A cableway unhooking grip self-adapting PID control method and system

Abstract

The present application relates to the technical field of cableway equipment, and proposes a cableway unhooking grip rope device adaptive PID control method and system, which comprises: collecting a steel wire rope vibration signal in the operation process of a cableway cable car; decomposing to obtain a plurality of component signals, dividing each component signal into a plurality of segmented signals and a plurality of time period signals; obtaining the relative deviation between the time period signals of adjacent segmented signals of the same component signal; screening to obtain a plurality of stable time period signals of each component signal; obtaining vibration inherent information of an operation cycle, and then obtaining vibration random information of each operation cycle; according to the correlation between the vibration inherent information of each operation cycle and the original vibration signal, adjusting the loss function of the LSTM model and training, real-time predicting the load value and performing PID feedforward control. The present application aims to solve the problem that the error signal suddenly changes and lags due to impact load, thereby affecting the accuracy of PID feedforward control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cableway equipment, in particular to a self-adaptive PID control method and system for a cableway dismounting and holding grip. BACKGROUND

[0002] The dismounting and holding grip is the core actuator of a circulating passenger cableway (such as a chairlift or a gondola lift), which dismounts the carrier (chair / gondola) from the high-speed running steel wire rope at the entrance of a station, so that the carrier can pass through the station curve and the boarding and alighting area at a low speed; at the exit of the station, the carrier needs to be accurately, smoothly and reliably connected to the high-speed running steel wire rope, and the dynamic characteristics of the dismounting and holding process directly affect the safety and comfort of passengers, so it is necessary to continuously experimentally simulate and test different working conditions to ensure the safety of the dismounting and holding grip.

[0003] The electrical control technology of the commonly used experimental control device for the cableway dismounting and holding grip mainly focuses on the power loading mode, parameter adjustment accuracy, multi-working condition simulation capability and safety protection mechanism; the traditional test bench usually uses fixed oil cylinders or weight loading, which cannot flexibly adjust the loading angle and lateral friction, resulting in a large deviation between the test parameters and the actual working conditions, and at the same time, due to the inability to realize digital control, the test efficiency is low. The self-adaptive PID control for the cableway dismounting and holding grip can rely on the feedback of errors to realize feedforward control to ensure the stable operation of the dismounting and holding grip, but the impact load will cause the error signal to mutate and then produce hysteresis, so it is necessary to introduce an LSTM model to effectively control the hysteresis effect of the impact load, and then effectively predict the periodic load change through training of an accurate LSTM model to realize self-adaptive PID control. SUMMARY

[0004] The present application provides a self-adaptive PID control method and system for a cableway dismounting and holding grip to solve the problem of error signal mutation and hysteresis caused by impact load, which affects the accuracy of PID feedforward control, and the technical solution adopted is as follows:

[0005] The present application provides a self-adaptive PID control method and system for a cableway dismounting and holding grip to solve the problem of error signal mutation and hysteresis caused by impact load, which affects the accuracy of PID feedforward control, and the technical solution adopted is as follows:

[0006] Collect the steel wire rope vibration signal during the operation of the cable car, and record the cycle length of several cycles of the cable car and the running time between adjacent stations of the cable car in each cycle;

[0007] The vibration signal is decomposed by EMD to obtain a plurality of component signals, and each component signal is divided into a plurality of segmented signals and a plurality of period signals according to the cycle length and the running length; based on the extreme point distribution in each segmented signal of each component signal, a key point curve of each period signal is obtained, and then the matching result of the key point curves between the period signals of adjacent segmented signals of the same order is analyzed to obtain the relative deviation between the period signals of adjacent segmented signals of the same component signal;

[0008] Based on the relative deviation between the period signals of different segmented signals of the same component signal, a plurality of stable period signals of each component signal are screened; based on the distribution of the stable period signals in each component signal, a weight is constructed, and the original vibration signal corresponding to the time interval of each stable period signal in each segmented signal is weighted and fused to obtain the vibration inherent information of the running cycle, and then the vibration random information of each running cycle is obtained;

[0009] According to the correlation between the vibration inherent information of each running cycle and the original vibration signal, the loss function of the LSTM model is adjusted and trained to predict the load value in real time and perform PID feedforward control.

[0010] Optionally, the plurality of component signals obtained by decomposing the vibration signal by EMD comprises the following specific method:

[0011] The steel wire rope vibration signal is decomposed by EMD to obtain a plurality of IMF components, and each IMF component is taken as a component signal.

