Method, device and system for controlling the tension of water supply and drainage pipeline robots against the inner wall

By constructing a real-time mileage coordinate system and a neural network model, abnormal pipe wall conditions are detected in real time, and pressure compensation bias is generated. This solves the adaptability problem of existing tension control methods under complex working conditions and improves the passability and stability of pipeline robots.

CN122331641BActive Publication Date: 2026-11-17NINGBO WATER ENVIRONMENT ENGINEERING CONSTRUCTION CO LTD OPERATION & MAINTENANCE BRANCH
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202610444093.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-04-07
Publication Date
2026-11-17
Estimated Expiration
2046-04-07

AI Technical Summary

Technical Problem

Existing tension control methods for pipeline robots are difficult to adapt to complex working conditions such as sudden changes in pipe wall friction coefficient, local protrusions, or deposits, leading to drive wheel slippage, insufficient climbing ability, or motor overload, which affects the robot's travel stability and detection accuracy.

Method used

By constructing a real-time mileage coordinate system and combining the operating parameters of the guide wheel and follower wheel, abnormal characteristics of the pipe wall status are detected in real time. Abnormal information is stored using a spatiotemporal memory matrix. By combining a piecewise continuous mapping function and a radial basis function neural network model, a pressure compensation bias is generated to achieve precise adjustment of the tension of the follower wheel.

Benefits of technology

It significantly shortens the response time to local pipe wall anomalies, improves the passability and operational stability of pipeline robots under non-uniform pipe wall conditions, and enhances their anti-disturbance capabilities.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122331641B_ABST
    Figure CN122331641B_ABST
Patent Text Reader

Abstract

This invention discloses a method, device, and system for controlling the tension of a pipeline robot against its inner wall, relating to the field of pipeline robot technology. The method includes: acquiring operating parameters; dividing the drive wheels into guide wheels and follower wheels; constructing a real-time mileage coordinate system; calculating the difference between the drive motor current of the guide wheel and the stable reference current; if the difference exceeds an anomaly detection threshold, locking the current mileage coordinate as an anomaly spatial node; extracting anomaly features of the pipe wall state and storing them in a spatiotemporal memory matrix; tracking the mileage coordinates of the follower wheels; extracting features when reaching an anomaly spatial node and calculating a pressure compensation bias using a piecewise continuous mapping function; inputting the operating parameters of the follower wheels into a neural network model to obtain a basic target pressure; superimposing this pressure with the pressure compensation bias to obtain the target control pressure; and driving a servo motor adjustment mechanism to adjust the tension. This invention achieves precise tension adjustment, enhancing the operational stability of the pipeline robot under non-uniform pipe wall conditions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of pipeline robot technology, and more specifically, to a method, device, and system for controlling the tension force of a water supply and drainage pipeline robot against the inner wall. Background Technology

[0002] In the inspection and maintenance of water supply and drainage pipelines, pipeline robots need to carry inspection equipment and move inside the pipe wall. To ensure the robot's passability and stability, a suitable normal tension must be maintained between the drive wheel and the inner wall of the pipe: if the tension is too small, the drive wheel will slip and the climbing ability will be insufficient; if the tension is too large, it may cause motor overload, mechanical wear, or even damage to the inner wall of the pipe.

[0003] Existing pipeline robots mostly use constant pressure setting or feedback adjustment based on the current wheel state for tension control. The constant pressure method is difficult to adapt to complex working conditions such as sudden changes in the friction coefficient of the pipe wall, local bulges or deposits. The feedback adjustment method only responds after the drive wheel has entered the abnormal area, which has a lag and is prone to instantaneous slippage or jamming, affecting the robot's travel stability and detection accuracy.

[0004] To address the above problems, this invention proposes a solution. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method, device and system for controlling the tension of a robot closely adhering to the inner wall of a water supply and drainage pipeline, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for controlling the tension of a water supply and drainage pipeline robot against the inner wall includes the following steps: obtaining the operating parameters of the pipeline robot at the current moment; dividing the multiple drive wheels of the pipeline robot into guide wheels and follow wheels according to the physical arrangement of the travel direction; based on the integral of the travel speed over time in the operating parameters, combined with the constant axle distance between the front and rear wheels along the pipeline axis between the guide wheels and the follow wheels and the cosine value of the robot's real-time attitude tilt angle, constructing a real-time mileage coordinate system, and obtaining the real-time mileage coordinates of the guide wheels and the follow wheels respectively;

[0008] The difference between the drive motor current of the guide wheel and the preset guide wheel steady-state reference current is calculated. When the absolute value of the difference exceeds the preset anomaly detection threshold, the current mileage coordinate of the guide wheel is locked as an anomaly space node, the difference is extracted as a pipe wall state anomaly feature, and the mapping relationship between the mileage coordinate of the anomaly space node and the corresponding pipe wall state anomaly feature is stored in the spatiotemporal memory matrix. Among them, when the pipe wall state anomaly feature is positive, it represents a slippage anomaly; when the pipe wall state anomaly feature is negative, it represents a stagnation anomaly.

[0009] Track the real-time mileage coordinates of the following wheel; when it is determined that the following wheel has reached the mileage coordinates of an abnormal spatial node, extract the corresponding abnormal pipe wall state feature from the spatiotemporal memory matrix, input it into the preset piecewise continuous mapping function, and calculate the pressure compensation bias.

[0010] The operating parameters of the follower wheel are input into a pre-trained neural network model to obtain the basic target pressure; the basic target pressure is algebraically superimposed with the pressure compensation bias to obtain the target control pressure; and sent to the controller to drive the servo motor adjustment mechanism corresponding to the follower wheel to adjust the tension of the follower wheel on the inner wall of the pipe.

[0011] In a preferred embodiment, the operating parameters include the travel speed obtained by an encoder mounted on the drive wheel shaft, the attitude tilt angle obtained by an inertial measurement unit mounted at the center of the robot body, and the armature current value obtained by sampling resistors connected in series in the power supply circuit of each drive motor; wherein the constant wheelbase is a fixed distance along the pipe axis between the axis plane of the guide wheel set and the axis plane of the follow wheel set; the real-time mileage coordinates of the guide wheel are obtained by accumulating the product of the travel speed and the sampling period period by sampling period; the real-time mileage coordinates of the follow wheel are obtained by subtracting the product of the constant wheelbase and the cosine of the current attitude tilt angle from the real-time mileage coordinates of the guide wheel.

[0012] In a preferred embodiment, the method for determining the mileage coordinates of the following wheel reaching the abnormal space node is as follows: a coordinate matching tolerance is introduced, which is the product of the maximum designed travel speed of the pipeline robot and the sampling period; when the absolute value of the difference between the real-time mileage coordinates of the following wheel and the mileage coordinates of the abnormal space node is not greater than the coordinate matching tolerance, it is determined that the following wheel has reached the abnormal space node; the abnormal space node that has triggered the matching is marked as consumed in the current forward travel process.

[0013] In a preferred embodiment, the piecewise continuous mapping function uses a preset dead zone filtering threshold as the segment boundary and includes the following three output rules:

[0014] When the absolute value of the abnormal pipe wall condition characteristic is not greater than the dead zone filtering threshold, the output pressure compensation bias is zero.

[0015] When the abnormal characteristic quantity of the pipe wall condition is greater than the dead zone filtration threshold, the output positive pressure compensation bias is proportional to the part exceeding the dead zone filtration threshold, and is constrained by the preset positive compensation upper limit value, that is, it does not exceed the preset positive compensation upper limit value.

[0016] When the abnormal characteristic quantity of the pipe wall condition is less than the negative dead zone filtering threshold, the output is a negative pressure compensation bias that is proportional to the portion below the negative dead zone filtering threshold, and is constrained by a preset negative compensation lower limit value, i.e., it is not lower than the preset negative compensation lower limit value.

