A self-adaptive control system for modifying dehydration and stirring parameters of a bored pile mud

Abstract

The present application relates to the technical field of construction waste slurry treatment, and discloses a self-adaptive control system for modified dehydration and stirring parameters of cast-in-place pile slurry, which comprises driving execution, baseline decoupling, phase state analysis, feature extraction, optimization decision and wall breaking control modules. The system constructs a feedforward decoupling matrix by collecting the stator active current and rotating speed data of the motor, extracts the transient total phase difference to calculate the rheological phase state index of the slurry, and calculates the equivalent macroscopic rotational inertia. Based on the above parameters, the system dynamically adjusts the dosing duty cycle command, evaluates the flocculation process in combination with the drug slurry concentration ratio data, and issues the discharging command. At the same time, the system calculates the rotating speed prediction residual in real time, and issues the alternating pulse torque when the residual breaks through the adaptive dynamic threshold. The present application eliminates mechanical background interference, realizes dynamic quantification of the rheological state of the slurry and closed-loop control of the dosing process, can prevent the slurry from sticking to the shaft and mechanical overload during the stirring process, and ensures the continuous and stable operation of the equipment.

Description

Technical Field

[0001] This invention relates to the field of construction waste mud treatment technology, specifically an adaptive control system for the mixing parameters of grouting pile mud modification and dehydration. Background Technology

[0002] In the construction of cast-in-place piles in building foundation engineering, a large amount of waste mud with high water content is generated. In order to meet the requirements of environmental protection and transportation, it is necessary to continuously add modifiers to the mud and mix it to promote the flocculation, dehydration and initial solidification of suspended particles in the mud. The control level of this mixing process directly determines the final treatment quality and dehydration efficiency of the mud.

[0003] Existing mud modification mixing equipment uses fixed operating parameters or relies solely on the observation experience of on-site personnel to manually adjust the dosing rate and mixing speed. However, after the mud is mixed with the modifying agent, the internal components undergo physicochemical reactions, resulting in the mud viscosity exhibiting strong nonlinear rheological characteristics. Furthermore, the entire flocculation process is in a complex dynamic state. Traditional fixed parameter control methods lack real-time monitoring and quantification of the evolution of the mud properties inside the mixer, and the system cannot synchronously adjust its operation based on fluctuations in the actual mud concentration.

[0004] Because the system cannot adapt to the dynamic rheological characteristics of the mud, adding chemicals at a fixed rate will cause the chemicals to become oversaturated in the local fluid, resulting in uneven flocculation and mixing. At the same time, in the high-concentration stage of mud treatment, a large amount of mud may adhere to the agitator blades or even cause them to seize up. At this time, conventional control logic cannot identify this sudden change in resistance in time and take proactive intervention measures, which will lead to mechanical overload of the agitator motor and transmission mechanism, ultimately causing the dewatering process to stop. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an adaptive control system for the mixing parameters of grouting pile mud modification and dewatering. This system solves the problems of existing mud modification mixing control methods being unable to quantify in real time and adapt to changes in the dynamic rheological properties of mud, leading to uneven mixing of reagents in certain areas and frequent blade seizure and mechanical overload during high mud concentration stages.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an adaptive control system for the mixing parameters of grouting pile mud modification and dewatering, comprising: The drive execution module is used to send operation control commands to the AC asynchronous stirring motor and collect transient data of stator active current and rotor speed feedback data. The baseline decoupling module is used to construct a two-dimensional lookup table based on the stator active current transient data and rotor speed feedback data to generate a feedforward decoupling matrix; The phase state analysis module is used to extract the transient total phase difference based on the stator active current transient data and rotor speed feedback data, and to calculate the mud rheological phase state index in combination with the feedforward decoupling matrix; The feature extraction module is used to calculate the equivalent macroscopic moment of inertia based on the transient data of stator active current and the feedback data of rotor speed, generate dosing control commands based on the mud rheological phase state index, and obtain the slurry concentration ratio data. The optimization decision module is used to send stop dosing and discharge control commands based on mud rheological phase state index, equivalent macroscopic rotational inertia and slurry concentration ratio data. The cell wall breaking control module is used to calculate the speed prediction residual in real time based on the stator active current transient data, rotor speed feedback data and equivalent macroscopic rotational inertia, and to issue an alternating pulse torque command to the drive execution module when the speed prediction residual exceeds the adaptive dynamic discrimination threshold.

[0007] This invention utilizes baseline decoupling to eliminate background interference from mechanical transmission, dynamically analyzes the phase evolution and rotational inertia of mud based on electrical parameters, constructs multidimensional state characteristics to evaluate the flocculation reaction process, and introduces a state observation mechanism to identify adhesion disturbances in real time and execute active torque impact, thus completing closed-loop control and equipment protection for the entire mud modification process.

[0008] Furthermore, the drive execution module is internally configured with an industrial controller and a frequency converter. The frequency converter collects stator phase current, decouples it to obtain stator active current transient data, and uploads the stator active current transient data and rotor speed feedback data to the industrial controller. The industrial controller calls a digital low-pass filter to perform anti-aliasing processing on the stator active current transient data and rotor speed feedback data. When the industrial controller issues a run control command to the frequency converter, it records a first system timestamp. When it receives the stator active current transient data and rotor speed feedback data, it records a second system timestamp. The industrial controller calculates the difference between the first system timestamp and the second system timestamp as the total transmission delay of the link and converts it into a step size index to perform timing synchronization alignment.

[0009] This invention employs a digital low-pass filter and a synchronization alignment mechanism based on timestamp differences to filter out high-frequency electromagnetic switch noise and compensate for data lag caused by bus transmission and node operations. This mechanism ensures that electrical sampling parameters and control commands are strictly aligned on the time axis, avoiding deviations in state space calculation caused by timing misalignment.

[0010] Furthermore, the baseline decoupling module sends a frequency sweep test command containing multiple step frequency points to the drive execution module and maintains a steady-state dwell time at the step frequency points. The baseline decoupling module extracts the data array within a fixed time window, accumulates and averages it to obtain the mean steady-state current and the mean base speed. Based on the mean steady-state current and the mean base speed, the baseline decoupling module performs a discrete Fourier transform to calculate the mechanical inherent phase lag. The baseline decoupling module constructs a two-dimensional lookup table with the mean base speed as the horizontal axis index and the mean steady-state current as the vertical axis index. The calculated mechanical inherent phase lag is used as the node value to fill the two-dimensional lookup table to generate the feedforward decoupling matrix.

[0011] By conducting multi-frequency dwell tests and frequency domain transformation, the fundamental phase component under steady-state conditions is extracted, and the viscous frictional resistance and inherent damping characteristics of the stirring mechanism are removed, thereby providing a mechanical bias compensation benchmark for rheological property analysis.

[0012] Furthermore, the phase state analysis module sends a perturbation angular velocity command to the drive execution module. The phase state analysis module calls the orthogonal demodulation operation unit to perform orthogonal demodulation calculation to extract the transient total phase difference. The phase state analysis module performs a subtraction algebraic operation on the transient total phase difference and the mechanical inherent phase lag obtained by calling the feedforward decoupling matrix to obtain the rheological phase difference parameter. The phase state analysis module performs normalization calculation on the rheological phase difference parameter to generate a dimensionless characteristic index, and outputs the mud rheological phase state index accordingly. The system superimposes the perturbation command on the base rotation speed, uses orthogonal demodulation to extract the response phase and subtracts the baseline lag, and transforms the electromechanical coupling parameter into a characteristic index reflecting the viscoelastic damping of the mud, thereby quantifying the phase state change process after chemical dosing.

