A paving thickness control method and system for a paving machine

Abstract

The present application relates to the technical field of automatic control of road construction equipment, and particularly relates to a paving thickness control method and system for a paver, the method comprising: constructing a history window, constructing a mapping function of average driving speed and target lag parameter based on the history window; constructing a current window, calculating a target lag parameter and a peak confidence of the current window, obtaining a current fitting lag value according to the mapping function, when the peak confidence is greater than a preset threshold, updating an old residual error, taking a sum of the fitting lag value and the updated residual error as an adaptive lag value; when the peak confidence is less than or equal to the preset threshold, taking a sum of the fitting lag value and the old residual error as the adaptive lag value; applying the adaptive lag value to a Smith predictor and a PID controller, and outputting a control quantity. The present application can adaptively adjust, so that the control system can more accurately compensate for the control delay caused by lag, and improve the control precision of the paving thickness.

Description

Technical Field

[0001] This invention relates to the field of automatic control technology for road construction equipment. In particular, it relates to a method and system for controlling the paving thickness of a paver. Background Technology

[0002] During paver paving, there is a significant hysteresis between control parameters (such as the angle of attack of the screed or the height of the traction point) and the final pavement thickness. This means that when the control parameters are adjusted, the pavement thickness does not change immediately, but rather after a period of time. This hysteresis is affected by many factors, such as the paver's travel speed, the temperature and viscosity of the mixture, and the condition of the aggregate head; moreover, these factors are constantly changing during construction.

[0003] Existing methods for controlling paving thickness mainly include conventional PID control (Proportional-Integral-Derivative). Since PID controllers adjust the control quantity based on the current error, and the characteristic of pure time-delay systems is that the change of error is delayed, it is difficult for PID controllers to adjust the control quantity in a timely and accurate manner, which can easily lead to problems such as overshoot (paving too thickly) or slow recovery (the speed of adjusting the thickness is very slow). Summary of the Invention

[0004] To address the technical problems of delay mismatch, overshoot, slow recovery, and thickness tailing that easily occur in the above-mentioned paving thickness control methods, the present invention provides solutions in the following aspects.

[0005] In the first aspect, a method for controlling the paving thickness of a paver includes:

[0006] Collect historical construction data of the paver and divide the historical construction data into multiple historical windows according to the preset time length. Each historical window includes the control quantity time sequence, the travel speed time sequence, and the pavement thickness time sequence.

[0007] Define the range of hysteresis parameters, iterate through all hysteresis parameters based on a single historical window, find the hysteresis parameter corresponding to the peak value of the cross-correlation between the control quantity time series and the ply thickness time series in the historical window, and mark it as the target hysteresis parameter; construct a mapping function between the average driving speed and the target hysteresis parameter based on all historical windows;

[0008] Construct the current window with the current time as the endpoint, calculate the target lag parameter and peak confidence of the current window, obtain the fitted lag value corresponding to the current average driving speed according to the mapping function, and update the old residual when the peak confidence is greater than the preset threshold. The sum of the fitted lag value corresponding to the current window and the updated residual is used as the adaptive lag value; otherwise, the sum of the fitted lag value corresponding to the current window and the old residual is used as the adaptive lag value.

[0009] The final adaptive hysteresis value is applied to the Smith predictor to predict the ply thickness, generating the predicted ply thickness. The difference between the actual measured ply thickness and the predicted ply thickness is calculated as the prediction error. The prediction error is input into the PID controller, which outputs a control quantity and executes the paving action according to the control quantity.

[0010] Preferably, before iterating through all hysteresis parameters, the process further includes preprocessing the control quantity time series and ply thickness time series within the historical window. The preprocessing process includes:

[0011] The fitting function between the control quantity and the time point is fitted based on the control quantity time series; the difference between the control quantity data in the control quantity time series and the fitted control quantity obtained based on the fitting function is calculated to obtain the difference sequence; the difference sequence is standardized to complete the preprocessing.

[0012] The preprocessing process for the layup thickness time series is the same as the preprocessing process for the control quantity time series.

