allocates 64 bytes at address 0xA030. After parsing all nodes, the interpreter begins linking. Based on the linking information in the instruction sequence, it sets the pointer at memory address 0xA01C, representing the output of `Filter_Input_Light`, as the source address of the second input parameter of the `Subtract` function. Similarly, it sets the pointer at memory address 0xA02C, representing the output of the `Subtract` function, as the source address of the input parameter of the `PID_Standard` function. By establishing this series of forced assignment links, a new executable logic control topology containing filtering, summation, and proportional-integral-derivative control logic is built in memory and ready to receive real-time data for computation.

[0189] In one embodiment of the present invention, the instruction execution module is configured to perform the following steps:

[0190] The graded particle size error electrical signal and the ash flow concentration oscillation amplitude electrical signal of the physical entity are extracted according to a fixed scanning frequency and used as real-time deviation acquisition feedback values.

[0191] The graded granularity error electrical signal is converted into a digital quantity and injected into the data input node of the executable logic control topology.

[0192] Along the preset data transmission path, data smoothing filtering and proportional amplification deviation numerical simulation operations are performed sequentially to obtain the control solution value;

[0193] The calculated control values ​​are pushed to the digital-to-analog converter chip for electrically isolated output, and then converted into hierarchical equipment control commands.

[0194] Specifically, the steps of reading the real-time deviation acquisition feedback value of the environmental sensor probe in the grading process, driving the executable logic control topology to perform step-by-step interpretation and calculation actions, and generating grading equipment control instructions to be output to the physical equipment adjustment end are executed in real time by the instruction execution module deployed in the solid-state logic controller that has completed logic reconstruction.

[0195] The controller first activates a set of environmental sensor probes at a preset physical location—the pipeline discharge terminal of the fly ash grading and recycling system—through its analog input module. This array primarily consists of a laser particle size scattering sensor array and a pipeline concentration detector. The laser particle size scattering sensor array is an online analytical instrument that infers particle size distribution by measuring the angle and intensity distribution of scattered light generated after a laser beam passes through an airflow containing particulate matter.

[0196] Pipeline concentration detectors typically employ microwave or capacitive methods to monitor the solid content or flow density of powdered materials in real time by measuring changes in the dielectric constant or microwave attenuation within the pipeline. The controller acquires raw electrical signals representing the current physical state from these sensors at a fixed scanning frequency determined by an internal clock pulse, for example, once every 100 milliseconds. Specifically, this includes a grading particle size error signal reflecting the difference between the particle size distribution of the graded fly ash and the set target, and an ash flow concentration fluctuation amplitude signal reflecting the uniformity of material transport within the pipeline.

[0197] It should be noted that the fixed scanning frequency is the reciprocal of the sampling and control cycle of the control system. Its setting must ensure a rapid response to process changes while avoiding unnecessary computational burden and system oscillation caused by excessively high frequencies. It is usually determined based on the time constant of the controlled object. For powder conveying systems, 10 Hz is a common value. Meanwhile, the classification particle size error electrical signal is a voltage or current signal directly output by the sensor's internal circuit after comparing the measured particle size distribution with the built-in target distribution. Its amplitude is proportional to the degree of difference between the two. In addition, the ash flow concentration oscillation amplitude electrical signal is the AC component amplitude of the output signal of the pipeline concentration detector. Its magnitude reflects the pulsation or non-uniformity of the material flow.

[0198] Upon receiving these analog electrical signals, the analog-to-digital converter (ADC) inside the controller immediately converts them into digital quantities. The digital quantity corresponding to the graded granularity error electrical signal is identified by the system and automatically mapped to the data input node of the executable logic control topology constructed in the previous step. The data input node is a specific memory address reserved when constructing the executable logic control topology, serving as the starting point of the entire data processing chain. This operation is equivalent to sending real-time process measurement values ​​into the input terminal of the logic topology in memory. After data injection, the interpreter engine's running thread begins to drive the calculation of the entire logic chain. It calls the arithmetic logic unit of the microprocessor core along the preset data transmission path, sequentially executing data processing. For example, the data first flows through the memory block mapped to the filtering algorithm, performing numerical smoothing filtering to remove noise; then, the processed data flows to the subtraction operation block, comparing it with the preset target value to generate a deviation value.

[0199] Finally, the deviation value is fed into the analog operation block for dynamic adjustment calculation, either proportionally amplified or integrally accumulated. The calculation process is executed sequentially along the entire topology chain until the last node. The interpreter collects and reads the control solution value located in the output register at the tail end of the executable logic control topology. This value is the final adjustment amount calculated by the entire control logic chain for the current real-time deviation.