[0013] Furthermore, the phase analysis module monitors the numerical change gradient of the transient total phase difference output by adjacent calculation cycles in real time. When the phase analysis module detects that the absolute value of the phase jump between two adjacent discrete time points exceeds the preset threshold, it triggers the phase unwinding processing mechanism, which pulls the transient total phase difference at the current abnormal crossover point back to the continuous physical change trajectory range by adding or subtracting a preset period integer multiple.

[0014] This module monitors numerical faults caused by periodic range limitations in phase calculation, and reconstructs continuous phase change trajectories through whole-cycle compensation to prevent logic distortions in the dynamic demodulation process.

[0015] Furthermore, the feature extraction module receives the mud rheological phase state index and performs discrete-time derivative calculation to obtain the phase state evolution gradient. The feature extraction module calculates the difference between the time stamp of the phase state evolution gradient reaching the extreme point and the time stamp of the dosing start and defines it as the diffusion characteristic time constant. Based on the diffusion characteristic time constant, the feature extraction module adjusts the operating frequency of the dosing pump inversely to generate the dosing pump time duty cycle instruction as the dosing control instruction.

[0016] The diffusion characteristic time constant is obtained by calculating the time difference of the extreme points of the phase evolution gradient, which characterizes the hysteresis state of the mud response to the agent mixing. Based on this, the system dynamically adjusts the dosing duty cycle to prevent local oversaturation of the agent concentration and maintain uniform flocculation reaction.

[0017] Furthermore, during the stop interval corresponding to the dosing pump time duty cycle command, the feature extraction module sends a macroscopic transient speed step command to the drive execution module. The feature extraction module linearly and equivalently converts the stator active current transient data into transient electromagnetic torque data, and uses a differential algorithm to perform discrete-time differentiation on the rotor speed feedback data to generate transient angular acceleration. Based on the transient electromagnetic torque data and transient angular acceleration, the feature extraction module calculates the equivalent macroscopic moment of inertia, extracts the rising edge dynamic window, performs the averaging operation, and latches it.

[0018] During the interval between chemical dosing and shutdown, transient torque and angular acceleration differential calculations are performed to avoid aliasing interference between control commands. This process identifies the equivalent rotational inertia of the mud based on the dynamic equilibrium relationship and uses it as a state parameter characterizing the change in mud consistency.

[0019] Furthermore, the optimization decision module extracts the mud rheological phase state index, equivalent macroscopic rotational inertia, and slurry concentration ratio data to form a three-dimensional state input vector. The optimization decision module sends the three-dimensional state input vector into a three-layer feedforward neural network architecture to map and output a completion probability assessment value. The internal register of the optimization decision module is configured with a sliding judgment window and a shutdown trigger judgment threshold. When the completion probability assessment value fills the sliding judgment window and all sample extraction values ​​in the window are higher than the shutdown trigger judgment threshold, a system flocculation completion flag is generated and a stop dosing and discharge control command is sent.

[0020] This module uses a feedforward neural network to perform nonlinear mapping evaluation of rheological phase state, rotational inertia and concentration data. Combined with a sliding decision window mechanism, it filters out abrupt changes in single calculations caused by local fluid clumps, thereby accurately identifying the steady-state critical point at which mud flocculation is completed.

[0021] Furthermore, a nonlinear state observer is built inside the cell-breaking control module to convert the transient stator active current data returned by the drive execution module into transient electromagnetic torque data. Based on the transient electromagnetic torque data, the equivalent macroscopic moment of inertia, and the extracted basic load resistance torque, the ideal predicted speed is predicted. The cell-breaking control module calculates the deviation between the ideal predicted speed and the rotor speed feedback data to obtain the speed prediction residual. The cell-breaking control module extracts the moving average and standard deviation of the speed prediction residual within a continuous sampling period and performs a linear combination to generate an adaptive dynamic discrimination threshold.

[0022] By calculating the residual between the ideal predicted speed and the actual feedback speed, the adhesion disturbance experienced by the stirring impeller is quantified. The system combines sliding window statistics to generate an adaptive dynamic threshold, which can accommodate the increase in background disturbance when the mud concentration rises normally, and avoids false triggering caused by using a fixed threshold.

[0023] Furthermore, after the speed prediction residual exceeds the adaptive dynamic discrimination threshold and the nonlinear anomaly flag is captured, the wall-breaking control module generates an alternating pulse torque command and injects it into the torque feedforward loop of the drive execution module. The wall-breaking control module adopts an alternating execution strategy of intermittent pulse impact and closed-loop detection. When it detects that the speed prediction residual has fallen back to below the preset proportional baseline of the adaptive dynamic discrimination threshold, it cancels the alternating pulse torque command and simultaneously activates the internally configured forced cooling timer.

[0024] When the speed prediction residual exceeds the threshold, the system injects alternating pulse torque to break the mud adhesion layer and solve the shaft seizure problem. By alternating between intermittent impact and closed-loop detection, the system introduces forced cooling time after the anomaly is resolved to isolate high-frequency oscillation data and prevent it from affecting the normal operation of the upper-level optimization decision algorithm.

[0025] This invention provides an adaptive control system for the mixing parameters of grouting pile mud modification and dewatering. It has the following beneficial effects: 1. This invention constructs a feedforward decoupling matrix through a baseline decoupling module and extracts the transient total phase difference in the phase state analysis module by combining orthogonal demodulation, thereby calculating the mud rheological phase state index. This structure, based on eliminating the inherent mechanical friction and damping interference of the stirring mechanism, directly utilizes the active current and speed feedback data of the motor to achieve dynamic quantification of the nonlinear rheological process of mud. This overcomes the shortcomings of traditional processes that rely on manual experience or single physical quantity settings, and provides an accurate state judgment basis for the adaptive control of the system.

[0026] 2. This invention obtains the diffusion characteristic time constant by calculating the time difference of the extreme points of the phase evolution gradient through the feature extraction module, and adjusts the duty cycle command of the dosing pump inversely based on this. At the same time, it calculates the equivalent macroscopic rotational inertia of the mud during the dosing stop interval. This control method, which dynamically adjusts the dosing frequency based on the actual mixing response state of the mud, avoids the problem of local oversaturation of the agent caused by fixed-rate dosing, ensures uniform mixing of mud and modified agent in the stirred flow field, and improves the overall flocculation reaction quality.

[0027] 3. This invention uses a nonlinear state observer to calculate the rotational speed prediction residual through a cell breaking control module, and combines it with sliding window statistics to generate an adaptive dynamic discrimination threshold. When the residual exceeds the threshold, an alternating pulse torque command is sent to the drive execution module. This mechanism can accurately identify blade adhesion and shaft seizing phenomena that occur in the high concentration stage, while being compatible with the background disturbance when the mud concentration rises normally. The system actively injects alternating impact torque to destroy the mud adhesion layer, prevents equipment overload and deadlock, and ensures the continuous and stable operation of dewatering and mixing. Attached Figure Description

[0028] Figure 1 This is a diagram of the module architecture of the present invention; Figure 2 This is a flowchart of the method of the present invention; Figure 3 This is a timing diagram of the physical topology and data interaction of the drive execution module of the present invention; Figure 4 This is a timing diagram of the drug administration coordination record and macroscopic inertia estimation of the present invention; Figure 5 This is a logic diagram of the multi-dimensional state space fusion and decision-making process of the present invention; Figure 6 This is a diagram of the nonlinear anomaly detection and self-healing anti-disturbance logic structure of the present invention; Figure 7 This is a time-series dynamic response diagram of rotational speed and current before and after the cell-wall breaking control of the present invention; Figure 8 This is a tracking diagram of the adaptive dynamic discrimination threshold and rotational speed prediction residual of the present invention; Figure 9 This is a diagram showing the evolution of the output state of the neural network of the optimization decision module of this invention.