[0017] In a preferred embodiment, the positive pressure compensation bias is calculated as follows: the smaller of the product of the preset positive compensation gain coefficient and the difference between the abnormal pipe wall condition characteristic quantity and the dead zone filtering threshold, and the upper limit of the positive compensation.

[0018] The negative pressure compensation bias is calculated as follows: take the larger of the product of the preset negative compensation gain coefficient, the sum of the abnormal pipe wall condition characteristic quantity and the dead zone filtering threshold, and the negative compensation lower limit value; where the positive compensation gain coefficient represents the pressure compensation amount corresponding to the slippage anomaly, and the negative compensation gain coefficient represents the pressure release amount corresponding to the stagnation anomaly, and the positive compensation gain coefficient and the negative compensation gain coefficient take different values.

[0019] In a preferred embodiment, the neural network model is a radial basis function neural network model, and its hidden layers use Gaussian radial basis functions as activation functions. The input vector of the radial basis function neural network model includes the travel speed, attitude tilt angle, follower wheel motor current, and follower wheel motor torque obtained by multiplying the follower wheel motor current and the motor torque constant. The output of the radial basis function neural network model is used to characterize the future pressure fitting curve of the follower wheel in a preset future time domain that is consistent with the control period after the current sampling time. The basic target pressure is calculated from the future pressure fitting curve according to the target control time. The basic target pressure calculation logic is as follows: the target control time is obtained by adding the current sampling time to the preset control period, and the pressure value of the future pressure fitting curve at the target control time is taken as the basic target pressure. The control period is an integer multiple of the sampling period.

[0020] In a preferred embodiment, after receiving the target control pressure, the controller calculates the difference between the target control pressure and the measured pressure obtained by the pressure sensor; it calculates the control quantity through the feedforward channel and the feedback channel respectively, and superimposes them to form a composite control command; in the feedforward channel, the feedforward position increment is the product of the preset feedforward gain coefficient and the difference between the target control pressure and the measured value; in the feedback channel, the PID controller calculates the fine-tuning correction amount based on the error between the target control pressure and the measured value; the composite control command for the servo motor is the sum of the current position command value, the feedforward position increment, and the fine-tuning correction amount.

[0021] In a preferred embodiment, the guide wheel stable reference current is obtained through a reference calibration procedure: in a standard pipe sample with a known uniform wall condition, the pipe robot is driven to travel at a predetermined calibration distance at a rated working speed, the time series of the guide wheel motor current is continuously recorded, and the arithmetic mean of the time series is taken. The result is used as the guide wheel stable reference current; and the abnormal detection threshold is greater than the standard fluctuation range of the guide wheel motor current under normal pipe wall conditions.

[0022] The tension control device for the water supply and drainage pipeline robot that closely adheres to the inner wall includes the following modules: a data acquisition coordinate construction module, which is used to obtain the operating parameters of the pipeline robot at the current moment, distinguish between the guide wheel and the follow wheel, construct a real-time mileage coordinate system based on the integral of the travel speed over time and the constant axle distance between the front and rear wheels, and output the real-time mileage coordinates of the guide wheel and the follow wheel.

[0023] An anomaly detection memory mapping module is used to calculate the difference between the drive motor current of the guide wheel and the stable reference current of the guide wheel. When the absolute value of the difference exceeds the anomaly detection threshold, the current mileage coordinate of the guide wheel is locked as an anomaly space node and the abnormal feature quantity of the pipe wall state is extracted. The mapping relationship between the mileage coordinate of the anomaly space node and the corresponding abnormal feature quantity of the pipe wall state is stored in the spatiotemporal memory matrix.

[0024] The coordinate alignment offset calculation module is used to track the real-time mileage coordinates of the following wheel. When it is determined that the following wheel has reached the mileage coordinates of the abnormal space node, the corresponding pipe wall state abnormality feature quantity is extracted from the spatiotemporal memory matrix, and the pressure compensation offset is calculated through the piecewise continuous mapping function.

[0025] The pressure fusion control execution module has a pre-trained radial basis function neural network model embedded in it. It outputs a basic target pressure based on the running parameters of the follower wheel, algebraically superimposes the basic target pressure with the pressure compensation bias to obtain the target control pressure, and encapsulates the target control pressure into a standard control command frame and sends it to the servo drive controller. This drives the servo motor adjustment mechanism corresponding to the follower wheel to precisely adjust the normal tension of the follower wheel on the inner wall of the pipe.

[0026] A tension control system for a robot that closely adheres to the inner wall of a water supply and drainage pipeline, characterized in that it includes:

[0027] The robot body is equipped with guide wheels and follower wheels along the direction of travel. Each drive wheel is tightly attached to the inner wall of the pipe through an independent tensioning mechanism. The robot body includes a servo actuator and a cantilever arm. The servo actuator consists of a servo motor and a nut-screw transmission pair, which is used to drive the cantilever arm to generate radial linear displacement, thereby adjusting the tension of the drive wheels on the inner wall of the pipe.

[0028] The sensor group includes an encoder mounted on the drive wheel shaft to acquire the travel speed, an inertial measurement unit mounted in the center of the robot body to acquire the attitude tilt angle, a sampling resistor connected in series in the power supply circuit of each drive motor to acquire the armature current value, and a pressure sensor mounted on the drive wheel suspension arm to acquire the measured pressure.

[0029] A tension control device for the water supply and drainage pipeline robot, which is in close contact with the inner wall, is communicatively connected to the sensor group and each of the tensioning mechanisms, and is used to implement the tension control method for the water supply and drainage pipeline robot in close contact with the inner wall.

[0030] The technical effects and advantages of the present invention regarding the method, device, and system for controlling the tension of a robot closely adhering to the inner wall of a water supply and drainage pipeline are as follows:

[0031] This invention associates the abnormal pipe wall condition features detected by the guide wheel with mileage coordinates and stores them in a spatiotemporal memory matrix. When the following wheel reaches the same spatial node, the feature values ​​are extracted and a pressure compensation bias is generated. This achieves a shift from passive feedback to spatial predictive feedforward adjustment mode, significantly shortening the response time to local pipe wall anomalies. A piecewise continuous mapping function is used to generate compensation values ​​with opposite directions and adjustable gains for slippage and stagnation anomalies, making tension adjustment more targeted and adaptable. Combined with the algebraic superposition of the basic target pressure output by the radial basis function neural network and the compensation bias, the invention effectively improves the anti-disturbance capability under abrupt changes while ensuring control accuracy under normal operating conditions, significantly enhancing the pipeline robot's passability and operational stability under non-uniform pipe wall conditions. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the process for controlling the tension of the robot in close contact with the inner wall of the water supply and drainage pipeline according to the present invention.

[0033] Figure 2 This is a schematic diagram of the internal structure of the tension control system for the water supply and drainage pipeline robot that adheres closely to the inner wall, as per the present invention.

[0034] Figure 3 This is a schematic diagram of the tension control device for the water supply and drainage pipeline robot that closely adheres to the inner wall according to the present invention.

[0035] Figure 4 This is a schematic diagram illustrating the prediction principle of the neural network model of the present invention;

[0036] Figure 5 This is a simplified schematic diagram of the wheel assembly arrangement under the longitudinal section view of the pipeline used in this invention. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] Example: Please refer to Figure 1 As shown, this invention discloses a method for controlling the tension of a robot tightly adhering to the inner wall of a water supply and drainage pipeline, comprising the following steps:

[0039] Step 1: Obtain the operating parameters of the pipeline robot at the current moment; divide the multiple drive wheels of the pipeline robot into guide wheels and follow wheels according to the physical arrangement of the direction of travel; based on the integral of the travel speed with respect to time in the operating parameters, combined with the constant wheelbase between the guide wheels and the follow wheels along the pipeline axis and the cosine value of the robot's real-time attitude tilt angle, construct a real-time mileage coordinate system, and obtain the real-time mileage coordinates of the guide wheels and the follow wheels respectively;

[0040] After the pipeline robot starts up inside the pipeline, its onboard data acquisition system runs continuously at a preset fixed frequency; the sensor sampling frequency is defined as... In this embodiment, The initial value is set to 200Hz, corresponding to the sampling period. It is the reciprocal of the sensor sampling frequency, i.e. The sampling period determines the time resolution for the system to acquire external information, and all subsequent calculations based on time integration use this period as the basic step size. At the beginning of each sampling period, the data acquisition system synchronously performs the following three types of parameter reading operations.