[0200] The controller pushes this digital adjustment to the digital-to-analog converter chip. As can be understood, the digital-to-analog converter chip is a key interface device connecting the digital control core and the analog actuator, converting the binary code output by the processor into smoothly varying voltages or currents. This chip performs electrically isolated output operations, converting purely digital quantities into analog control signals or pulse signals with clear physical meaning that can directly drive physical equipment. These signals are the final control commands for the classifying equipment, such as pulse width modulation signals used to control the rotation of the main air valve servo motor by a specific step, or standard 4-20 mA current signals used to adjust the power supply frequency of the classifier's variable frequency traction motor.

[0201] It should be noted that the control commands for the grading equipment are the final manifestation of the control actions. They directly act on actuators such as air valves and motor frequency converters, adjusting the entire grading process by changing their physical state.

[0202] In one specific embodiment, during a certain control cycle, the solid-state logic controller scans at a frequency of 10 Hz. It first reads a -0.5 volt graded particle size error signal from the laser particle size scattering sensor array at the pipeline terminal, indicating that the average particle size of the current finished fly ash is coarser than the target value. This -0.5 volt signal is converted into a digital value, such as -512, after analog-to-digital conversion. This number -512 is directly written to the data input node of the executable logic control topology, i.e., memory location at address 0xA000. After data injection, the interpreter engine begins execution. It first calls the Filter_Input_Light algorithm module to perform a weighted average of the input value -512 and historical values ​​from the previous few cycles, outputting a smoothed value, such as -500, and storing it at address 0xA01C. Next, the data flows to the Subtract module, which reads the preset target value 0 from memory, executes 0 - (-500), obtains the deviation value +500, and stores it at address 0xA02C. Finally, this deviation value of +500 is sent to the PID_Standard module. This module calculates the deviation value based on its internally configured gain parameters, such as a proportional gain of 0.01, and outputs an adjustment of +5. The evolution of real-time deviation data processing by each algorithm block within the aforementioned underlying topology can be intuitively seen in [the documentation / reference needed]. Figure 9 .

[0203] The system compares the absolute value of the adjustment with the digital value corresponding to the preset 0.5% dead zone boundary limit threshold. If it determines that the adjustment exceeds the dead zone ignore range, it is confirmed as a valid action. This calculated control value +5 is then stored in the tail output register at address 0xA06C. The controller reads this value and pushes it to the digital-to-analog converter chip. This chip converts the digital value +5 into an incremental adjustment for the output range of 4 to 20 mA. Assuming the current output is 12 mA, an increase of 5 units corresponds to an increase of 0.5 mA, resulting in a new output current of 12.5 mA. This 12.5 mA current signal is the control command for the classifier. It is sent to the frequency controller of the classifier's variable frequency traction motor, causing it to increase its power supply frequency accordingly. This increases the classifier's speed, drawing more coarse particles back into the classification zone to correct particle size deviations.

[0204] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A fly ash grading control system based on object identifier resolution, characterized in that, include: The identifier construction module is used to obtain displacement signals of fly ash storage and conveying nodes, extract process condition snapshot data of associated equipment, and generate unique object identifiers. The feature decoding module is used to parse unique object identifiers, extract historical operating parameters, and decode historical operating parameters through a physical data hybrid model to obtain multi-dimensional feature vectors. The strategy matching module is used to obtain the current mechanical load parameters of the staged fan equipment. Based on the multi-dimensional feature vector and the current mechanical load parameters, it performs matching in the built-in control primitive library to extract the target operation primitive set. The sequence encapsulation module is used to perform topological recombination on the target operation primitive set and, in combination with preset process parameters, generate a batch instruction control sequence containing a digital signature. The authorization verification module is used to send the batch instruction control sequence to the logic execution interpretation engine, verify the digital signature, and generate an authorization running status signal; The topology reconfiguration module is used to load batch instruction control sequences into the microcontroller unit based on authorized operating status signals, and construct an executable logic control topology. The instruction execution module is used to acquire the real-time deviation acquisition feedback value of the environmental sensor probe, drive the executable logic control topology to calculate the real-time deviation acquisition feedback value, and generate hierarchical equipment control instructions.

2. The fly ash grading control system based on object identifier resolution according to claim 1, characterized in that, The identifier constructor module is specifically used for: Monitor the mechanical opening and closing action of the discharge valve of the upstream ash storage silo, and simultaneously accumulate the flow pulse value output by the bottom conveyor belt scale. When the flow pulse value exceeds the preset batch tonnage threshold, generate a physical batch start cutoff event. Obtain the timestamp parameter and hardware device code of the moment when the start truncation event occurs, extract key frames of fire images within the same time window based on the timestamp parameter, and construct process condition snapshot data; The timestamp parameter, hardware device code, and process condition snapshot data are serialized and concatenated, and then fused using a hash function to generate a unique object identifier.