[0029] Among them, 10 is the drive execution module; 20 is the baseline decoupling module; 30 is the phase analysis module; 40 is the feature extraction module; 50 is the optimization decision module; and 60 is the cell wall breaking control module. Detailed Implementation

[0030] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] See attached document Figure 1The present invention provides an adaptive control system for the mixing parameters of the mud modification and dewatering of cast-in-place piles. The system may include: a drive execution module 10, a baseline decoupling module 20, a phase state analysis module 30, a feature extraction module 40, an optimization decision module 50, and a cell wall breaking control module 60.

[0032] The drive execution module 10 includes an industrial controller and a variable frequency drive with closed-loop vector control function. The drive execution module 10 is used to receive the operation control commands output by each calculation and decision module and send them to the AC asynchronous stirring motor. It synchronously collects the stator active current transient data and rotor speed feedback data of the AC asynchronous stirring motor and sends the collected stator active current transient data and rotor speed feedback data back to the corresponding calculation module.

[0033] The baseline decoupling module 20 is used to issue a frequency sweep test command to the drive execution module 10 during the initial pumping stage of mud without the addition of modified agents. It also uses the mud pumping pump to change the pure mud level in the mixing tank in steps to provide multiple sets of different basic load resistance torques. Based on the stator active current transient data and rotor speed feedback data returned by the drive execution module 10, a two-dimensional lookup table is constructed to generate a feedforward decoupling matrix and output it to the phase state analysis module 30.

[0034] The phase state analysis module 30 is used to send a perturbation angular velocity command with a low-frequency sine waveform to the drive execution module 10. Based on the returned stator active current transient data and rotor speed feedback data, it performs a phase-locked demodulation algorithm to analyze and obtain the transient total phase difference. After compensation by calling the feedforward decoupling matrix output by the baseline decoupling module 20, it calculates the mud rheological phase state index and outputs the mud rheological phase state index to the feature extraction module 40, the optimization decision module 50 and the wall breaking control module 60.

[0035] The feature extraction module 40 is used to receive the mud rheological phase state index during the mud dosing stage to calculate the diffusion characteristic time constant and output the dosing pump time duty cycle command to the drive execution module 10. The feature extraction module 40 also sends a macroscopic transient speed step command to the drive execution module 10 during the dosing pump stop interval. Based on the returned stator active current transient data and rotor speed feedback data, it calculates the equivalent macroscopic rotational inertia of the stirring transmission mechanism and outputs the equivalent macroscopic rotational inertia to the optimization decision module 50 and the cell wall breaking control module 60.

[0036] The optimization decision module 50 is used to receive the mud rheological phase state index output by the phase state analysis module 30 and the equivalent macroscopic rotational inertia and slurry concentration ratio data output by the feature extraction module 40, construct a three-dimensional state input vector and send it into the feedforward neural network. When the completion probability evaluation value output by the network satisfies the flocculation steady-state critical point condition, it sends a stop dosing and discharge control command to the drive execution module 10.

[0037] The cell disruption control module 60 receives the stator active current transient data returned by the drive execution module 10 and internally converts it into transient electromagnetic torque data. It also monitors the rotor speed feedback data and the equivalent macroscopic moment of inertia output by the feature extraction module 40. When it detects that the speed prediction residual calculated based on the state observer exceeds the adaptive dynamic discrimination threshold and thus causes a nonlinear anomaly, the cell disruption control module 60 intercepts the judgment process of the optimization decision module 50, sends an alternating pulse torque command with asymmetric amplitude containing the lower limit safety frequency parameter to the drive execution module 10, and sends a data sampling shielding countdown signal to the phase state analysis module 30 and the feature extraction module 40 to suspend the data sampling state after execution.

[0038] See attached document Figure 2 This invention provides an adaptive control method for the mixing parameters of grouting pile mud modification and dewatering, comprising the following steps: S1, the baseline decoupling module 20 controls the AC asynchronous stirring motor to run in the preset basic speed range through the drive execution module 10, and coordinates the control of the mud pump to change the pure mud level in the mixing tank in steps to provide multiple sets of different basic load resistance torques. It extracts the mechanical inherent phase lag based on the stator active current transient data and rotor speed feedback data to construct a two-dimensional lookup table, completes the generation of the feedforward decoupling matrix, and configures it as the basic parameter for system calls. S2, the phase state analysis module 30 sends a micro-perturbation angular velocity command to the drive execution module 10, performs orthogonal demodulation calculation based on the returned stator active current transient data and rotor speed feedback data to obtain the transient total phase difference, calls the feedforward decoupling matrix to perform subtraction compensation operation, calculates the mud rheological phase state index and outputs it to the feature extraction module 40. S3, the feature extraction module 40, based on the received mud rheological phase state index, controls the dosing pump to output test reagent through the drive execution module 10, calculates the difference time when the time derivative of the mud rheological phase state index reaches the preset threshold to obtain the diffusion feature time constant, and sets the dosing pump time duty cycle instruction of the drive execution module 10 to control the dosing pump according to the diffusion feature time constant. S4, the feature extraction module 40 sends a macroscopic transient speed step command to the drive execution module 10 within the stop gap corresponding to the dosing pump time duty cycle command set according to the diffusion feature time constant. It extracts the stator active current transient data time integral parameters and rotor speed feedback data change rate to calculate the equivalent macroscopic moment of inertia and outputs it. S5 performs multi-dimensional state space fusion and anti-counterfeiting identification. The optimization decision module 50 receives the mud rheological phase state index obtained by analysis and the equivalent macroscopic rotational inertia and the slurry concentration ratio data obtained by calculation to construct a three-dimensional state input vector. It outputs the probability evaluation value of the completion of the mud solidification reaction through feedforward neural network mapping to determine whether the flocculation steady-state critical point condition is met. The cell wall breaking control module 60 synchronously receives the transient electromagnetic torque data converted from the stator active current transient data, the rotor speed feedback data and the equivalent macroscopic rotational inertia to calculate the speed prediction residual in real time and perform nonlinear anomaly identification. S6, if the completion probability assessment value meets the flocculation steady-state critical point condition and the cell wall breaking control module 60 does not detect nonlinear anomalies, the optimization decision module 50 executes the final stop dosing and discharge control command through the drive execution module 10. If the cell wall breaking control module 60 detects that the speed prediction residual exceeds the adaptive dynamic discrimination threshold and continues to exceed the set time tolerance, the cell wall breaking control module 60 cuts off the output control path of the optimization decision module 50 and sends an alternating pulse torque command with a frequency lower limit constraint to the drive execution module 10. After the set data sampling shielding countdown signal corresponding to the delay period, the system is made to return to the data extraction and cyclic judgment process of mud rheological phase state index and equivalent macroscopic rotational inertia.

[0039] See attached document Figure 3 The drive execution module 10 realizes data interaction between the system computing module and the underlying hardware by constructing the underlying hardware driver topology. In this embodiment, the drive execution module 10 is internally configured with an industrial controller as the master station and a frequency converter as the slave station. The industrial controller establishes a real-time communication link with the frequency converter through an industrial Ethernet bus. The frequency converter has a closed-loop vector control algorithm embedded in it and is connected to the AC asynchronous stirring motor set to 10 Hz to 50 Hz via a power line. The industrial controller receives the operation control instructions issued by other computing modules and parses them into bus messages. After receiving the messages, the frequency converter converts them into a three-phase inverter drive voltage for the AC asynchronous stirring motor. The coordinate transformation and space vector pulse width modulation process in the closed-loop vector control can be conventionally configured according to the principles of motor mechanics.

[0040] The high-speed analog-to-digital conversion channel inside the variable frequency drive acquires the stator phase current and rotor encoder pulse signal of the AC asynchronous mixer motor in real time. The variable frequency drive decouples the stator phase current through Parker transformation to obtain the stator active current transient data and packages it with the rotor speed feedback data and uploads it to the industrial controller. As a preferred method, after receiving the raw feedback data containing wideband electromagnetic noise, the industrial controller calls a digital low-pass filter to perform anti-aliasing processing to filter out high-frequency glitches caused by the high-frequency switching of the inverter. Specifically, an infinite impulse response filtering algorithm is used to smooth the stator active current transient data and rotor speed feedback data.