[0041] The first type of parameter is the real-time travel speed of the pipeline robot. The unit is m / s; this speed is obtained by a photoelectric incremental encoder mounted on the drive wheel shaft. The photoelectric incremental encoder is a sensor that measures angular velocity and angular displacement by detecting the number of pulses generated during shaft rotation, enabling non-contact real-time detection of the drive wheel's rotational speed. The total number of pulses generated per revolution of the encoder is defined as the encoder line count. In this embodiment, The initial value is set to 1024 pulses / rev; the outer diameter of the drive wheel is defined as the drive wheel diameter. In this embodiment, the diameter of the drive wheel The initial value is set to 150mm, or 0.15m; in the first... Within each sampling period, the encoder pulse increment recorded by the system counter is speed of travel The calculation method is as follows: the arc travel of the drive wheel within one sampling period is equal to the ratio of the pulse increment to the encoder line count, multiplied by the circumference of the drive wheel, and finally divided by the sampling period, i.e.: ;

[0042] Taking numerical values ​​as an example, if the encoder count increments within a certain sampling period... ,but This speed value represents the real-time travel status of the pipeline robot at the current sampling moment.

[0043] The second type of parameter is the real-time attitude tilt angle of the pipeline robot. The unit is radians (rad). This tilt angle is calculated by a six-axis inertial measurement unit (IMU) installed at the center of the robot body. Physically, it represents the pitch angle between the robot's longitudinal axis and the horizontal plane. The IMU integrates a three-axis accelerometer and a three-axis gyroscope sensor module, which can calculate motion parameters such as the vehicle's attitude angle and angular velocity. In this embodiment, the IMU outputs attitude angle data after internal Kalman filtering once per sampling cycle, and the system directly reads this pitch angle value as... When the robot travels in a horizontal pipe section, the pitch angle value Approaching 0; when the robot is on an uphill pipe section, the pitch angle value The value is positive; when the robot is on a downhill pipe section, the pitch angle value is... It is a negative value;

[0044] The third type of parameter is the real-time armature current value of each drive motor. The unit is ampere (A). The pipeline robot is equipped with multiple independently driven drive wheels, each driven by a DC servo motor. A precision sampling resistor is connected in series in the power supply circuit of each motor. The system obtains the instantaneous armature current flowing through the motor by measuring the voltage drop across the resistor and dividing it by the resistance value. The precision sampling resistor is a resistor with a low temperature coefficient and high resistance value, used to convert the motor armature current into a measurable voltage signal.

[0045] The direction in which the pipeline robot travels along the pipeline axis is defined as the positive direction. Observing the robot's physical structure along the positive direction, the pipeline robot body is equipped with several drive wheels along the pipeline's travel direction, and each drive wheel is tightly attached to the inner wall of the pipeline through its own independent tensioning mechanism; please refer to... Figure 5This invention uses a simplified wheel arrangement diagram shown in a longitudinal sectional view of a pipe. It illustrates two sets of drive wheels. Based on the physical arrangement of each drive wheel on the robot body along the direction of travel, the drive wheel located at the front (i.e., the side that first contacts the pipe wall along the direction of travel) is defined as the guide wheel; the drive wheel located at the rear (i.e., the side that lags behind the guide wheel in reaching the same pipe wall position along the direction of travel) is defined as the follower wheel. The fixed distance along the pipe axis between the axial plane of the guide wheel set and the axial plane of the follower wheel set is defined as the constant wheelbase between the front and rear wheels. This value is uniquely determined by the robot's mechanical structure and remains constant throughout the robot's operation; in this embodiment, The initial value is set to 350mm, or 0.35m; the tensioning mechanism includes a servo actuator and a suspension arm; the servo actuator consists of a servo motor and a nut-screw transmission pair, used to drive the suspension arm to generate radial linear displacement, thereby adjusting the tension force of the drive wheel on the inner wall of the pipe;

[0046] Based on the above role definition, the drive motor current of the guide wheel is independently labeled as the guide wheel motor current. The drive motor current of the following wheel is independently labeled as the following wheel motor current. At each sampling time The system also outputs the guide wheel parameter set. and following wheel parameter set Among them, speed and attitude angle For the overall motion parameters of the robot, the two parameter sets share these two components; while the current... and Each reflects the independent mechanical interaction state between its respective drive wheel and the pipe wall;

[0047] Furthermore, in order to transform all sensor data from a purely temporal sequence into information with spatial positioning capabilities, this step constructs a one-dimensional relative mileage coordinate system with the pipeline inlet as the origin; the coordinate axes of this coordinate system extend along the pipeline axis, the unit of the coordinate values ​​is meters, and the physical meaning is the cumulative travel distance of a specific point on the pipeline robot relative to the pipeline inlet.

[0048] Real-time odometer coordinates of the guide wheel By measuring the speed of travel It is obtained by performing discrete-time integration; the specific calculation method is as follows: multiply the velocity value at the current sampling time by the sampling period, and add the resulting product to the mileage coordinate value at the previous sampling time. The mathematical expression is shown below: ;in, The previous sampling time The odometer coordinates of the guide wheel;

[0049] At system startup At that time, the guide wheel's odometer coordinates are initialized to zero, that is... This indicates that the guide wheel is currently located at the pipe inlet;

[0050] Follow wheel's real-time odometer coordinates Instead of independent integration calculations, the distance is directly obtained by subtracting the constant axial projection distance of the front and rear wheel axles from the mileage coordinates of the guide wheel. This calculation relationship is derived from rigid body kinematic constraints and considers the influence of the robot's tilt angle on the axial projection of the wheel axle: at the same moment, the position difference between the front and rear wheel sets along the pipe axis is always equal to the projection value of their fixed installation distance on the robot body along the pipe axis. The mathematical expression can be seen as follows: ;in, This represents the robot's real-time tilt angle at the current sampling moment;

[0051] The final output data of this step includes three sets: the real-time odometer coordinates of the guide wheel at the current moment. Real-time mileage coordinates following the wheel at the current moment. and the guide wheel parameter set and following wheel parameter set The above data is refreshed once at the end of each sampling period and synchronously transmitted to steps two and three as input.

[0052] Step 2: Calculate the difference between the drive motor current of the guide wheel and the preset guide wheel stable reference current; when the absolute value of the difference exceeds the preset abnormal detection threshold, lock the current mileage coordinate of the guide wheel as an abnormal space node, extract the difference as the pipe wall state abnormal feature quantity, and store the mapping relationship between the mileage coordinate of the abnormal space node and the corresponding pipe wall state abnormal feature quantity into the spatiotemporal memory matrix.