3. The fly ash grading control system based on object identifier resolution according to claim 1, characterized in that, The feature decoding module is specifically used for: Use the unique object identifier as the query index, and extract the bound timestamp parameter and hardware device code as historical running parameters; Based on the timestamp parameter, the boiler combustion temperature time series and the current fluctuation curve of the electrostatic precipitator station are extracted from the associated production cycle. A frequency domain transformation operation is performed on the boiler combustion temperature time series to extract the combustion main frequency fluctuation amplitude parameter, and the spark discharge variance index of the current fluctuation curve is calculated and combined to form an intermediate feature vector. The intermediate feature vector is standardized and then input into the physical data hybrid model to infer the estimated content of fly ash glass microspheres. The combustion frequency fluctuation amplitude parameter, spark discharge variance index and the estimated content of fly ash glass microspheres are then concatenated to generate a multidimensional feature vector.

4. The fly ash grading control system based on object identifier resolution according to claim 3, characterized in that, The characteristic decoding module is specifically used for calculating the spark discharge variance index of the current fluctuation curve in the following ways: Obtain the current sample value at each data point in the current fluctuation curve, and calculate the average current value of all current sample values; Calculate the difference between each current sample value and the average current value, and square the difference to obtain the squared difference of each data point; The squared differences of all data points are calculated to obtain the spark discharge variance index.

5. A fly ash grading control system based on object identifier resolution according to claim 1, characterized in that, The strategy matching module is specifically used for: The estimated content of vitreous microspheres in fly ash was extracted from the multidimensional eigenvectors. Obtain the preset reference glass microsphere content and reference operating current reference value, calculate the first ratio of the reference glass microsphere content to the estimated value of fly ash glass microsphere content, and the second ratio of the current mechanical load parameter to the reference operating current reference value; Based on preset weighting coefficients, the first ratio and the second ratio are weighted and summed to obtain the flow state early warning coefficient. Based on the numerical range mapped by the flow state early warning coefficient, the built-in tags of the built-in control primitive library are retrieved, the matching processing node sequence is extracted, and the target operation primitive set is generated.

6. A fly ash grading control system based on object identifier resolution according to claim 1, characterized in that, The sequence encapsulation module is specifically used for: Analyze the data input / output interface configuration of each independent processing node within the target operation primitive set, sort the nodes according to the control causal transmission logic, and establish a control logic chain. Configure the deviation target setpoint and feedforward reference offset at the injection end of the control logic chain, and write the action dead zone boundary limit threshold at the output end of the tail to form batch control logic script code. Extract the unique object identifier and the device factory verification code of the target controller, fuse them to generate a symmetric encryption derivative key, combine it with the obtained current clock random number factor, perform encryption operation on the batch control logic script code, generate a digital signature, and encapsulate it to obtain the batch instruction control sequence.

7. A fly ash grading control system based on object identifier resolution according to claim 1, characterized in that, The authorization verification module is specifically used for: Write the batch instruction control sequence into the high-speed receive buffer hardware area of ​​the logic execution interpretation engine, and lock the current control output to pause logic switching; Read the device factory verification code of the target controller, and use the device factory verification code to perform reverse comparison and verification of the digital signature carried by the batch instruction control sequence; When the verification is successful, the memory addressing lock state of the protected area is released and an authorized running status signal is output; when the verification fails, a mechanical lock is triggered and an out-of-bounds alarm pulse is output.

8. A fly ash grading control system based on object identifier resolution according to claim 1, characterized in that, The topology reconfiguration module is specifically used for: Extract the level state transition event of the authorized running status signal by listening to the pins, and clear the preset proportional-integral-differential parameter table in memory; Write the batch instruction control sequence into the instruction temporary register, and scan the control node step-by-step execution pointer of the batch instruction control sequence line by line; In the memory stack area, an independent local simulation storage block is allocated for each atomic-level algorithm subroutine. A data assignment link is established from the output of the preceding process algorithm subroutine to the input of the following process algorithm subroutine, generating an executable logic control topology.

9. A fly ash grading control system based on object identifier resolution according to claim 1, characterized in that, The instruction execution module is specifically used for: The graded particle size error electrical signal and the ash flow concentration oscillation amplitude electrical signal of the physical entity are extracted according to a fixed scanning frequency and used as real-time deviation acquisition feedback values. The graded granularity error electrical signal is converted into a digital quantity and injected into the data input node of the executable logic control topology. Along the preset data transmission path, data smoothing filtering and proportional amplification deviation numerical simulation operations are performed sequentially to obtain the control solution value; The calculated control values ​​are pushed to the digital-to-analog converter chip for electrically isolated output, and then converted into hierarchical equipment control commands.

10. A fly ash grading control system based on object identifier resolution according to claim 9, characterized in that, When the calculated control values ​​are pushed to the digital-to-analog converter chip for electrically isolated output, and converted into hierarchical equipment control commands, they are specifically used for: Convert the calculated control values ​​into analog control signals; The analog control signal is used as the control command for the classifying equipment and output to the main air valve servo motor to drive the rotation step, or output to the classifier variable frequency traction motor to adjust the power supply frequency.

Images