[0041] The formula for calculating the discrete-time filtering is as follows: ; In the formula, The filtered data output value represents the current sampling period; The raw feedback data input value represents the current sampling period; This represents the filtered data output value from the previous sampling period; This represents the filter coefficients calculated based on the sampling frequency and the system cutoff frequency. The specific value range is set as follows ; The calculation formula is: , In the formula, To set a low-pass cutoff frequency in the range of 100 Hz to 300 Hz; The sampling period of the underlying hardware is used as the basis for this filtering formula, which is applied independently to the processing links of stator active current transient data and rotor speed feedback data.

[0042] The industrial controller performs communication delay calibration and timing synchronization alignment mechanisms to ensure data timeline matching. Due to the time overhead of industrial Ethernet bus transmission and internal calculations of the frequency converter, the industrial controller is equipped with a clock synchronization channel and a delay compensation buffer queue. When the industrial controller issues a running control command to the frequency converter, it records the first system timestamp. When it receives the filtered stator active current transient data and rotor speed feedback data returned by the frequency converter, it records the second system timestamp. The industrial controller calculates the difference between the two timestamps as the total link transmission delay and converts it into a step index, which is then injected into the delay compensation buffer queue.

[0043] The industrial controller shifts the currently transmitted stator active current transient data and rotor speed feedback data on the time axis based on the step size index, aligning them to the exact phase point when the corresponding command is issued. When the buffer data addressing the corresponding step size index is empty, the industrial controller executes a linear interpolation algorithm to generate a temporary value to fill the queue in combination with the buffer data from the previous two cycles, preventing the timing alignment operation from triggering deadlock. After the timing synchronization alignment is completed, the stator active current transient data and rotor speed feedback data are transmitted back to the corresponding computing module by the industrial controller for direct use by subsequent algorithms.

[0044] The baseline decoupling module 20 is activated during the initial system startup and the pure mud pumping stage without the addition of modified agents. In this embodiment, there is gear backlash and bearing friction damping inside the stirring transmission mechanism. The baseline decoupling module 20 directly sends a frequency sweep test command containing multiple discrete step frequency points and also superimposed with low-frequency sinusoidal micro-perturbation waves to the drive execution module 10. It also coordinates the control of the mud pumping pump to change the pure mud liquid level height in the mixing tank stepwise to provide multiple sets of different basic load resistance torques. The test frequency band corresponding to the frequency sweep test command is set to 10Hz to 50Hz, and the span between two adjacent discrete step frequency points is set to 5Hz. The baseline decoupling module 20 controls the drive execution module 10 to maintain a preset steady-state dwell time of 5 to 10 seconds at each step frequency point.

[0045] During the dwell period at each step frequency point, the baseline decoupling module 20 continuously receives the stator active current transient data and rotor speed feedback data transmitted back by the drive execution module 10. The baseline decoupling module 20 removes the dynamic overshoot transient data of speed switching in the first part of the dwell time, and only extracts the data array within a fixed time window of 2 to 4 seconds in the second part of the dwell time for smoothing. By averaging the data in this time window, the mean steady-state current and mean base speed corresponding to the current step frequency point are obtained. This averaging operation removes the random data deviation caused by mechanical transmission vibration.

[0046] As a preferred method, the basic principle of obtaining the phase shift characteristics caused by pure mechanical damping is to regard the stator current as a response signal containing the system's inherent damping characteristics, and to perform orthogonal multiplication and integration with the sine and cosine reference signals of the reference speed to extract specific fundamental phase components. Based on this physical principle, the baseline decoupling module 20 performs discrete Fourier transform based on the extracted mean steady-state current and mean base speed to calculate the mechanical inherent phase hysteresis under the corresponding operating conditions.

[0047] The formula for calculating the phase difference extraction in the discrete domain is as follows: ; In the formula, This represents the calculated mechanical inherent phase hysteresis. This represents the angular frequency corresponding to the perturbation angular velocity command. This represents the discrete sequence of transient stator active current data within the steady-state window. It is the arctangent function; representing the angular frequency corresponding to the low-frequency sinusoidal perturbation wave; Represents the sampling period of the underlying hardware; Represents the index of discrete sampling points within a time window; This represents the theoretical reference phase angle corresponding to the rotor speed feedback data; the baseline offset caused by pure mechanical damping is extracted using this formula.

[0048] After traversing all the step frequency points of the sweep frequency test command, the baseline decoupling module 20 reorganizes the acquired data into a grid. Specifically, it constructs a two-dimensional lookup table with the average base speed as the horizontal axis index and the average steady-state current as the vertical axis index. The calculated mechanical inherent phase hysteresis of each group is used as the node value to fill the coordinate grid corresponding to the two-dimensional lookup table. This two-dimensional lookup table completely maps the viscous friction resistance characteristics in the entire mechanical transmission space in a physical sense. The baseline decoupling module 20 then solidifies the initialized two-dimensional lookup table into a static feedforward decoupling matrix.

[0049] After the feedforward decoupling matrix is ​​constructed, the baseline decoupling module 20 incorporates a bilinear interpolation algorithm and a boundary clamping protection mechanism to handle data requests from non-fixed grid nodes. When the rotational speed or current coordinates of the system's real-time operating point do not accurately hit the fixed grid nodes of the matrix, the baseline decoupling module 20 uses the mechanical inherent phase lag of the four adjacent nodes to perform distance-weighted average calculation to output a smooth compensation value. If the operating point exceeds the matrix boundary, it is forcibly clamped to the value of the outermost node. The configured feedforward decoupling matrix is ​​directly output by the baseline decoupling module 20 and mapped to the memory space where the phase state analysis module 30 is located.

[0050] The phase state analysis module 30 is activated after receiving the status signal that the system has entered the mud dosing stage. In this embodiment, the phase state analysis module 30 sends a perturbation angular velocity command to the drive execution module 10. The perturbation angular velocity command contains a low-frequency sine wave component. The phase state analysis module 30 directly superimposes the low-frequency sine wave component into the current steady-state base speed command of the frequency converter in digital form. As a preferred method, the injection frequency of the low-frequency sine wave component is set between 0.5Hz and 2.0Hz, and its amplitude is set to 2% to 5% of the steady-state base speed command. The selection and configuration of this frequency band and amplitude avoid the high-frequency rigid resonance region of the mechanical transmission and form a low-frequency shear field that satisfies the viscoelastic deformation of the mud polymer flocculation process.

[0051] The drive execution module 10 drives the AC asynchronous mixer motor to run according to the superimposed instructions and synchronously transmits the transient data of stator active current and the feedback data of rotor speed. The phase state analysis module 30 receives this dynamic data stream and inputs it into the quadrature demodulation unit. Based on the physical principle of bandpass mixing and filtering of stator response current and drive reference signal, the discrete mathematical expression of the quadrature demodulation calculation performed by the phase state analysis module 30 is as follows: ; In the formula, Represents the transient total phase difference, including mechanical background characteristics; It is the arctangent function; This represents a discrete sequence of transient data of the stator active current excited by perturbation. This represents the angular frequency corresponding to the perturbation angular velocity command. Represents the sampling period of the underlying hardware; Represents the index of the sampling point within the sliding calculation window; The total length of the data in the sliding calculation window and its value corresponds to the entire perturbation sine period; The reference reference phase angle is reconstructed from the rotor speed feedback data. The reconstruction logic of the reference reference phase angle is as follows: the phase analysis module 30 uses the same low-frequency sine wave component and the corresponding cosine wave component with a 90-degree phase shift to independently perform the same orthogonal multiplication and time window integration arctangent operation on the rotor speed feedback data sequence to extract the basic speed response phase.