[0012] Optionally, the plurality of component signals obtained by decomposing the vibration signal by EMD comprises the following specific method:

[0013] The average cycle length of all the cable cars in each running cycle is taken as a first segmented length, and for the running length of the same order in each running cycle, the average running length of the order in each running cycle is taken as a second segmented length, a plurality of second segmented lengths in each running cycle are obtained, and the second segmented lengths are arranged in order;

[0014] For any component signal, the component signal is divided into a plurality of segmented signals by the first segmented length; for any segmented signal, the segmented signal is segmented by each second segmented length in order from the first data point, wherein after a period signal is obtained by each second segmented length, the next second segmented length is segmented in order, and the plurality of period signals of the component signal and the segmented signal are obtained by segmenting each second segmented length in order.

[0015] Optionally, the key point curve of each period signal is obtained by the following specific method:

[0016] For any component signal and any segment of the signal, obtain several extreme points in the segment, connect the extreme points that are adjacent in time sequence to obtain the extreme point curve of the segment, and truncate the extreme point curve of each time period in the segment to obtain the key point curve of each time period.

[0017] Optionally, the specific method for obtaining the relative deviation between time-segment signals of adjacent segments with the same component signal includes:

[0018] For any two time period signals in any order of any two adjacent segment signals in the component signal, DTW matching is performed on the key point curves of the two time period signals. Matching is performed on the corresponding extreme points in the key point curves of the two time period signals to obtain several matching extreme point pairs. The time corresponding to the first data point in each segment signal is taken as the zero time of each segment signal. The time point of each extreme point in the corresponding segment signal is obtained. The matching time point pair is obtained for each matching extreme point pair.

[0019] For any pair of matching time points between any two adjacent segment signals in any component signal, the absolute value of the difference between the two time points is obtained as the basic time deviation, and the basic time deviation corresponding to each pair of matching time points is obtained.

[0020] Using the baseline time deviations as the independent variables of the cross-correlation function, a cross-correlation function is constructed for the key point curves of the signals from the two time periods. The expression is:

[0021] ;

[0022] The domain of the cross-correlation function integral is... The value range is the base time deviation corresponding to each matching time point. Indicates the number of matching time point pairs. Indicates the first The time point in the corresponding time period of the previous segment signal in the adjacent segment signal of the matching time point pair. Indicates the first The time point in the corresponding time period of the next segment signal in the adjacent segment signal of the matching time point pair. This represents the signal value in the corresponding time period of the preceding segment of the signal in an adjacent segment. This represents the signal value in the corresponding time period of the next segment of the adjacent segmented signal. It is an exponential function with the natural constant as its base. The attenuation coefficient;

[0023] acquiring the basic time deviation corresponding to the maximum value of the cross-correlation function output value, and taking the basic time deviation as the final time deviation of the two time period signals; acquiring the final time deviation of the two time period signals of each order in each adjacent segmented signal in the component signal, and performing linear normalization on all final time deviations to obtain a result as the relative deviation of the two time period signals.

[0024] Optionally, the screening obtains several stable time period signals of each component signal, and the specific method comprises the following steps:

[0025] For any time period signal in any segmented signal in any component signal, if there is at least one relative deviation less than or equal to the deviation threshold value in the two relative deviations corresponding to the time period signals of the same order in the adjacent previous segmented signal and the adjacent subsequent segmented signal of the time period signal and the segmented signal, the time period signal is taken as a stable time period signal of the component signal; the time sequence signals of each segmented signal in the component signal are subjected to threshold value judgment to obtain several stable time period signals of the component signal.

[0026] If there is a time period signal of any order that is not judged as a stable time period signal in each segmented signal, the mean value of the relative deviations corresponding to each time period signal in the several time period signals of the order of each segmented signal in the component signal is acquired, and the time period signal corresponding to the minimum value of all the obtained mean values of the relative deviations is taken as a stable time period signal.

[0027] Optionally, the weight is constructed based on the distribution performance of the stable time period signals in each component signal, and the specific method comprises the following steps:

[0028] For any stable time period signal in any component signal, the number of component signals in which the time period signal corresponding to the time interval of the stable time period signal is a stable time period signal is acquired, and a sum value obtained by adding 1 to the number is taken as the occurrence frequency of the stable time period signal, and a ratio of the occurrence frequency to the number of component signals is taken as the occurrence frequency of the stable time period signal.