[0013] By preprocessing the control quantity time series and ply thickness time series within the historical window, trend interference in the data is eliminated, making the cross-correlation analysis more accurate. The target lag parameter is determined by the peak value of the cross-correlation value, which can more accurately identify the time alignment point between the control quantity and the ply thickness, further improving the control accuracy.

[0014] Preferably, the construction process of the mapping function is as follows:

[0015] Set the first parameter, the second parameter, and the third parameter. Use the sum of the average driving speed and the third parameter as the denominator and the second parameter as the numerator to form a fraction. Calculate the sum of the first parameter and the fraction to obtain the mapping function.

[0016] Preferably, an objective function is constructed, and all the first, second, and third parameters are iterated. When the difference between the objective lag parameter of the historical window and the fitted lag value obtained through the mapping function is minimized, the corresponding first, second, and third parameters are the optimal parameters. The optimal parameters are then applied to the mapping function to obtain the optimal mapping function.

[0017] Preferably, the old residuals are updated using a single exponential smoothing method, and the update process includes:

[0018] Calculate the difference between the target lag parameter of the current window and the fitted lag value obtained through the mapping function, and use it as the new residual; set the smoothing coefficient, calculate the product of the difference between 1 and the smoothing coefficient and the old residual, calculate the product of the smoothing coefficient and the new residual, and calculate the sum of the two products to obtain the updated residual.

[0019] By using a single exponential smoothing update method, the residual update process becomes smoother, avoiding drastic changes in control input caused by abrupt changes, thereby improving the stability of the control system.

[0020] Preferably, after obtaining the updated residual, the method further includes: if the updated residual exceeds a preset range, then limiting the updated residual to the preset range.

[0021] This prevents the residuals from becoming too large or too small, thus avoiding instability in the system due to drastic changes in the residuals.

[0022] Preferably, the PID controller operates under an anti-integral saturation mechanism.

[0023] The PID controller operates under an anti-integral saturation mechanism. When the actuator reaches its physical limit, the integral term is frozen to prevent integral saturation. This further enhances the stability of the system and avoids system instability caused by an excessively large integral term.

[0024] Preferably, the calculation process of the peak confidence level includes:

[0025] Calculate the average of all cross-correlation values ​​within the current window, and use the ratio of the peak value to the average value as the peak confidence level.

[0026] In a second aspect, a paver paving thickness control system includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the paver paving thickness control method described in any one of the claims is implemented.

[0027] The present invention has the following effects:

[0028] 1. This invention fits a mapping function model to historical construction data and updates the fitted lag value in real time when the paver is running online. This allows the obtained adaptive lag value to dynamically reflect the actual lag time of the paver at different travel speeds. This adaptive adjustment enables the control system to more accurately compensate for the control delay caused by lag, thereby improving the control accuracy of the pavement thickness.

[0029] 2. The adaptive hysteresis value of the present invention changes continuously with the travel speed of the paver without vibration. This continuous change enables the control system to respond more quickly to changes in the actual construction process, thereby shortening the system's settling time. Moreover, since the adaptive hysteresis value is closer to the actual value, the phase margin is improved, which enables the control system to adjust the control quantity more accurately, thereby reducing the occurrence of overshoot and tailing phenomena.

[0030] 3. When updating residuals, this invention determines whether the target lag parameter of the current window is reliable by calculating the peak confidence level. Only when the peak confidence level is greater than a preset threshold is the old residual updated. This mechanism avoids erroneous updates caused by noise or disturbances, thereby improving the stability of the system.

[0031] 4. This invention does not require the addition of new sensors or changes to the mechanical structure of the paver. It only requires replacing the control algorithm and can be implemented using existing travel speed, pavement thickness, and control parameters. This reduces the system modification cost and implementation difficulty, making the technical solution of this invention easier to promote and apply. Attached Figure Description

[0032] Figure 1 This is a flowchart of steps S1-S5 in a paver thickness control method according to an embodiment of the present invention. Detailed Implementation

[0033] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0034] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0035] Reference Figure 1 A method for controlling the paving thickness of a paver, comprising steps S1-S6, as detailed below:

[0036] S1. Collect historical construction data of the paver and divide the historical construction data into multiple historical windows according to the preset time length. Each historical window includes the control quantity time sequence, the travel speed time sequence, and the pavement thickness time sequence.