[0053] Determining abnormal pipe wall conditions requires a clear benchmark. Therefore, before the pipeline robot officially enters the pipeline to perform inspection tasks, the system performs a benchmark calibration procedure. The calibration process is as follows: in a standard pipe sample with a known uniform wall condition and no obvious defects, the robot is driven to travel a calibration distance at a constant speed at its rated working speed. In this embodiment, the initial value of the calibration distance is set to 3m. During this calibration journey, the system continuously records the current of the guide wheel motor. The time series was calculated, and the arithmetic mean of the time series was taken. The result was defined as the guide wheel steady-state reference current. This parameter characterizes the steady-state current consumption level of the guide wheel when traveling at rated speed under normal wall conditions; in this embodiment, after calibration... The initial value is 2.5A; Once calibrated, it is written into the system parameter storage area and remains unchanged throughout the entire detection task cycle;

[0054] After the pipeline robot enters the pipeline to be inspected and begins its formal movement, the system performs the following difference calculation operation in each sampling period; retrieves the current sampling time. current of the guide wheel motor To stabilize the reference current of the guide wheel Given the minuend, calculate the algebraic difference between the two. ,Right now ;

[0055] The difference The physical meaning is as follows: when the guide wheel passes through the area where the friction coefficient of the pipe wall is normal, the current of the guide wheel motor is... Approximate guide wheel smooth reference current Difference The absolute value approaches zero;

[0056] When the guide wheel passes through an area where the friction coefficient of the pipe wall is significantly reduced, such as when the pipe wall surface is slippery due to grease, water film, or biofilm, the frictional resistance on the drive wheel decreases, and the current required for the motor to maintain the same speed decreases. Difference It is a positive value;

[0057] When the guide wheel passes through areas where there are local geometric abrupt changes in the pipe wall, such as local protrusions, deposit adhesion, or a reduction in pipe diameter that increase the traveling resistance, the motor outputs a larger current to overcome the additional resistance torque. Difference It is negative;

[0058] Then the difference The absolute value is compared with a preset anomaly detection threshold; the anomaly detection threshold is defined as... This threshold is used to distinguish between minute fluctuations caused by sensor noise under normal operating conditions and meaningful deviations caused by actual changes in the pipe wall condition. The numerical setting is based on the following: it should be greater than the standard fluctuation range of the guide wheel motor current under normal pipe wall conditions; in this embodiment, through statistical analysis of the current sequence during the calibration phase, the guide wheel motor current under normal operating conditions... The standard deviation is approximately 0.15A, based on which... The initial value is set to the sum of three times the standard deviation and an engineering margin. This value ensures that 99.7% of normal fluctuations will not be misjudged as abnormal, while retaining the ability to detect deviations greater than 0.5A.

[0059] The criteria for determining abnormal pipe wall condition are as follows: when the guide wheel reaches a specific mileage coordinate... The sampling time, that is, there exists a certain sampling time. Make If the absolute value of the difference calculated at that moment Greater than the anomaly detection threshold The system will then perform a coordinate locking operation, locking the mileage coordinates. Mark it as an anomalous spatial node;

[0060] At mileage coordinates At the same moment a node is locked as an anomalous spatial node, the system will trigger a coordinate lock on the corresponding difference. We directly extract and assign explicit physical labels to them, defining them as anomalous features of pipe wall condition. That is, its mathematical expression is: Among them, the abnormal characteristic quantity of pipe wall condition It is a signed scalar value whose sign carries information about the type of pipe wall anomaly, specifically:

[0061] When the pipe wall condition is abnormal characteristic quantity Characterizes slippage-type anomalies, such as wall lubrication and low friction;

[0062] When the pipe wall condition is abnormal characteristic quantity Characterizing obstructive anomalies, such as wall roughness, protrusions, and deposits; and characteristic quantities of abnormal pipe wall conditions. The absolute value carries information about the degree of abnormality: anomalous characteristic quantity of pipe wall condition. The larger the absolute value, the more serious the deviation of the pipe wall condition from the normal baseline at that mileage coordinate.

[0063] Furthermore, during the initialization phase, the system maintains a dedicated data storage area in memory, defined as a spatiotemporal memory matrix. The matrix's data structure is a list of key-value pairs, with each record containing two fields: the first field is the mileage coordinate key. Its spatiotemporal memory matrix The first field is a floating-point number in meters (m), and the second field is the abnormal characteristic value of the pipe wall condition. Its spatiotemporal memory matrix The value in the middle is a floating-point number, in units of A; the spacetime memory matrix. Each coordinate key A single feature value corresponds to a unique feature value. Under the same coordinate key, multiple different feature values ​​are not stored. If the same position is triggered multiple times, the value of the later one will overwrite the value of the previous one.

[0064] Once a new anomalous spatial node is locked, the system immediately checks the spatiotemporal memory matrix. If a new record is inserted, the mapping relationship between the mileage coordinates and the abnormal characteristic values ​​of the pipe wall condition is expressed as follows: ;

[0065] Take a specific scenario as an example: a pipeline robot travels within a pipeline to the guide wheel's mileage coordinates. At that time, Then the stored record is ;

[0066] As the robot continues to move, the guide wheel may trigger multiple anomaly detections at different mileage coordinates; each trigger generates an independent record; for example, the guide wheel at... Detected at the location That is, the slippery type, in Detected at the location That is, the blocking type, in Detected at the location For severe slippage, the spacetime memory matrix... Three records are stored in the middle: ;

[0067] The final output of this step is: a continuously updated spatiotemporal memory matrix. This matrix stores all marked abnormal spatial nodes from the pipe inlet to the current position of the guide wheel and their corresponding abnormal pipe wall status features. This matrix is ​​stored in the system's shared memory for real-time querying in step three. The computational logic of this step is executed once in each sampling period, synchronously with the data acquisition in step one, without consuming additional waiting time.

[0068] Step 3: Track the real-time mileage coordinates of the following wheel; when it is determined that the following wheel has reached the mileage coordinates of the abnormal space node, extract the corresponding abnormal pipe wall state feature from the spatiotemporal memory matrix, input it into the preset piecewise continuous mapping function, and calculate the pressure compensation bias.

[0069] Within each sampling period, the system acquires the real-time mileage coordinates of the following wheel from step one. The coordinate value monotonically increases as the robot continues to move, meaning the system analyzes situations where the robot is moving forward and not backward; the system will... The current value and the spatiotemporal memory matrix The coordinate keys of all abnormal space nodes stored in the system are compared one by one to determine whether the following wheel has reached any marked abnormal position of the pipe wall.

[0070] Because the data acquisition system operates with discrete sampling periods, the mileage coordinates of the following wheel change by a finite number of steps between two adjacent sampling times. Therefore, the coordinate value of the following wheel in the sampling sequence may not be exactly equal to a certain coordinate key stored in the spatiotemporal memory matrix. To address this discretization problem, the system introduces a coordinate matching tolerance parameter. Define coordinate matching tolerance This represents the maximum distance the following wheel can travel within one sampling period; it is calculated based on the robot's maximum designed travel speed. With sampling period The product of In this embodiment, the exemplary pipeline robot's maximum design travel speed is... The initial value is 0.6 m / s, therefore That is, 3mm;

[0071] The rules for determining mileage coordinate alignment are as follows: For the spatiotemporal memory matrix A coordinate key in If at the current sampling time The absolute value of the difference between the following wheel's mileage coordinate and the coordinate key must not exceed the coordinate matching tolerance. The system then determines that the following wheel has reached the abnormal space node. The physical location is determined, triggering the analysis program to proceed to the next step; this determination condition applies to the matrix in each sampling period. All unconsumed coordinate keys in the array are traversed and checked; once a coordinate key is found... Matching was triggered, and the record was marked as consumed during this forward movement to prevent the follow wheel from triggering it repeatedly in subsequent sampling cycles because the coordinates are still within the tolerance range. The consumption mark is valid for a single forward movement. When the robot detects reverse movement, i.e., the encoder pulse increment is negative or it restarts from the pipe inlet, all consumed marks in the spatiotemporal memory matrix are automatically cleared, and the abnormal spatial nodes are restored to the unconsumed state, which can trigger coordinate matching and compensation again.

[0072] After the analysis program is triggered, the system retrieves data from the spatiotemporal memory matrix. Extract the coordinates that match the current position. Abnormal characteristics of the bound pipe wall condition Continuing from the example, the extracted result is: The system uses this feature as the sole independent variable and inputs it into a predefined piecewise continuous mapping function. The function converts the current-dimensional characteristic quantity, which characterizes the degree of pipe wall abnormality, into the pressure-dimensional bias quantity, which characterizes the mechanical compensation requirement.