[0029] The occurrence frequencies of all stable time period signals of any order in the component signal are subjected to weight normalization to obtain a result as the confidence weight of each stable time period signal.

[0030] Optionally, the vibration inherent information of the running cycle is obtained, and then the vibration random information of each running cycle is acquired, and the specific method comprises the following steps:

[0031] The original vibration signals under the time interval of each stable time period signal are subjected to weighted summation according to the confidence weight, and a result is taken as the fixed time period information of the order.

[0032] The fixed time period information is acquired in sequence, the fixed time period information is combined in sequence to obtain a combined signal of a time range corresponding to a segmented signal, the time range corresponding to each segmented signal is regarded as each operation cycle, and the combined signal is regarded as vibration inherent information of the operation cycle.

[0033] A segmented vibration signal corresponding to each operation cycle in the original vibration signal is acquired as the segmented vibration signal of each operation cycle; and a difference signal obtained by subtracting the vibration inherent information of the operation cycle from the segmented vibration signal of the operation cycle is regarded as vibration random information of the operation cycle.

[0034] Optionally, the loss function of the LSTM model is adjusted and trained according to the correlation between the vibration inherent information of each operation cycle and the original vibration signal, and the specific method comprises the following steps:

[0035] For any operation cycle, a Pearson correlation coefficient is calculated between the vibration inherent information of the operation cycle and the segmented vibration signal of the operation cycle, and is regarded as information proportion of the operation cycle; and a difference obtained by subtracting the information proportion from 1 is regarded as random proportion of the operation cycle.

[0036] An LSTM model is constructed, a training data set is constructed based on the segmented vibration signal of each operation cycle in the steel wire rope vibration signal, and is regarded as input of the LSTM model, and a signal prediction value is output in a training process; for the segmented vibration signal of any operation cycle, the signal prediction value is output under the condition of being input into the LSTM model; L2 norms of inherent information in the vibration inherent information of the operation cycle and the signal prediction value are acquired, and L2 norms of random information in the vibration random information of the operation cycle and the signal prediction value are acquired; information proportion is regarded as weight of the L2 norm of the inherent information, and random proportion is regarded as weight of the L2 norm of the random information; a loss function is constructed by weighted summation, the LSTM model is trained through the segmented vibration signal of each operation cycle, and a trained LSTM model is obtained.

[0037] The application further provides a cableway dismounting and mounting cable gripper adaptive PID control system, which comprises a stud sleeve, a rod end joint bearing, a dismounting and mounting cable gripper, a platform frame, a proximity switch, a switch frame and a dismounting and mounting cable gripper test bed electrical control system.

[0038] The beneficial effects of the present application are: the present application analyzes the impact load hysteresis effect by monitoring the vibration signal of the cableway steel wire rope, wherein the component signals are obtained by EMD decomposition of the steel wire rope vibration signal, and the segmented signal division and time period signal acquisition of the component signals are carried out according to the cycle length and the running length; the stress release time is reflected according to the extreme point distribution in the segmented signal, the matching similarity analysis between the extreme points in the time period signals of the same order in the adjacent segmented signals is carried out, the matching results of the extreme point corresponding time points are analyzed in correlation considering the impact load attenuation effect, and the relative deviation of the time period signals of the same order in the adjacent segmented signals of the same component signal is quantitatively obtained; the stable time period signal is selected on the basis of the relative deviation, so as to reflect that the hysteresis effect of the impact load on the corresponding component signal is small, and the periodic and time performance is more regular; the frequency of the stable time period signal in the same time interval in different component signals is quantified, the overall stability of the corresponding time interval in the original vibration signal is reflected, and then the original vibration signal is weighted and fused in the time range corresponding to the different segmented signals, i.e. each vibration signal in the running cycle, so as to extract the vibration inherent information in the original vibration signal which is not affected by the impact load, and obtain the vibration random information of each running cycle, so as to train the LSTM model, solve the problem that the LSTM model cannot predict the sudden impact load, so that the PID feedforward control can be based on the hysteresis effect of the impact load in the historical load data, can respond to the possible impact load in time in the load prediction process, and can be fed forward and compensated, and then the accuracy of the PID adaptive control of the cableway decoupling grip is improved. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0040] Figure 1 A cableway decoupling grip adaptive PID control method flow chart provided by an embodiment of the present application;

[0041] Figure 2 A cableway decoupling grip adaptive PID control system structure diagram provided by another embodiment of the present application;

[0042] Figure 3 An internal structure block diagram of a decoupling grip test control box used in the present application;

[0043] Figure 4 An adaptive PID control algorithm architecture flow chart used in the present application.