[0037] Historical construction data is extracted from the paver's historical construction log and arranged in chronological order according to the original timestamps to obtain the control quantity time series, travel speed time series, and pavement thickness time series. The control quantity is the equivalent of the screed angle of attack or traction point height. The control quantity time series, travel speed time series, and pavement thickness time series are divided into multiple historical windows of 10 seconds each, according to a preset time length (e.g., 10 seconds). Each historical window includes a control quantity time series segment, a travel speed time series segment, and a pavement thickness time series segment. The preset time length can be adaptively adjusted according to the specific implementation scenario.

[0038] For a single historical window, the least squares method is used to perform linear fitting on the time series of control quantities within the historical window to obtain a fitting function between the control quantities and time points. Similarly, the least squares method is used to perform linear fitting on the time series of ply thickness within the historical window to obtain a fitting function between the ply thickness and time points. The fitting functions can describe the trends of the control quantities and ply thickness over time within that time period.

[0039] Substituting the time points into the two fitting functions above yields the fitted data for the control quantity and ply thickness, respectively. Subtracting the corresponding fitted data for the control quantity from the data within the time series yields a difference sequence. Standardizing this difference sequence yields the control quantity sequence. Similarly, subtracting the corresponding fitted data for the ply thickness from the data within the time series yields a difference sequence. Standardizing this difference sequence yields the ply thickness sequence.

[0040] If the data in the control quantity time series or the ply thickness time series show a significant upward or downward trend, direct analysis may be affected by this trend. By fitting a straight line and subtracting this line, this trend can be eliminated, making the data more stable.

[0041] S2. Set the range of hysteresis parameters. Based on a single historical window, traverse all hysteresis parameters, find the hysteresis parameter corresponding to the peak value of the cross-correlation between the control quantity time series and the layup thickness time series in the historical window, and mark it as the target hysteresis parameter.

[0042] The range of the hysteresis parameter is set by the maximum possible delay time between the control quantity and the ply thickness. For example, if we know from experience and engineering knowledge that the change in the control quantity will not affect the ply thickness for more than 2 seconds, then the range of the hysteresis parameter can be set to [-2, 2].

[0043] A minimum hysteresis step size is preset as the minimum time interval to be considered when calculating the cross-correlation value. The hysteresis parameter is a value at different multiples of the minimum hysteresis step size. By setting a smaller minimum hysteresis step size, possible hysteresis parameters can be searched more meticulously, thereby estimating the actual hysteresis time between the control quantity and the layup thickness more accurately.

[0044] Calculate the cross-correlation values ​​between the data in the control quantity sequence and the data in the ply thickness sequence under different lag parameters, and take the peak value among all cross-correlation values ​​as the target lag parameter of the historical window.

[0045] For example, the hysteresis parameter is expressed as The maximum possible time delay between the data in the control sequence and the data in the ply thickness sequence is ,So For each possible hysteresis parameter ,calculate and The cross-correlation value, Indicates a point in time The data in the corresponding control quantity sequence, This indicates the time points in the ply thickness sequence. Offset hysteresis parameter The peak value among all cross-correlation values ​​is then taken as the target lag parameter for the historical window.

[0046] By calculating the cross-correlation value between the control quantity and the ply thickness, the optimal time alignment point between the two signals can be identified, i.e., the peak value of the cross-correlation value. The hysteresis parameter corresponding to this peak value is considered to be the most likely delay in the influence of the control quantity on the ply thickness.

[0047] After calculating all cross-correlation values ​​within each historical window, the peak confidence score for each historical window is calculated for subsequent steps. For a single historical window, the average of all cross-correlation values ​​within the window is calculated, and the ratio of the peak value to the average value is calculated to obtain the peak confidence score for that historical window. The higher the peak confidence score, the more significant the peak value of the found cross-correlation value, meaning a stronger correlation between the control variable and the driving speed under the target lag parameter.