[0052] In the dynamic computation sequence of continuously executing the orthogonal demodulation algorithm to extract the transient total phase difference, there is a phase entanglement phenomenon. The phase state analysis module 30 monitors the numerical change gradient of the transient total phase difference output between two adjacent calculation cycles in real time. When the phase state analysis module 30 detects that the absolute value of the phase jump between two adjacent discrete time points exceeds a preset threshold, At that time, the phase state analysis module 30 automatically triggers the phase unwinding processing mechanism, by adding or subtracting 2 from the transient total phase difference at the current abnormal boundary point. The integer multiples of the value pull it back to the continuous physical change trajectory range. This unwinding judgment and compensation logic avoids waveform discontinuity and upper logic misjudgment dead zone caused by the periodic limitation of the arctangent function range.

[0053] After obtaining the continuous transient total phase difference data sequence, the phase state analysis module 30 enters the feedforward decoupling compensation stage. The phase state analysis module 30 extracts the steady-state mean values ​​of the base speed and current before the injection of micro-perturbation under the current working condition as index coordinates. The phase state analysis module 30 uses the coordinates to address and call the feedforward decoupling matrix stored in the baseline decoupling module 20 to query the corresponding mechanical inherent phase lag. The phase state analysis module 30 performs a subtraction algebra operation on the transient total phase difference and the queried mechanical inherent phase lag to remove the rigid viscous resistance interference of the transmission mechanism. The difference obtained after the subtraction compensation operation is the rheological phase difference parameter. The rheological phase difference parameter extracts the viscoelastic damping dissipation characteristics of the microstructure of the mud flocculation network under the action of modified agents at the physical level.

[0054] Phase analysis module 30 performs normalization calculations on the extracted rheological phase difference parameters to generate dimensionless characteristic indices, the calculation formula of which is: ; In the formula, The mud flow phase state index represents the calculated output; Represents the rheological phase difference parameter after the removal of mechanical interference; The lower limit of phase difference is the reference physical measurement value extracted during the solidification stage of peak concentration mud. For the upper limit of phase difference reference physical measurement value extracted during the pure water no-load trial operation, the phase state analysis module 30 calculates the obtained value. The amplitude limiting clamping operation is performed to forcibly constrain its numerical boundary within a closed interval of 0 to 1. This amplitude limiting logic prevents normalization from going out of bounds due to sudden changes in high-frequency electromagnetic noise, which could lead to abnormal dead zones in downstream calculations. The resulting mud rheological phase state index is written into the common buffer by the phase state analysis module 30 and simultaneously distributed to the optimization decision module 50 and the wall breaking control module 60 to drive subsequent state transitions.

[0055] See attached document Figure 4 The feature extraction module 40 starts parallel monitoring when the system starts the synchronous dosing control process. In this embodiment, the feature extraction module 40 reads the transient flow feedback signal of the dosing pump in real time according to the basic control rhythm through the standard analog hard-wired interface. The feature extraction module 40 performs discrete integral accumulation operation on the transient flow feedback signal in the time dimension to obtain the absolute cumulative dosing amount at the current moment. The absolute cumulative dosing amount, combined with the pre-entered mud basic volume parameters, is directly converted into real-time mud concentration ratio data. The feature extraction module 40 stores the mud concentration ratio data along with a unified timestamp generated by the system global clock in the internal data buffer array. This data synchronization and coordination mechanism quantifies the external dosing physical process in time sequence and establishes the basic material boundary conditions for the macroscopic change of mud rheological properties.

[0056] During this synergistic dosing phase, the feature extraction module 40 is also responsible for dynamically adjusting the dosage of the agent based on the phase evolution rate. Specifically, the feature extraction module 40 receives the mud rheological phase index and performs discrete-time differentiation to obtain the phase evolution gradient. The feature extraction module 40 records the timestamp of the phase evolution gradient reaching the preset extreme point and calculates the difference between it and the dosing start timetamp. This difference is defined as the diffusion characteristic time constant of the agent in the current flow field. This time constant physically characterizes the spatial mixing response hysteresis of the mud network structure to the agent. The feature extraction module 40 adjusts the operating frequency of the dosing pump inversely according to this diffusion characteristic time constant, generates the corresponding dosing pump time duty cycle command through the built-in mapping function and issues it, thereby ensuring the uniformity of the overall flocculation reaction while avoiding local oversaturation of the agent.

[0057] As a preferred approach, the feature extraction module 40 periodically sends macroscopic transient speed step commands to the drive execution module 10 during the shutdown interval corresponding to the above-mentioned dosing pump time duty cycle command, at a preset time interval of 30 to 60 seconds. To prevent dynamic response aliasing caused by multi-source speed command, the feature extraction module 40 sends a forced suspension signal to the phase state analysis module 30 before sending the step command to temporarily cut off the superposition injection of low-frequency sine wave components.

[0058] The amplitude range of the macroscopic transient speed step command is set between 10% and 20% of the current steady-state base speed. The feature extraction module 40 synchronously collects high-frequency data during the forced speed jump executed by the drive execution module 10 and cancels the forced suspension signal after the step steady-state response ends to restore the perturbation injection. The stator active current transient data and rotor speed feedback data corresponding to the variable frequency drive side are used to extract the basic physical principle of the mechanical torque characteristics of the drive side, which is the interaction between the air gap magnetic field of the motor and the rotor current. Based on this principle, the feature extraction module 40 calls the pole pair number and flux linkage parameters preset by the motor as electromechanical energy conversion constants and linearly and equivalently converts the collected stator active current transient data into the transient electromagnetic torque data output by the drive shaft.

[0059] The feature extraction module 40 uses a differential algorithm to perform discrete-time differentiation on the acquired rotor speed feedback data to generate the real-time transient angular acceleration of the system. The time interval between adjacent sampling points in the differential calculation is equal to the basic sampling period of the underlying hardware. Since mathematical differential operation has high-pass filtering characteristics in the discrete domain and amplifies the high-frequency quantization noise in the digital acquisition circuit, the feature extraction module 40 is cascaded with a digital low-pass filter with a cutoff frequency set to 10 Hz to 20 Hz after the differential calculation link. The transient angular acceleration data generated after bandpass differential calculation and smoothing filtering presents the dynamic physical state of the stirring blade overcoming fluid resistance and completing the speed increase in the mud system. This transient angular acceleration data, together with the transient electromagnetic torque data obtained by the aforementioned calculation, is sent by the feature extraction module 40 to the built-in dynamic calculation unit for macroscopic parameter identification.

[0060] The feature extraction module 40 calculates the equivalent macroscopic moment of inertia under the current operating condition based on the dynamic balance relationship between the output torque and the load resistance torque. The calculation formula is as follows: ; In the formula, Represents the equivalent macroscopic moment of inertia; Represents a discrete sequence of transient electromagnetic torque; This represents the basic load resistance torque before the speed jump; The transient angular acceleration represents the value after filtering; the feature extraction module 40 performs the division on the denominator before the division is executed. An absolute value threshold check is performed. This absolute value threshold is set based on the minimum measured value calculated by the physical resolution of the system encoder. When the absolute value of angular acceleration is lower than this lower limit, the feature extraction module 40 skips the current calculation cycle and stops updating.