[0073] Piecewise continuous mapping function The output is a pressure-compensated bias. The unit is Newtons (N); the function consists of three consecutively joined intervals, each interval corresponding to a processing rule for a type of pipe wall condition; the function design requires the introduction of the following preset parameters;

[0074] Piecewise continuous mapping function Filtering threshold with a dead zone As a segment boundary, the dead zone filtering threshold is used to shield sensor noise or minor disturbances in characteristic quantities. When the absolute value of an abnormal characteristic quantity of the pipe wall condition is not greater than this threshold, no pressure compensation bias is generated. In this embodiment, the dead zone filtering threshold... The initial value is set to 0.3A; this value is less than the anomaly detection threshold in step two. Therefore, all features filtered in step two and stored in the spatiotemporal memory matrix have absolute values ​​greater than 1. After entering the mapping function, all functions can cross the dead zone and produce non-zero compensated bias outputs. The existence of the dead zone makes the function have a flat zero output region near the origin, which ensures the robustness of the function.

[0075] Then piecewise continuous mapping function The segmented output is defined as follows:

[0076] When the absolute value of the abnormal pipe wall condition characteristic is not greater than the dead zone filtering threshold, the system determines that the pipe wall condition deviation corresponding to the abnormal pipe wall condition characteristic is a minor disturbance, insufficient to produce a perceptible mechanical effect, and the mapping function output pressure compensation bias is zero. When, piecewise continuous mapping function Entering the dead zone and outputting pressure compensation bias. ;

[0077] When the abnormal characteristic quantity of the pipe wall condition exceeds the dead zone filtering threshold, it indicates that there is a risk of reduced friction coefficient of the pipe wall at that coordinate and possible slippage and misalignment of the drive wheel; piecewise continuous mapping function The output is a positive pressure compensation bias that is proportional to the portion of the abnormal pipe wall condition characteristics that exceeds the dead zone filtering threshold, and is constrained by the upper limit of the positive compensation, i.e.: When, piecewise continuous mapping function Enter the positive compensation phase and output pressure compensation bias. ;in, To preset the upper limit of positive compensation, This is the positive compensation gain coefficient, expressed in N / A, representing the pressure compensation amount corresponding to each unit of positive pipe wall abnormality characteristic, i.e., slippage-type abnormality; in this embodiment... The initial value is set to 15.0 N / A; the physical basis of this coefficient is: when the current of the guide wheel is 1 A lower than the reference, an additional normal pressure of about 15 N is required to compensate for the lack of friction when the follow wheel reaches the same position.

[0078] When the abnormal characteristic quantity of the pipe wall condition is less than the negative dead zone filtering threshold, it indicates that there is a protrusion, deposit, or blockage in the pipe wall at that coordinate, resulting in increased travel resistance; piecewise continuous mapping function The output is a negative pressure compensation bias that is proportional to the portion of the pipe wall condition abnormality characteristic quantity that is below the negative dead zone filtering threshold, and is constrained by the negative compensation lower limit, i.e.: When, piecewise continuous mapping function Enter the negative compensation section and output pressure compensation bias. ;in, This is the lower limit of negative compensation. This is the negative compensation gain coefficient, expressed in N / A, representing the pressure release per unit of negative pipe wall abnormality characteristic, i.e., the pressure release corresponding to the stagnation-type abnormality; in this embodiment... The initial value is set to 12.0 N / A; and Different values ​​are chosen because the impact risk caused by the protrusion of the pipe wall to the wheel body is higher than that caused by insufficient friction. The gain of pressure release is deliberately set to be slightly lower than the gain of pressure increase, so as to maintain a certain wall adhesion force while reducing the impact risk.

[0079] It should be noted that, in this embodiment, the initial value of the positive compensation upper limit is... The initial value of the negative compensation lower limit is These two parameters saturate and limit the output of the mapping function to prevent the compensation bias from exceeding the physical adjustment capability of the servo mechanism or causing structural overload due to abnormal characteristics of extreme pipe wall conditions.

[0080] The final output of this step is: the coordinates of the abnormal space node reached by the currently following wheel. One-to-one pressure compensation bias If the following wheel fails to reach any marked abnormal spatial nodes within the current sampling period (i.e., the coordinate alignment judgment fails), this step will not trigger the analysis procedure. Keeping it at zero, step four is controlled solely based on the neural network's baseline predictions.

[0081] Please see Figure 4The schematic diagram of the prediction principle of the neural network model of the present invention is shown. Step 4: Input the running parameters of the following wheel into the pre-trained neural network model to obtain the basic target pressure; algebraically superimpose the basic target pressure and the pressure compensation bias to obtain the target control pressure; and send it to the controller to drive the servo motor adjustment mechanism corresponding to the following wheel to adjust the tension of the following wheel on the inner wall of the pipe.

[0082] At the same sampling moment when the coordinate alignment determination is completed in step three. The system extracts the following wheel's coordinates at the current mileage from the output of step one. Complete set of operating parameters To match the input dimension of the basic pressure prediction system of this invention, the system also needs to calculate the output torque of the follower wheel drive motor. The motor torque constant is defined as follows: The unit is Nm / A, and this value is provided by the motor manufacturer. In this embodiment... The initial value is 0.8 Nm / A; the output torque of the follower motor is calculated as the product of the motor torque constant and the follower motor current. ;

[0083] The above parameters are combined into an input vector. ; input vector The first neural network model, which has been pre-trained, is fed into the system. In this embodiment, the model is a Radial Basis Function Neural Network (RBFNN), whose network structure has been determined and solidified during the offline training phase. Only forward computation is performed during online operation. The offline training process of the RBFNN is as follows: The training dataset consists of operating parameters collected from the pipeline robot under five typical working conditions: horizontal, uphill (e.g., θ=30° / 60°), downhill (θ=-30° / -60°), slight slippage, and slight obstruction. The sample size for each working condition is no less than 1000 groups. The dataset is divided into training, validation, and test sets in a 7:2:1 ratio. The mean squared error (MSE) is used as the loss function, and the convergence threshold is set to... Training is considered convergent when the loss function value on the training set is below the convergence threshold for 100 consecutive iterations and the loss on the validation set does not increase significantly; the coefficient of determination on the model's test set... To ensure that the model's prediction accuracy meets engineering requirements, the structural parameters of this RBFNN are as follows:

[0084] The input layer contains four nodes, each corresponding to a different travel speed. Attitude and tilt angle Follow wheel motor current and following wheel motor torque Define the number of hidden layer nodes as In this embodiment The initial value is set to 20; the hidden layer uses a Gaussian radial basis function as the activation function; the first... The center vector of each hidden layer node is And it is a four-dimensional vector, and , define the first The width parameter of each hidden layer node is And it is a positive real scalar; the central vector and width parameter The number of clusters is determined during the offline training phase using the K-means clustering algorithm and nearest neighbor distance calculation; for example, the number of clusters is set to... After clustering convergence Each cluster center serves as the center vector for each neuron; the width parameter Take the first The Euclidean distance between each center vector and its nearest neighbor center vector multiplied by the width scaling factor The value after that, The initial value is ;

[0085] Hidden layer Each node pairs the input vector The response output is as follows: The square of the Euclidean distance between the input vector and the center vector of the node is calculated, divided by twice the square of the node's width parameter, and the negative of this value is used to calculate the natural exponential function value, which can be obtained through the following formula: ;

[0086] The output layer contains one node; define the first... The connection weights from each hidden layer node to the output layer node are: Define the output layer bias as The network output is a weighted linear combination of the response values ​​of all hidden layer nodes plus a bias term. This output is used to characterize the future pressure fitting curve of the following wheel in the preset future time domain after the current sampling time, which is consistent with the control period. The control system calculates the basic target pressure of the following wheel at the current coordinate from the future pressure fitting curve according to the target control time. That is, its mathematical expression is: Among them, connection weights and bias The solution is obtained through regularized least squares during the offline training phase; the regularization coefficient is defined as... In this embodiment The initial value is set to 0.01 to suppress overfitting. After training, the weights and biases are stored in the system's non-volatile memory and can be directly loaded and used during online runtime.