[0044] Figure 2 In the figure, 1 is a stud sleeve, 2 is a rod end joint bearing, 3 is a detachable rope gripper, 4 is a platform frame, 5 is a proximity switch, 6 is a switch frame, and 7 is an electrical control system of a detachable rope gripper test bench.

[0045] Figure 3 In the figure, 301 is a man-machine interface, 302 is a logical relationship processing circuit, 303 is a control relay, 304 is a motor, 305 is an alarm relay, 306 is an alarm, 307 is a data processing unit, and 308 is a sensor signal conversion unit. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0047] Please refer to Figure 1 which shows a flow chart of a ropeway detachable rope gripper adaptive PID control method provided by an embodiment of the present application. The method comprises the following steps:

[0048] In step S001, the steel wire rope vibration signal in the running process of the ropeway cable car is collected, and the cycle time of several running cycles of the cable car and the running time between adjacent platforms in each running cycle are recorded.

[0049] The purpose of the embodiment is to analyze the hysteresis effect of the steel wire rope vibration signal on the impact load in the running process of the ropeway cable car, to realize the feedforward response control of the detachable rope gripper by adjusting and training the LSTM model and combining the PID control. First, the steel wire rope vibration data need to be obtained. The running of the ropeway cable car is periodic, and it has passed through multiple platforms. In order to facilitate the analysis of the inherent information of the impact load and vibration, the cycle time of the running cycle and the running time between adjacent platforms need to be extracted.

[0050] Specifically, the vibration data of the steel wire rope in the operation process of the cableway cable car is collected by a vibration sensor on the dismounting gripper test bench, and a steel wire rope vibration signal is formed according to the time sequence. The vibration data of the steel wire rope in all historical and current tests is collected, and a complete steel wire rope vibration signal is formed according to the time sequence. Each test includes multiple operation cycles, and the operation cycle is the time period when the cable car reaches the same position after one round of cableway operation. The cycle length of each operation cycle, i.e., the length of the corresponding time period, is obtained. At the same time, the cable car passes through multiple stations in a single operation cycle, and the time interval between adjacent stations, i.e., the length of time from one station to another station, is obtained as the operation length between adjacent stations.

[0051] It should be noted that the LSTM network is good at learning long-term sequences and can well predict periodic load changes (such as periodic stress fluctuations when the gripper is dismounted), but the duration of some impact loads (such as the impact at the moment of dismounting) is relatively short. Impact load and periodic load are both acting on the steel wire rope, and the vibration signal of the steel wire rope reflects the dismounting process of the dismounting gripper of the cableway and the unique periodic process of the cableway cable car operation. Therefore, by analyzing the vibration signal collected by the steel wire rope in the data set, and the impact load and periodic load are both reflected in the different frequency information of the vibration signal, the vibration signal is decomposed by EMD to obtain different IMF components as component signals, each component signal represents a frequency information, and the EMD decomposition retains the phase information of the signal, which is translationally invariant to the instantaneous impact of the steel wire rope.

[0052] Step S002, a plurality of component signals are obtained by EMD decomposition on the vibration signal, and each component signal is divided into a plurality of segmented signals and a plurality of period signals according to the cycle length and the operation length; based on the extreme point distribution in each segmented signal of each component signal, the key point curve of each period signal is obtained, and then the matching result of the key point curves between the period signals of the adjacent segmented signals of the same order is analyzed to obtain the relative deviation between the period signals of the adjacent segmented signals of the same component signal.

[0053] Preferably, in an embodiment of the present application, a plurality of component signals are obtained by EMD decomposition on the steel wire rope vibration signal, and each component signal is divided into a plurality of segmented signals and a plurality of period signals according to the cycle length and the operation length, and the specific method comprises:

[0054] The steel wire rope vibration signal is decomposed by EMD to obtain several IMF components, each of which is taken as a component signal; and the average of the cycle length obtained in each running cycle of the cable car is taken as the first subsection length. For the running length of the same order in each running cycle, i.e. the running length between the adjacent platforms corresponding to the same order in different running cycles, the average of the running length of the order in each running cycle is taken as a second subsection length. Several second subsection lengths in each running cycle are obtained according to the above method. The second subsection lengths are arranged in order and correspond to the order between the adjacent platforms in each running cycle, i.e. the second subsection lengths correspond to the adjacent platforms in each segment in the running cycle. It should be noted that the sum of the second subsection lengths should be equal to the first subsection length to avoid deviation in the process of obtaining the period signal of the segmented signal. In the process of calculating the second subsection length in the embodiment, if the sum is less than the first subsection length, the remaining length is added to the last second subsection length, and if the sum is greater than the first subsection length, the last second subsection length is deleted to make the sum equal to the first subsection length. This is the method for ensuring that the sum of the second subsection lengths is equal to the first subsection length in the embodiment, and other methods can be used to ensure the equality according to the actual situation.