[0061] To ensure the representativeness of the selected parameters, the feature extraction module 40 only extracts the rising edge dynamic window of the rotational speed from the start of the jump to 90% of the target rotational speed. It performs an averaging operation on the equivalent macroscopic rotational inertia of all calculation cycles within this window and latches it as the final identification result of this step test. This mechanism avoids the deadlock caused by the denominator approaching zero, which leads to system numerical calculation exceeding limits, and parameter drift caused by steady-state tail noise. The feature extraction module 40 subtracts the calculated average of the equivalent macroscopic rotational inertia from the baseline steady-state inertia under the initial no-load state of the system. This baseline steady-state inertia is pre-determined and stored in the read-only memory area during the clean water no-load calibration operation. The subtraction algebraic operation physically removes the mechanical distributed inertia inherent in the pure mechanical transmission components. The equivalent macroscopic rotational inertia obtained after baseline stripping is defined by the feature extraction module 40 as a macroscopic mud viscosity index characterizing the fluid state. The feature extraction module 40 packages and fuses the macroscopic consistency index of the mud along with the concentration ratio data of the slurry obtained in the collaborative process according to the unified timestamp of the system. The macroscopic state feature data group after the data frame is packaged and fused is written into the system shared memory area by the feature extraction module 40 and pushed to the optimization decision module 50 in real time as the core state input basis for multidimensional control decision.

[0062] See attached document Figure 5 After receiving the feature data stream synchronously pushed by the front-end, the optimization decision module 50 triggers a cyclic decision mechanism. In this embodiment, the optimization decision module 50 extracts the mud flow change phase index output by the phase state analysis module 30 and the mud macro-consistency index and slurry concentration ratio data output by the feature extraction module 40 in real time. These three sets of data are aligned on the time axis by a unified system clock and spliced ​​together to form a single-dimensional real-time multi-dimensional state input vector. Before sending the three-dimensional vector into the nonlinear fusion calculation link, the optimization decision module 50 calls the front-end logic to perform the input data timeliness verification operation. When the optimization decision module 50 detects that any dimension of the input vector presents a null value or a non-numerical abnormal form, it triggers the alarm register and calls the historical cache data of the previous cycle for forced replacement operation. This front-end data replacement logic avoids the system deadlock caused by non-numerical overflow in network matrix operation due to packet loss in the underlying transmission.

[0063] As a preferred approach, the optimization decision module 50 internally constructs a three-layer feedforward neural network architecture to perform nonlinear phase space mapping on the preprocessed three-dimensional state input vector. This neural network model, in terms of hierarchical topology, includes an input layer with three input nodes, a hidden layer with sixteen single-layer operation nodes, and an output layer with a single output node. After the three-dimensional state input vector is buffered and distributed by the input layer, it is fully connected to the hidden layer weight matrix for multiplication and addition operations and then input into the modified linear unit activation function. This process extracts the comprehensive nonlinear variation law of various physical parameters exhibited by mud flocculation. The sixteen-dimensional tensor output after the hidden layer is activated is transmitted to the output layer and processed by the log-probability activation function to be mapped to a continuous numerical range from zero to one. This continuous floating-point value represents the probability assessment value of the completion degree of the current mud flocculation and solidification reaction at the system physical level.

[0064] Before embedded deployment, the feedforward neural network inside the optimization decision module 50 undergoes an offline parameter training process relying on an external computing platform. The sample dataset required for network training comes from the multi-dimensional sensor synchronous time-series records saved in multiple batches of historical dosing and mixing operations. Based on the qualified benchmark measured by the laboratory offline rheometer, the system labels the physical solidification state corresponding to each time-series node as a binary label. The points that reach the solidification shear standard are marked as one, and the remaining initial and intermediate transition states are marked as zero. The training process calls the adaptive moment estimation algorithm to perform backpropagation iteration and uses the binary cross-entropy loss function in the parameter configuration to quantify the distribution deviation between the predicted output and the true label. After multiple rounds of batch gradient descent and the loss value of the independent validation set converges and stabilizes, the neural network model stops updating parameters and the final connection weights are solidified in non-volatile memory.

[0065] The optimization decision module 50 executes the dosing shutdown judgment logic based on the completion probability assessment value output by the feedforward neural network. The internal register of the optimization decision module 50 is configured with a sliding judgment window and a shutdown trigger judgment threshold. The data length of the sliding judgment window is set to ten to fifty sampling points based on the system sampling period and the time span required for solidification steady state confirmation. The shutdown trigger judgment threshold is fixed in the range of 0.95 to 0.98 based on the safety margin required by the process. When the optimization decision module 50 polls and monitors that the probability assessment value output by the neural network model in real time continuously fills the entire sliding judgment window and all sample extraction values ​​in the window are higher than the shutdown trigger judgment threshold, a system flocculation completion flag is generated. This smooth judgment mechanism based on the time series window filters out single calculation jumps caused by the flow of local mud clumps through the detection area and outputs a steady state control decision conclusion.

[0066] After generating the system flocculation completion flag, the optimization decision module 50 synchronously sends a shutdown cutoff command to the drive execution module 10 and the external dosing control unit via the field industrial bus. The discrete mathematical equivalent expression of its comprehensive time window judgment decision operation is as follows: ; In the formula, The output state variable representing the comprehensive shutdown decision is determined if and only if this summed variable... The value is equal to the length of the sliding window. When the system determines that the status meets the conditions, it will initiate a shutdown action. The data length parameter represents the sliding window's decision window; This represents the probability evaluation sequence of completion outputs for the current node and historical nodes within the sliding window range; This represents the set threshold parameter for triggering a shutdown. This is the global index for the current discrete sampling node; It represents a discrete unit step function and outputs one when the independent variable is greater than or equal to zero, otherwise it outputs zero. The optimization decision module 50 blocks the power supply circuit of the dosing pump based on the above triggering conditions and coordinates the control of the frequency converter to execute ramp deceleration to zero speed.

[0067] See attached document Figure 6 The cell wall breaking control module 60 is in a background parallel monitoring state during the normal dosing and stirring process of the system. In this embodiment, the cell wall breaking control module 60 receives the stator active current transient data and rotor speed feedback data transmitted back by the drive execution module 10 in real time, and linearly converts the stator active current transient data into transient electromagnetic torque data. The cell wall breaking control module 60 has a nonlinear state observer based on the mechanical motion equation of the motor built inside. The observer predicts the ideal rotational speed at the next discrete time node based on the transient electromagnetic torque data of the current cycle, the equivalent macroscopic moment of inertia obtained from the pre-process identification, and the basic load resistance torque of the previous steady-state cycle. Its core physical mechanism lies in quantifying the sudden nonlinear adhesion disturbance encountered by the stirring blades in the slurry by comparing the deviation between the output of the ideal model and the feedback from the actual sensor. The wall-breaking control module 60 calculates the discrete mathematical expression of the rotational speed prediction residual as follows: ; In the formula, The speed prediction residual represents the current sampling period; The ideal predicted rotational speed represents the output of the nonlinear state observer; This represents the actual rotor speed feedback data collected. Representing the discrete time node index, this speed prediction residual physically characterizes the transient torque distortion physical quantity caused by the phenomenon of clumps wrapping around the blades or gripping the shaft within the flow field.

[0068] The cell wall breaking control module 60 performs absolute value processing on the calculated rotational speed prediction residual and sends it to the anomaly discrimination device. As a preferred approach, the cell wall breaking control module 60 adopts an adaptive dynamic threshold generation mechanism based on local statistical characteristics to replace the solidified constant parameter to accommodate the rheological physical process of gradually increasing mud concentration. The cell wall breaking control module 60 extracts the moving average and standard deviation of the rotational speed prediction residual in real time within a continuous sampling period and generates an adaptive dynamic discrimination threshold through linear combination. The combination calculation formula is as follows: ; In the formula, For the generated adaptive dynamic discrimination threshold; This is the moving average of the residual sequence; The standard deviation of the residual sequence; The weighting coefficient is fixed in the range of 2.0 to 3.0 based on the anti-disturbance sensitivity requirement. When the cell wall breaking control module 60 detects that the absolute value of the residual in the current cycle exceeds the adaptive dynamic discrimination threshold and continues to exceed the time tolerance set to 0.5 seconds to 1.5 seconds, the nonlinear anomaly flag is activated. This adaptive threshold mechanism based on sliding statistics avoids the judgment deadlock caused by the normal rise in mud concentration leading to an increase in background disturbance and thus false triggering of the fixed threshold.