[0048] The paver's travel speed directly affects the paving layer thickness. After obtaining the target lag parameter, the average speed of the historical window is calculated: for a single historical window, the average speed is calculated based on the travel speed time series within the historical window. This yields the target lag parameter and average speed for each historical window, which are then used in subsequent calculations.

[0049] S3. Construct a mapping function between average driving speed and target lag parameter based on all historical windows.

[0050] A mapping function between the average driving speed and the target lag parameter is constructed using the nonlinear least squares method. The mapping function is specifically expressed as follows:

[0051]

[0052] In the formula, This represents a mapping function that describes the fitting lag value of the paver at different travel speeds; Indicates average driving speed; The first parameter represents the average driving speed. The minimum value of the fit lag that cannot be eliminated as it approaches infinity; The second parameter represents the dominant strength of the influence of average driving speed on the fitted lag value; The third parameter represents the numerical stability term, used to prevent the denominator from approaching zero at low speeds, thus ensuring the numerical stability of the function. , and These are learning parameters that need to be obtained through data fitting. Since the faster the paver travels, the smaller the lag time, therefore... , and Need to meet , , The conditions are set to ensure the monotonicity of the mapping function.

[0053] To find the optimal parameters , and Define an objective function to measure the mapping function. The difference between the target lag parameter and the objective parameter. The objective function is specifically expressed as follows:

[0054]

[0055] In the formula, Indicates the total number of history windows; Indicates the first A historical window; Indicates the first The weight value of each historical window; Indicates the first The average speed of a historical window; Represents the first value calculated by the mapping function. The fitting lag value of each historical window; Indicates the first The target lag parameter for each historical window, i.e., the lag time actually observed; Indicated by An exponential function with base 0.5, to prevent The presence of a value of 0 between these values ​​renders the weights invalid. and The closer, The smaller the value of , the smaller the difference between the fitted lag value obtained by the mapping function and the actual observed lag time, meaning the better the fitting effect.

[0056] in, , Indicates the first Peak confidence level of a historical window; This indicates a preset threshold, only... Greater than hour, Only the corresponding target lag parameter is valid, and its weight is calculated based on the difference between the peak confidence level and the preset threshold; otherwise... The corresponding historical window data is invalid, and the weight of the historical window is 0, indicating that its contribution to the fitting process is ignored; Take the empirical value of 0.6.

[0057] The objective function is solved using nonlinear least squares or other optimization algorithms (such as genetic algorithms) to find the optimal parameters. , and These parameters make and The objective function should be as close as possible to the target. The process of solving the objective function is based on existing technology and will not be described in detail here.

[0058] After obtaining the optimal parameters, the optimal mapping function model was obtained. This mapping function model can more accurately predict the lag time during the paving process at different travel speeds. This is crucial for compensating for the control delay caused by the lag. This predicted lag time will be used in the online control system to dynamically adjust the paver's control strategy to adapt to speed changes and other factors that may affect the paving quality, thereby controlling the pavement thickness more accurately.

[0059] S4. Construct the current window with the current time as the endpoint, calculate the target lag parameter and peak confidence of the current window, obtain the fitted lag value corresponding to the current average driving speed according to the mapping function, and update the old residual when the peak confidence is greater than the preset threshold. The sum of the fitted lag value corresponding to the current window and the updated residual is used as the adaptive lag value; otherwise, the sum of the fitted lag value corresponding to the current window and the old residual is used as the adaptive lag value.