[0087] The mapping relationship learned by the RBFNN model can be physically explained through an approximate interpretable linear expression: the model essentially performs an optimized nonlinear fusion of the positive contribution of travel speed to anti-skid pressure, the compensation of the attitude tilt angle sine value to the normal component of gravity, the correlation of the current square term to the electromagnetic load, and the response of the torque term to external resistance; taking the trained model in this embodiment as an example, statistical regression analysis on a large number of input-output data pairs yields the following approximate relationship:

[0088] ;

[0089] In the formula, k1 is the preset speed anti-slip weighting coefficient, which represents the positive contribution of the travel speed to the normal pressure required to maintain the wheel wall friction. k1 is the preset slope gravity compensation coefficient, which characterizes the magnitude of the compensation of the attitude tilt angle to the normal component of gravity along the pipe wall, in N; k2 is the preset electromagnetic load correlation coefficient, which characterizes the degree of correlation between the electromagnetic load intensity reflected by the square term of the motor current and the wheel wall pressure demand, in N / A²; k3 is the preset resistance torque response coefficient, which characterizes the influence weight of the external travel resistance reflected by the motor output torque on the normal pressure demand, in N / Nm. is a static base pressure constant, which characterizes the minimum base normal pressure required for the pipeline robot to maintain reliable contact between the drive wheel and the pipe wall when all dynamic disturbance terms are zero in a static horizontal pipe section. The unit is N.

[0090] It should be noted that in this embodiment, a detection robot with a drive wheel diameter of 150mm, a total weight of 40kg, and suitable for drainage pipes with a diameter of 600mm is used as the object. The RBFNN model is trained using historical operating data collected by this robot under various typical pipeline conditions. Statistical regression analysis is then performed on a large amount of input-output data after training. The fitting results of each coefficient are as follows: the initial value of the speed anti-slip weight coefficient k1 is 8.2 N·s / m, the initial value of the slope gravity compensation coefficient k2 is 150 N, the initial value of the electromagnetic load correlation coefficient k3 is 0.05 N / A², the initial value of the resistance torque response coefficient k4 is 1.5 N / Nm, and the static foundation pressure constant... The initial value is 200N;

[0091] The physical meaning of the above coefficients can be interpreted as follows: k1 = 8.2 N·s / m indicates that for every 1 m / s increase in travel speed, the required pressure predicted by the model increases by approximately 8.2 N. This increment is used to compensate for the risk of dynamic slippage caused by the increase in speed; k2 = 150 N indicates that when the robot is in a vertical upward state, i.e., θ = 90°, At this time, the pressure predicted by the model needs to be increased by an additional 150N relative to the horizontal travel state to completely offset the component of gravity in the normal direction of the pipe wall; k3=0.05N / A² reflects the weak coupling relationship between the square term of the motor current and the Joule heat and electromagnetic force load; k4=1.5N / Nm indicates that for every 1Nm increase in the motor output torque, i.e., an increase of 1Nm in the external resistance torque, the pressure predicted by the model increases by 1.5N accordingly; C=200N is a constant pressure base independent of the operating state, ensuring that even under the most ideal working conditions, the drive wheel still maintains sufficient wall adhesion force;

[0092] The system obtains the basic target pressure. Then, call the pressure compensation bias output from step three. Perform a direct algebraic superposition operation on the two to generate the corrected target control pressure. ,Right now ;

[0093] If step three is not triggered within the current sampling period, i.e., the follow wheel does not reach any abnormal space node, then Target control pressure Directly equal to The control system relies on the basic predictions of the neural network for adjustment;

[0094] Target control pressure Encapsulated as standard control command frames, the data is transmitted via the onboard communication bus. In this embodiment, the CANopen protocol, a high-level communication protocol based on the CAN bus, is used. This protocol is suitable for real-time data transmission between industrial devices and supports standardized communication for servo drives, sensors, and other devices. The initial bus communication rate is set to 500kbps and sent to the dedicated servo drive controller of the following wheel. Each following wheel is equipped with an independent servo actuator, consisting of a servo motor and a nut-screw transmission pair. The rotation of the servo motor is converted into radial linear displacement of the drive wheel suspension arm through the nut-screw mechanism, thereby changing the normal force exerted by the drive wheel on the inner wall of the pipe. The transmission ratio of the nut-screw mechanism is defined as follows: In this embodiment The initial value is 2mm / rev, which means that the axial displacement of the screw is 2mm for every revolution of the servo motor.

[0095] The servo drive controller receives the target control pressure. Then, based on the target control pressure and the current measured value of the pressure sensor... The difference between them, combined with the feedforward gain coefficient Calculate the feedforward position increment of the servo motor; where the feedforward gain coefficient is... The physical meaning is the amount of telescoping displacement of the nut and screw mechanism corresponding to a unit pressure difference. In this embodiment... The initial value was set to 0.02 mm / N, which was obtained through a force-displacement calibration test on the tensioning mechanism. The calibration test method can be as follows: the drive wheel and the tensioning mechanism are fixed on a mechanical test bench, a normal pressure of 0~500N is applied by a digital push-pull force gauge, and the extension and retraction displacement of the corresponding nut and screw is collected by a laser displacement sensor. Each measurement point is a calibration point every 50N, and each measurement point is repeated 3 times and the average value is taken. The pressure-displacement data is linearly fitted, and the slope of the fitted line is the feedforward gain coefficient. ;

[0096] The feedforward position increment is calculated as the product of the feedforward gain coefficient and the difference between the target control pressure and the measured pressure. ;

[0097] In the feedback channel, which operates in parallel with the feedforward channel, the pressure value measured by the pressure sensor in the next sampling cycle will be compared with... The residual error is compared, and a fine-tuning correction is generated by a low-gain PID controller. This, combined with the feedforward increment, forms the final composite control command. The PID parameters for the feedback channel are: proportional gain. The initial value is Integral gain The initial value is Differential gain The initial value is In this embodiment, the initial value of the differential term is set to zero to avoid amplification of high-frequency noise; the function of the feedback channel is to eliminate the steady-state deviation caused by model errors and mechanical transmission clearances in the feedforward circuit, ensuring that the actual pressure eventually tracks the target value. ;

[0098] It should be noted that a PID controller is a closed-loop controller composed of proportional, integral, and derivative components. It achieves precise adjustment of the controlled object by performing proportional, integral, and derivative operations on the deviation.

[0099] The entire process, from generating the feedforward position command to executing it on the servo motor, is completed within one control cycle; the control cycle is defined as... In this embodiment The initial value is set to 0.01s, meaning the system updates the target control pressure and outputs a control command every 10 milliseconds; the control period is an integer multiple of the sensor sampling period, for example... Each control command update is based on the average value of sensor data from the two most recent sampling periods, ensuring that the operating parameters used in the control loop are the latest, stable, and valid data.

[0100] This completes the control loop; the pipeline robot will then proceed with its operation based on the sampling period. With the minimum time step, the entire process from step one to step four is continuously executed in a loop. The guide wheel continuously scans the state of the pipe wall ahead and writes abnormal information into the spatiotemporal memory matrix. When the follow wheel reaches each marked position, it extracts information from the matrix and generates a compensation bias. After being fused with the basic target pressure of the RBFNN, it drives the servo mechanism to precisely adjust the tension. This collaborative mechanism of guide wheel detection and follow wheel compensation transforms the tension adjustment of the follow wheel from a traditional passive feedback response to an active feedforward adjustment based on spatial prediction information. When a local change occurs in the pipe wall state, the response time of pressure adjustment can be advanced from the hundreds of milliseconds of traditional feedback control to the moment when the guide wheel passes through the abnormal area.