[0069] After capturing the nonlinear anomaly flag, the cell wall breaking control module 60 immediately takes over the underlying variable frequency torque command channel of the drive execution module 10. The cell wall breaking control module 60 generates an alternating pulse torque command sequence with asymmetric amplitude according to the built-in strategy library and injects it directly into the torque feedforward loop of the variable frequency drive. The alternating pulse torque command contains a periodic alternating superposition of high-frequency positive impact torque and low-frequency reverse drag torque. The alternating pulse torque command is transmitted to the stirring impeller through the mechanical shaft of the motor rotor and acts directly on the attached mud agglomerate network structure. The asymmetric transient shear stress distribution forcibly tears the pseudo-solidified mud layer and the adhered boundary layer surrounding the blades in a physical mechanism. During the execution of the alternating pulse torque command, the underlying logic of the variable frequency drive forcibly suspends the conventional closed-loop speed control regulator operation node. This suspension blocking mechanism isolates the integral saturation of the control loop caused by abnormal speed feedback and outputs high-fidelity pulse torque.

[0070] The cell wall breaking control module 60 adopts an execution strategy of alternating intermittent pulse impact and closed-loop detection. After the cell wall breaking control module 60 issues an alternating pulse torque command with a single duration set to 3 to 5 seconds, it briefly restores the closed-loop speed control regulator of the frequency converter and makes it run under the steady-state base speed command to perform detection window sampling.

[0071] Within the detection window, the cell wall breaking control module 60 recalculates the speed prediction residual. If the residual is still higher than the threshold, it triggers the next round of alternating pulse torque injection. If the speed prediction residual is detected to fall below 50% of the adaptive dynamic discrimination threshold, this residual fall indicates that the nonlinear adhesion resistance around the stirring blades has been decoupled and the system flow field has returned to a macroscopically uniform suspension mixing phase. The cell wall breaking control module 60 then cancels the alternating pulse torque command and smoothly transfers the underlying control, allowing the frequency converter to resume executing the steady-state base speed command set by the front-end module. After exiting the intervention state, the cell wall breaking control module 60 synchronously activates the internally configured forced cooling timer. The timer's running time parameter is set based on three to five times the dominant time constant of the mechanical transmission chain. During this shielded cooling cycle, the cell wall breaking control module 60 ignores all new abnormal judgment requests.

[0072] Meanwhile, to prevent high-frequency nonlinear oscillation data generated during the cell breaking process from contaminating the upper-level optimization decision-making algorithm, the cell breaking control module 60 continuously sends a data sampling shielding countdown signal to the phase state analysis module 30 and the feature extraction module 40 throughout the entire cycle from the intervention start point to the end of the cooling timer. During the period when the shielding signal is received, the modules 30 and 40 forcibly suspend their respective data buffer queue updates and feature extraction operations. This mechanism prevents the transient mechanical adjustment fluctuations when the system just resumes normal speed from crossing the discrimination threshold, thereby causing the risk of infinite cell breaking and algorithm dead zones due to neural network misjudgment.

[0073] To aid in understanding the technical solution of this invention, a specific application example of an environmentally friendly treatment project for mud slurry from large-diameter cast-in-place piles of a cross-river bridge pier is provided below.

[0074] The construction of cast-in-place piles for cross-river bridge piers generates waste mud with high water content and high adhesion. The traditional treatment method is to add flocculant at fixed times and in fixed quantities and continuously stir. This process results in waste of flocculant, uneven stirring, and mud clumps adhering to the stirring shaft, which in turn causes motor overload damage. To solve the above technical problems, the adaptive control system of this invention is introduced.

[0075] The system starts up and pumps in unmodified pure mud. The baseline decoupling module controls the frequency converter to enable the AC asynchronous stirring motor to perform step operation at multiple discrete speeds. The system synchronously records the pure mechanical friction damping characteristic parameters of the gears and bearings. Based on these mechanical friction damping characteristic parameters, a two-dimensional lookup table is generated inside the system and solidified into a feedforward decoupling matrix.

[0076] The dosing pump begins injecting flocculant. The phase state analysis module sends a perturbation angular velocity command with superimposed low-frequency sinusoidal waveform to the frequency converter driver. The system extracts the response delay time of the stator active current transient data to the perturbation command to determine the mud rheological phase state index. The feature extraction module synchronously calculates the time constant of the agent diffusion characteristic and adjusts the output of the dosing pump to the pulse duty cycle mode. During the dosing pump stop interval, the system sends a macroscopic transient speed step command to the frequency converter driver. The equivalent macroscopic moment of inertia of the mud is calculated by extracting transient angular acceleration and transient electromagnetic torque.

[0077] When semi-solidified mud clumps adhere to the stirring blades, causing the rotor speed feedback data to fall below the ideal predicted speed output by the nonlinear state observer, the wall-breaking control module detects that the speed prediction residual exceeds the adaptive dynamic discrimination threshold (as shown in the attached figure). Figure 8 As shown in the figure, this illustrates the system's monitoring mechanism for the real-time sequence of speed prediction residuals. When localized mud adhesion or shaft seizure causes the residual amplitude to exceed the adaptive dynamic discrimination boundary, the system accurately captures the anomaly point and triggers the nonlinear anomaly flag (the working principle of the system). It then interrupts the conventional chemical dosing and mixing process. The cell-wall breaking control module sends an alternating pulse torque command sequence containing high-frequency positive impact torque and low-frequency reverse drag torque to the frequency converter driver. The asymmetric transient shear stress generated by the alternating pulse torque peels off the solidified mud layer encasing the blades. After detecting that the speed prediction residual has fallen back to the safe baseline, the system cancels the alternating pulse torque command and restores the steady-state base speed command (as shown in the attached figure). Figure 7 As shown in the figure, when the rotor speed drops abnormally due to the obstruction of the stirring shaft, the system forcibly tears the solidified mud layer by injecting an alternating electromagnetic torque pulse at an instantaneous high frequency, and then automatically restores the speed to the steady-state target speed without stopping. The global dynamic anti-disturbance response process is as follows: the cell wall breaking control module simultaneously activates the forced cooling timer to ignore new abnormal judgment requests within the nonlinear oscillation cycle.

[0078] The optimization decision-making module performs calculations and judgments by fusing the received mud flow phase state index, equivalent macroscopic rotational inertia, and slurry concentration ratio data through a feedforward neural network. When the completion probability assessment value output by the neural network continuously fills the sliding judgment window and exceeds the shutdown trigger judgment threshold, the system determines that the mud modification and dewatering operation has been completed and issues a deceleration shutdown and discharge command to the frequency converter driver and the dosing pump (as attached). Figure 9 As shown in the figure, this diagram illustrates the evolution trajectory of the completion probability assessment value output by the feedforward neural network as the dosing and solidification process gradually increases, and the timing control logic that the system immediately cuts off the normal process and precisely triggers the deceleration and shutdown discharge action when the assessment value continuously meets the two conditions of filling the sliding judgment window and exceeding the shutdown trigger judgment threshold.

[0079] Two sets of treatment equipment were configured at the same construction site to conduct a parallel comparative experiment on the treatment of mud from cast-in-place piles with a single batch volume of five cubic meters and a total of fifty batches. The control group was operated by a logic controller based on a set constant frequency stirring and a set constant time ratio for adding chemicals, while the experimental group was operated by an adaptive control system that included baseline decoupling, phase analysis, neural network optimization and cell wall breaking control logic.