[0060] To update the lag time estimate in real time and adapt to changes in paver operating conditions, during online operation, real-time construction data for a preset duration (e.g., 5 seconds) prior to the current moment is collected every preset interval (e.g., 3 seconds), with the current moment as the endpoint, to construct the current window. The current window also includes control quantity time series, travel speed time series, and pavement thickness time series. The data within the current window is preprocessed to obtain the control quantity sequence, travel speed sequence, and pavement thickness sequence. The cross-correlation values ​​between the data in the control quantity sequence and the data in the pavement thickness sequence under different lag parameters are calculated. The peak value of the cross-correlation value is taken as the target lag parameter for the current window, and the peak confidence score of the current window is calculated based on the cross-correlation value. The methods for obtaining the target lag parameter and peak confidence score for the current window are the same as those for the historical window.

[0061] Specifically, when the peak confidence level of the current window is greater than a preset threshold, the old residuals are updated using a single exponential smoothing method, which satisfies the following formula:

[0062]

[0063] In the formula, This represents the updated residual, and ; This represents the smoothing coefficient, used for exponential smoothing of residual updates, for example, In this case, each update will result in the residual being 80% of the old residual plus 20% of the new residual. This represents the old residual, obtained from the previous update process. At the initial moment when the system just starts running online, the old residual is 0. Indicates the target hysteresis parameter for the current window; Represents the average speed in the current window. Below, the fitted lag value is obtained based on the mapping function; This represents the difference between the target lag parameter and the fitted lag value under the current window, that is, the difference between the actual observed lag time and the predicted lag time, also known as the new residual, which represents the deviation between the actual observation and the model prediction. This indicates the preset range of the residual. The absolute value exceeds At that time, it is necessary to limit it to Internally, to prevent the updated residual from becoming too large or too small.

[0064] Calculate the average speed in the current window The adaptive lag value is obtained by summing the fitted lag value obtained from the mapping function with the updated residual, as shown in the following formula:

[0065]

[0066] Therefore, in the next update of the residuals, the old residuals used in the next update refer to those of this update. .

[0067] When the peak confidence level of the current window is less than or equal to a preset threshold, calculate the average velocity within the current window. The adaptive lag value is obtained by summing the fitted lag value obtained from the mapping function with the old residual, as shown in the following formula:

[0068]

[0069] Therefore, in the next residual update, the old residual used in the next update refers to the residual used in the current update. .

[0070] The old residuals are only updated when the peak confidence level of the current window exceeds a preset threshold. This means that fine-tuning of the old residuals is only performed when the target lag parameter of the current window is sufficiently reliable, avoiding erroneous updates caused by noise or disturbances, thereby improving the stability and reliability of the system. Furthermore, the use of small-step exponential smoothing updates ensures that each update is small in magnitude, avoiding drastic changes. Through this method, the paver's control system can respond more flexibly and accurately to changes during actual construction, thereby improving paving quality and efficiency.

[0071] S5. The final adaptive hysteresis value is applied to the Smith predictor to predict the ply thickness, generating the predicted ply thickness. The difference between the actual measured ply thickness and the predicted ply thickness is calculated as the prediction error. The prediction error is input into the PID controller, which outputs the control quantity and executes the paving action according to the control quantity.

[0072] A first-order link model describing the dynamic behavior of the paver thickness control system is established. The pure time lag in the Smith predictor is replaced with an adaptive lag value. The Smith predictor outputs the predicted pavement thickness based on the current control quantity and the first-order link model. The actual measured pavement thickness is obtained, and the difference between the actual measured pavement thickness and the predicted pavement thickness is calculated to obtain the prediction error. The PID controller calculates and adjusts the final control quantity based on the prediction error to make the system output closer to the desired value. Thus, the paver is controlled according to the final control quantity to achieve control of the pavement thickness.

[0073] The PID controller operates under an anti-integral saturation mechanism. When the actuator (such as the adjustment mechanism of an ironing plate) reaches its physical limit, further increasing the control quantity (especially the integral term) will no longer be effective and may lead to system instability. Therefore, when the actuator is saturated, the integral term needs to be frozen to prevent integral saturation.