[0101] Please see Figure 3 As shown, the present invention discloses a tension control device for a water supply and drainage pipeline robot that adheres closely to the inner wall, comprising the following modules: a data acquisition coordinate construction module, used to acquire the operating parameters of the pipeline robot at the current moment, distinguish between the guide wheel and the follow wheel, construct a real-time mileage coordinate system based on the integral of the travel speed over time and the constant axle distance between the front and rear wheels, and output the real-time mileage coordinates of the guide wheel and the follow wheel.

[0102] An anomaly detection memory mapping module is used to calculate the difference between the drive motor current of the guide wheel and the stable reference current of the guide wheel. When the absolute value of the difference exceeds the anomaly detection threshold, the current mileage coordinate of the guide wheel is locked as an anomaly space node and the abnormal feature quantity of the pipe wall state is extracted. The mapping relationship between the mileage coordinate of the anomaly space node and the corresponding abnormal feature quantity of the pipe wall state is stored in the spatiotemporal memory matrix.

[0103] The coordinate alignment offset calculation module is used to track the real-time mileage coordinates of the following wheel. When it is determined that the following wheel has reached the mileage coordinates of the abnormal space node, the corresponding pipe wall state abnormality feature quantity is extracted from the spatiotemporal memory matrix, and the pressure compensation offset is calculated through the piecewise continuous mapping function.

[0104] The pressure fusion control execution module has a pre-trained radial basis function neural network model embedded in it. It outputs a basic target pressure based on the running parameters of the follower wheel, algebraically superimposes the basic target pressure with the pressure compensation bias to obtain the target control pressure, and encapsulates the target control pressure into a standard control command frame and sends it to the servo drive controller. This drives the servo motor adjustment mechanism corresponding to the follower wheel to precisely adjust the normal tension of the follower wheel on the inner wall of the pipe.

[0105] Please see Figure 2 As shown, the present invention further provides a tension control system for a water supply and drainage pipeline robot that adheres tightly to the inner wall. The system comprises three components at the hardware level: the robot body, the sensor group, and the tension control device for the water supply and drainage pipeline robot that adheres tightly to the inner wall.

[0106] The robot body serves as the physical carrier of the entire system. The robot body comprises a main body structure and several drive wheels positioned along the pipe's travel direction. According to their physical arrangement in the travel direction, the drive wheels at the front that first contact the pipe wall are called guide wheels, and the drive wheels at the rear that lag behind the guide wheels in reaching the same pipe wall position are called follower wheels. The fixed distance along the pipe axis between the axis of the guide wheel group and the axis of the follower wheel group is the constant wheelbase, uniquely determined by the robot's mechanical structure and remaining constant throughout the robot's operation. Each drive wheel is equipped with an independent tensioning mechanism, including a servo actuator and a suspension arm. The rotational motion of the servo motor in the servo actuator is converted into radial linear displacement of the suspension arm through a nut-screw transmission pair, thereby changing the radial position of the drive wheel against the inner wall of the pipe. Each drive wheel is driven by an independent DC servo motor, and the rotation drive and tension adjustment of each drive wheel are independent of each other.

[0107] The sensor array provides the system with multi-source heterogeneous real-time operating data. The sensor array includes four types of sensors: the first type is a photoelectric incremental encoder, mounted on the drive wheel shaft, which outputs pulse incremental signals reflecting the rotation of the drive wheels. This signal is processed by the data acquisition and coordinate construction module to obtain the real-time travel speed of the pipeline robot. The second type is a six-axis inertial measurement unit, mounted in the center of the robot body, integrating a three-axis accelerometer and a three-axis gyroscope. After internal Kalman filtering, it outputs the pitch angle between the robot's longitudinal axis and the horizontal plane, i.e., the real-time attitude tilt angle. The third type is a precision sampling resistor, connected in series in the power supply circuit of each drive motor. By measuring the voltage drop across the resistor and dividing by the resistance value, the instantaneous armature current value flowing through each motor is obtained. The motor currents of the guide wheel and follower wheel are collected independently. The fourth type is a pressure sensor, mounted on the suspension arm of each drive wheel, used to measure the normal contact pressure between the drive wheel and the inner wall of the pipe in real time, i.e., the measured pressure.

[0108] The tension control device for the water supply and drainage pipeline robot, which closely adheres to the inner wall, is integrated into the main controller of the pipeline robot. It establishes a bidirectional communication connection with the sensor group and each tensioning mechanism through an onboard communication bus. In the data input direction, the tension control device receives the travel speed, attitude tilt angle, armature current value of each drive wheel, and measured pressure output by the sensor group. In the control output direction, the tension control device sends the target control pressure command frame to the servo drive controller corresponding to each following wheel through the communication bus. The tension control device for the water supply and drainage pipeline robot, which closely adheres to the inner wall, completes all operations in sequence according to the method flow from step one to step four: collecting operating parameters and constructing mileage coordinates, detecting abnormal pipe wall conditions on the guide wheel side and storing them in the spatiotemporal memory matrix, determining coordinate alignment and calculating pressure compensation offset on the following wheel side, and fusing the basic target pressure prediction and target control pressure output. After receiving the target control pressure, the servo drive controller of each following wheel generates the position command of the servo motor through a composite control strategy that works in parallel with the feedforward channel and the feedback channel. This drives the corresponding nut and screw transmission pair to adjust the radial displacement of the suspension arm, thereby precisely adjusting the normal tension of the following wheel on the inner wall of the pipeline.

[0109] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0110] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0111] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0112] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0113] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0114] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for controlling the tension force of a water supply and drainage pipeline robot against the inner wall, characterized in that, Includes the following steps: Obtain the operating parameters of the pipeline robot at the current moment; divide the multiple drive wheels of the pipeline robot into guide wheels and follow wheels according to the physical arrangement of the direction of travel; based on the integral of the travel speed with respect to time in the operating parameters, combined with the constant front and rear wheel wheel distance along the pipeline axis between the guide wheels and the follow wheels and the cosine value of the robot's real-time attitude tilt angle, construct a real-time mileage coordinate system, and obtain the real-time mileage coordinates of the guide wheels and the follow wheels respectively. Calculate the difference between the drive motor current of the guide wheel and the preset guide wheel steady reference current; When the absolute value of the difference exceeds the preset anomaly detection threshold, the current mileage coordinates of the guide wheel are locked as an anomaly space node, the difference is extracted as a pipe wall state anomaly feature quantity, and the mapping relationship between the mileage coordinates of the anomaly space node and the corresponding pipe wall state anomaly feature quantity is stored in the spatiotemporal memory matrix; where, when the pipe wall state anomaly feature quantity is positive, it represents a slippage anomaly; when the pipe wall state anomaly feature quantity is negative, it represents a stagnation anomaly. The system tracks the real-time mileage coordinates of the following wheel; when the mileage coordinates of the following wheel reaching an abnormal spatial node are determined, the corresponding abnormal pipe wall state feature quantity is extracted from the spatiotemporal memory matrix and input into a preset piecewise continuous mapping function to calculate the pressure compensation bias; wherein, the preset piecewise continuous mapping function uses a preset dead zone filtering threshold as the segment boundary and includes the following three output rules: When the absolute value of the abnormal pipe wall condition characteristic is not greater than the dead zone filtering threshold, the output pressure compensation bias is zero. When the abnormal characteristic quantity of the pipe wall condition is greater than the dead zone filtration threshold, the output positive pressure compensation bias is proportional to the part exceeding the dead zone filtration threshold, and is constrained by the preset positive compensation upper limit value, that is, it does not exceed the preset positive compensation upper limit value. When the abnormal characteristic quantity of the pipe wall condition is less than the negative dead zone filtering threshold, the output is a negative pressure compensation bias that is proportional to the portion below the negative dead zone filtering threshold, and is constrained by a preset negative compensation lower limit value, i.e., not lower than the preset negative compensation lower limit value. The operating parameters of the follower wheel are input into a pre-trained neural network model to obtain the basic target pressure; the basic target pressure is algebraically superimposed with the pressure compensation bias to obtain the target control pressure; and sent to the controller to drive the servo motor adjustment mechanism corresponding to the follower wheel to adjust the tension of the follower wheel on the inner wall of the pipe.