[0080] Table 1: Comparison of Overall Operational Efficiency and Energy Consumption

[0081] Table 2: Verification of Specific Functions of the Cell Wall Breaking Control Module

[0082] Based on the data in Tables 1 and 2, the comparative experimental data confirms that the adaptive control system uses multi-dimensional state space fusion and feedforward neural network decision control to stop the timing of chemical dosing. This control logic prevents excessive stirring from damaging the mud flocculation structure and reduces the consumption of modifying agents and system running time. The nonlinear anomaly identification and alternating pulse torque injection commands of the cell wall breaking control module resolve the stirring shaft seizure caused by mud agglomerate adhesion. The above functional unit modules work together to ensure the continuous operation stability of the mud dewatering and modification treatment equipment.

[0083] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive control system for the mixing parameters of grouting pile mud modification and dewatering, characterized in that, include: The drive execution module is used to send operation control commands to the AC asynchronous stirring motor and collect transient data of stator active current and rotor speed feedback data. The baseline decoupling module is used to construct a two-dimensional lookup table based on the stator active current transient data and the rotor speed feedback data to generate a feedforward decoupling matrix; The phase state analysis module is used to extract the transient total phase difference based on the stator active current transient data and the rotor speed feedback data, and to calculate the mud rheological phase state index in combination with the feedforward decoupling matrix; The feature extraction module is used to calculate the equivalent macroscopic moment of inertia based on the transient data of the stator active current and the feedback data of the rotor speed, generate a dosing control command based on the mud rheological phase state index, and obtain the slurry concentration ratio data. The optimization decision module is used to send stop dosing and discharge control commands based on the mud rheological phase state index, the equivalent macroscopic rotational inertia and the slurry concentration ratio data. The cell wall breaking control module is used to calculate the speed prediction residual in real time based on the stator active current transient data, the rotor speed feedback data and the equivalent macroscopic moment of inertia, and to issue an alternating pulse torque command to the drive execution module when the speed prediction residual exceeds the adaptive dynamic discrimination threshold.

2. The adaptive control system for the mixing parameters of the grouting pile mud modification and dewatering according to claim 1, characterized in that, The drive execution module is internally configured with an industrial controller and a variable frequency drive. The variable frequency drive collects stator phase current, decouples it to extract stator active current transient data, and uploads the stator active current transient data and rotor speed feedback data to the industrial controller. The industrial controller calls a digital low-pass filter to perform anti-aliasing processing on the stator active current transient data and rotor speed feedback data. When the industrial controller issues the operation control command to the variable frequency drive, it records a first system timestamp. When it receives the stator active current transient data and rotor speed feedback data, it records a second system timestamp. The industrial controller calculates the difference between the first system timestamp and the second system timestamp as the total link transmission delay and converts it into a step size index to perform timing synchronization alignment.

3. The adaptive control system for the mixing parameters of the grouting pile mud modification and dewatering according to claim 1, characterized in that, The baseline decoupling module issues a frequency sweep test command containing multiple step frequency points to the drive execution module and maintains a steady-state dwell time at the step frequency points. The baseline decoupling module extracts data arrays within a fixed time window, accumulates and averages them to obtain the mean steady-state current and the mean base speed. The baseline decoupling module performs a discrete Fourier transform based on the mean steady-state current and the mean base speed to calculate the mechanical inherent phase lag. The baseline decoupling module constructs the two-dimensional lookup table with the mean base speed as the horizontal axis index and the mean steady-state current as the vertical axis index. The calculated mechanical inherent phase lag is used as a node value to fill the two-dimensional lookup table to generate the feedforward decoupling matrix.

4. The adaptive control system for the mixing parameters of the grouting pile mud modification and dewatering according to claim 1, characterized in that, The phase state analysis module sends a perturbation angular velocity command to the drive execution module. The phase state analysis module calls the orthogonal demodulation operation unit to perform orthogonal demodulation calculation to extract the transient total phase difference. The phase state analysis module performs a subtraction algebra operation on the transient total phase difference and the mechanical inherent phase lag amount queried by calling the feedforward decoupling matrix to obtain the rheological phase difference parameter. The phase state analysis module performs normalization calculation on the rheological phase difference parameter to generate a dimensionless characteristic index, and outputs the mud rheological phase state index accordingly.

5. The adaptive control system for the mixing parameters of the grouting pile mud modification and dewatering according to claim 4, characterized in that, The phase analysis module monitors the numerical change gradient of the transient total phase difference output by adjacent calculation cycles in real time. When the phase analysis module detects that the absolute value of the phase jump between two adjacent discrete time points exceeds a preset threshold, it triggers a phase unwinding processing mechanism. By adding or subtracting a preset period integer multiple to the transient total phase difference at the current abnormal crossover point, the transient total phase difference is pulled back to the continuous physical change trajectory range.

6. The adaptive control system for the mixing parameters of the grouting pile mud modification and dewatering according to claim 1, characterized in that, The feature extraction module receives the mud rheological phase state index and performs discrete-time derivative operation to obtain the phase state evolution gradient. The feature extraction module calculates the difference between the time stamp of the phase state evolution gradient reaching the extreme point and the time stamp of the dosing start and defines it as the diffusion characteristic time constant. The feature extraction module adjusts the operating frequency of the dosing pump inversely according to the diffusion characteristic time constant to generate the dosing pump time duty cycle instruction as the dosing control instruction.

7. The adaptive control system for the mixing parameters of the grouting pile mud modification and dewatering according to claim 6, characterized in that, The feature extraction module sends a macroscopic transient speed step command to the drive execution module during the stop interval corresponding to the dosing pump time duty cycle command. The feature extraction module linearly and equivalently converts the stator active current transient data into transient electromagnetic torque data, and uses a differential algorithm to perform discrete-time differentiation on the rotor speed feedback data to generate transient angular acceleration. Based on the transient electromagnetic torque data and the transient angular acceleration, the feature extraction module calculates the equivalent macroscopic moment of inertia, extracts the rising edge dynamic window, performs an averaging operation, and latches it.

8. The adaptive control system for the mixing parameters of the grouting pile mud modification and dewatering according to claim 1, characterized in that, The optimization decision module extracts the mud rheological phase state index, the equivalent macroscopic moment of inertia, and the slurry concentration ratio data to form a three-dimensional state input vector. The optimization decision module sends the three-dimensional state input vector into a three-layer feedforward neural network architecture to map and output a completion probability evaluation value. The internal register of the optimization decision module is configured with a sliding judgment window and a shutdown trigger judgment threshold. When it is detected that the completion probability evaluation value fills the sliding judgment window and all sample extraction values ​​in the window are higher than the shutdown trigger judgment threshold, a system flocculation completion flag is generated and the stop dosing and discharge control command is sent.

9. The adaptive control system for the mixing parameters of the grouting pile mud modification and dewatering according to claim 1, characterized in that, The cell-breaking control module internally constructs a nonlinear state observer, which converts the transient stator active current data returned by the drive execution module into transient electromagnetic torque data. Based on the transient electromagnetic torque data, the equivalent macroscopic moment of inertia, and the extracted basic load resistance torque, the ideal predicted speed is predicted. The cell-breaking control module calculates the deviation between the ideal predicted speed and the rotor speed feedback data to obtain the speed prediction residual. The cell-breaking control module extracts the moving average and standard deviation of the speed prediction residual within a continuous sampling period and performs a linear combination to generate the adaptive dynamic discrimination threshold.

10. The adaptive control system for the mixing parameters of the grouting pile mud modification and dewatering according to claim 9, characterized in that, After the speed prediction residual exceeds the adaptive dynamic discrimination threshold and the nonlinear anomaly flag is captured, the cell breaking control module generates the alternating pulse torque command and injects it into the torque feedforward loop of the drive execution module. The cell breaking control module adopts an alternating execution strategy of intermittent pulse impact and closed-loop detection. When it detects that the speed prediction residual has fallen back to below the preset proportional baseline of the adaptive dynamic discrimination threshold, it cancels the alternating pulse torque command and simultaneously activates the internally configured forced cooling timer.

Images