[0074] This invention avoids erroneous updates and oscillations by using monotonic constraint fitting of historical construction data, cross-correlation confidence gating, residual preset range, and anti-integral saturation measures, making the control system more stable and reliable. The adaptive lag value changes continuously with the paver's travel speed without jitter, and the adaptive lag value is closer to the actual value, thus improving the phase margin, shortening the settling time, and reducing overshoot and tailing phenomena. No new sensors or changes to the paver's mechanical structure are required; only the control algorithm needs to be replaced, and it can be implemented using the existing travel speed, pavement thickness, and control variables.

[0075] The system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the paver paving thickness control method according to the first aspect of the present invention.

[0076] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0077] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for controlling the paving thickness of a paver, characterized in that, include: Collect historical construction data of the paver and divide the historical construction data into multiple historical windows according to the preset time length. Each historical window includes the control quantity time sequence, the travel speed time sequence, and the pavement thickness time sequence. Set the range of hysteresis parameters, traverse all hysteresis parameters based on a single historical window, find the hysteresis parameter corresponding to the peak value of the cross-correlation between the control quantity time series and the ply thickness time series in the historical window, and mark it as the target hysteresis parameter; Construct a mapping function between average driving speed and target lag parameter based on all historical windows; Construct the current window with the current time as the endpoint, calculate the target lag parameter and peak confidence of the current window, obtain the fitted lag value corresponding to the current average driving speed according to the mapping function, update the old residual when the peak confidence is greater than the preset threshold, and take the sum of the fitted lag value corresponding to the current window and the updated residual as the adaptive lag value. Otherwise, the sum of the fitted lag value and the old residual corresponding to the current window is used as the adaptive lag value; The final adaptive hysteresis value is applied to the Smith predictor to predict the ply thickness, generating the predicted ply thickness. The difference between the actual measured ply thickness and the predicted ply thickness is calculated as the prediction error. The prediction error is input into the PID controller, which outputs a control quantity and executes the paving action according to the control quantity. The process of constructing the mapping function is as follows: Set the first parameter, the second parameter, and the third parameter; use the sum of the average driving speed and the third parameter as the denominator and the second parameter as the numerator to form a fraction; calculate the sum of the first parameter and the fraction to obtain the mapping function. Construct an objective function, iterate through all the first, second, and third parameters, and when the difference between the objective lag parameter of the historical window and the fitted lag value obtained through the mapping function is the smallest, the corresponding first, second, and third parameters are the optimal parameters. Apply the optimal parameters to the mapping function to obtain the optimal mapping function.

2. The method for controlling the paving thickness of a paver according to claim 1, characterized in that, Before iterating through all hysteresis parameters, preprocessing is performed on the control quantity time series and ply thickness time series within the history window. The preprocessing process includes: The fitting function between the control quantity and the time point is fitted based on the control quantity time series; the difference between the control quantity data in the control quantity time series and the fitted control quantity obtained based on the fitting function is calculated to obtain the difference sequence; the difference sequence is standardized to complete the preprocessing. The preprocessing process for the layup thickness time series is the same as the preprocessing process for the control quantity time series.

3. The method for controlling the paving thickness of a paver according to claim 1, characterized in that, The old residuals are updated using a single exponential smoothing method. The update process includes: Calculate the difference between the target lag parameter of the current window and the fitted lag value obtained through the mapping function, and use it as the new residual; set the smoothing coefficient, calculate the product of the difference between 1 and the smoothing coefficient and the old residual, calculate the product of the smoothing coefficient and the new residual, and calculate the sum of the two products to obtain the updated residual.

4. The method for controlling the paving thickness of a paver according to claim 3, characterized in that, After obtaining the updated residual, the method also includes: if the updated residual exceeds a preset range, then the updated residual is limited to the preset range.

5. The method for controlling paving thickness using a paver according to claim 1, characterized in that, The PID controller operates under an anti-integral saturation mechanism.

6. The method for controlling the paving thickness of a paver according to claim 1, characterized in that, The calculation process for the peak confidence level includes: Calculate the average of all cross-correlation values ​​within the current window, and use the ratio of the peak value to the average value as the peak confidence level.

7. A paver thickness control system, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement the paver paving thickness control method according to any one of claims 1-6.

Images