2. The method for controlling the tension of a robot tightly adhering to the inner wall of a water supply and drainage pipeline according to claim 1, characterized in that, The operating parameters include the travel speed obtained by the encoder mounted on the drive wheel shaft, the attitude tilt angle obtained by the inertial measurement unit mounted in the center of the robot body, and the armature current value obtained by the sampling resistor connected in series in the power supply circuit of each drive motor; the constant wheelbase is the fixed distance between the axis plane of the guide wheel group and the axis plane of the follow wheel group along the pipe axis; the real-time mileage coordinates of the guide wheel are obtained by accumulating the product of the travel speed and the sampling period in each sampling period; the real-time mileage coordinates of the follow wheel are obtained by subtracting the product of the constant wheelbase and the cosine of the current attitude tilt angle from the real-time mileage coordinates of the guide wheel.

3. The method for controlling the tension of a robot closely adhering to the inner wall of a water supply and drainage pipeline according to claim 2, characterized in that, The method for determining the mileage coordinates of the following wheel reaching the abnormal space node is as follows: a coordinate matching tolerance is introduced, which is the product of the maximum design travel speed of the pipeline robot and the sampling period; when the absolute value of the difference between the real-time mileage coordinates of the following wheel and the mileage coordinates of the abnormal space node is not greater than the coordinate matching tolerance, it is determined that the following wheel has reached the abnormal space node. Abnormal space nodes that have triggered matching are marked as consumed during this forward movement.

4. The method for controlling the tension of a robot closely adhering to the inner wall of a water supply and drainage pipeline according to claim 3, characterized in that, The positive pressure compensation bias is calculated as follows: take the smaller of the product of the preset positive compensation gain coefficient and the difference between the abnormal pipe wall condition characteristic quantity and the dead zone filtering threshold, and the upper limit of the positive compensation. The negative pressure compensation bias is calculated as follows: take the larger of the product of the preset negative compensation gain coefficient, the sum of the abnormal pipe wall condition characteristic quantity and the dead zone filtering threshold, and the negative compensation lower limit value; where the positive compensation gain coefficient represents the pressure compensation amount corresponding to the slippage anomaly, and the negative compensation gain coefficient represents the pressure release amount corresponding to the stagnation anomaly, and the positive compensation gain coefficient and the negative compensation gain coefficient take different values.

5. The method for controlling the tension of a robot tightly adhering to the inner wall of a water supply and drainage pipeline according to claim 1, characterized in that, The neural network model is a radial basis function (RBF) neural network model, with Gaussian radial basis functions used as activation functions in its hidden layers. The input vector of the RBF neural network model includes the travel speed, attitude tilt angle, follower wheel motor current, and follower wheel motor torque obtained by multiplying the follower wheel motor current by the motor torque constant. The output of the RBF neural network model is used to characterize the future pressure fitting curve of the follower wheel in the preset future time domain after the current sampling time, which is consistent with the control period. The basic target pressure is calculated from the future pressure fitting curve based on the target control time. The basic target pressure calculation logic is as follows: the target control time is obtained by adding the current sampling time to the preset control period, and the pressure value of the future pressure fitting curve at the target control time is taken as the basic target pressure. The control period is an integer multiple of the sampling period.

6. The method for controlling the tension of a robot tightly adhering to the inner wall of a water supply and drainage pipeline according to claim 1, characterized in that, After receiving the target control pressure, the controller calculates the difference between the target control pressure and the measured pressure obtained by the pressure sensor; it calculates the control quantity through the feedforward channel and the feedback channel respectively, and superimposes them to form a composite control command; in the feedforward channel, the feedforward position increment is the product of the preset feedforward gain coefficient and the difference between the target control pressure and the measured value; in the feedback channel, the PID controller calculates the fine-tuning correction amount based on the error between the target control pressure and the measured value; the composite control command for the servo motor is the sum of the current position command value, the feedforward position increment, and the fine-tuning correction amount.

7. The method for controlling the tension of a robot closely adhering to the inner wall of a water supply and drainage pipeline according to claim 1, characterized in that, The guide wheel stable reference current is obtained through a reference calibration procedure: In a standard pipe sample with a known uniform wall condition, the pipe robot is driven to travel at a predetermined calibration distance at a rated working speed. The time series of the guide wheel motor current is continuously recorded and the arithmetic mean of the time series is taken. The result is used as the guide wheel stable reference current. The abnormal detection threshold is greater than the standard fluctuation range of the guide wheel motor current under normal pipe wall conditions.

8. A tension control device for a water supply and drainage pipeline robot, used to implement the tension control method for a water supply and drainage pipeline robot against the inner wall as described in any one of claims 1-7, characterized in that, It includes the following modules: a data acquisition coordinate construction module, which is used to obtain the operating parameters of the pipeline robot at the current moment, distinguish between the guide wheel and the follow wheel, construct a real-time mileage coordinate system based on the integral of the travel speed over time and the constant wheelbase of the front and rear wheels, and output the real-time mileage coordinates of the guide wheel and the follow wheel. An anomaly detection memory mapping module is used to calculate the difference between the drive motor current of the guide wheel and the stable reference current of the guide wheel. When the absolute value of the difference exceeds the anomaly detection threshold, the current mileage coordinate of the guide wheel is locked as an anomaly space node and the abnormal feature quantity of the pipe wall state is extracted. The mapping relationship between the mileage coordinate of the anomaly space node and the corresponding abnormal feature quantity of the pipe wall state is stored in the spatiotemporal memory matrix. The coordinate alignment offset calculation module is used to track the real-time mileage coordinates of the following wheel. When it is determined that the following wheel has reached the mileage coordinates of the abnormal space node, the corresponding pipe wall state abnormality feature quantity is extracted from the spatiotemporal memory matrix, and the pressure compensation offset is calculated through the piecewise continuous mapping function. The pressure fusion control execution module has a pre-trained radial basis function neural network model embedded in it. It outputs a basic target pressure based on the running parameters of the follower wheel, and then algebraically superimposes the basic target pressure with the pressure compensation bias to obtain the target control pressure. The target control pressure is then encapsulated into a standard control command frame and sent to the servo drive controller. This drives the servo motor adjustment mechanism corresponding to the follower wheel to precisely adjust the normal tension of the follower wheel on the inner wall of the pipe.

9. A tension control system for a water supply and drainage pipeline robot that closely adheres to the inner wall, characterized in that, include: The robot body is equipped with guide wheels and follower wheels along the direction of travel. Each drive wheel is tightly attached to the inner wall of the pipe through an independent tensioning mechanism. The robot body includes a servo actuator and a cantilever arm. The servo actuator consists of a servo motor and a nut-screw transmission pair, which is used to drive the cantilever arm to generate radial linear displacement, thereby adjusting the tension of the drive wheels on the inner wall of the pipe. The sensor group includes an encoder mounted on the drive wheel shaft to acquire the travel speed, an inertial measurement unit mounted in the center of the robot body to acquire the attitude tilt angle, a sampling resistor connected in series in the power supply circuit of each drive motor to acquire the armature current value, and a pressure sensor mounted on the drive wheel suspension arm to acquire the measured pressure. The water supply and drainage pipeline robot's tension control device is communicatively connected to the sensor group and each of the tensioning mechanisms, and is used to implement the water supply and drainage pipeline robot's tension control method against the inner wall as described in any one of claims 1 to 7.

Citation Information

Patent Citations

  • Magnetic traction type intelligent pipeline detection robot and method

    CN121539703A

  • KR20250032844A