[0055] It should be noted that different scale periods are divided into large periods and small periods. The large period represents the time length when the same cable car moves to the same position, i.e. a running cycle, and the small period represents the time interval when the cable car reaches the platform each time, i.e. the time interval between adjacent cable cars. The second subsection length corresponding to the adjacent platforms, i.e. the time interval of the adjacent cable cars, is obtained.

[0056] Further, for any component signal, the component signal is divided into several segmented signals by the first subsection length. If the number of remaining data points is less than the first subsection length, the remaining data points are discarded, i.e. the segmented signal needs to ensure a complete running cycle. For any segmented signal, the segmented signal is segmented by each second subsection length in order from the first data point, and each second subsection length obtains a period signal. Then the segmented signal is segmented by the next second subsection length in order, and the segmented signal is segmented by each second subsection length in order to obtain several period signals of the segmented signal.

[0057] It is further needed to be explained that, for each component signal, the segmented signal is obtained, then the time interval signal is obtained for each segmented signal, the similarity of the time interval signals of the same order between adjacent segmented signals is calculated to judge the time deviation, and the greater the time deviation is, the more the other component signals in the time interval of the corresponding time interval signal contain some additional information (in the process of normal impact load, the inherent frequency information changes regularly, and the information of one component changes, according to the principle of EMD decomposition, the corresponding component information also changes), and the reason represented by the time deviation is that the impact load excites stress waves in the structure, and when the wave front reaches different positions, the vibration response of the local area has a time difference, resulting in that the signals of different measuring points in the same period are not synchronized, causing phase shift.

[0058] Preferably, in an embodiment of the present application, the key point curve of each time interval signal is obtained based on the extreme point distribution in each segmented signal of each component signal, and then the matching result of the key point curves between the time interval signals of the same order between adjacent segmented signals is analyzed to obtain the relative deviation between the time interval signals of adjacent segmented signals of the same component signal, and the specific method includes:

[0059] It is needed to be explained that the similarity of the time interval signals needs to be analyzed, and the hysteresis is also considered, that is, the normal steel wire rope vibrates regularly with the movement of the cable car, and for the current time interval signal, if the change range data of the previous time interval signal is unstable, the subsequent time window will produce hysteresis effect; since there are key nodes in the vibration of the steel wire rope, these key nodes correspond to the key moments in the running process, that is, the moments when the system is operated, that is, the stress release moments, therefore, when comparing the similarity, the time interval between the key nodes is quantified, that is, the envelope peak appears at the operation moment, then the key point curve is constructed by the extreme points, and the similarity between the time interval signals is analyzed.

[0060] Specifically, for any segmented signal of any component signal, a plurality of extreme points in the segmented signal are obtained, the time-sequentially adjacent extreme points are connected to obtain the extreme point curve of the segmented signal, the corresponding time interval of each time interval signal in the segmented signal is truncated in the extreme point curve to obtain the key point curve of each time interval signal; for two time interval signals of any order in any two adjacent segmented signals of the component signal, the key point curves of the two time interval signals are matched by DTW, that is, the corresponding extreme points in the key point curves of the two time interval signals are matched to obtain a plurality of matched extreme point pairs; the moment corresponding to the first data point in each segmented signal is taken as the zero moment of each segmented signal, the time points corresponding to each matched extreme point pair are obtained (the difference between the corresponding moment and the moment corresponding to the first data point in the segmented signal), and then the corresponding matched time point pairs are obtained for each matched extreme point pair.

[0061] It should be further explained that since the key point corresponds to the stress release moment, and the impact load on the wire rope is constantly decaying, when calculating the time interval lag after obtaining the key node matching results, the decay characteristics need to be considered. That is, the impact load fluctuates more in the early stage and gradually stabilizes in the later stage, which is the early stage of the time window. Therefore, when comparing time intervals, the decay change needs to be considered.

[0062] Specifically, for any pair of matched time points between any two adjacent segmented signals in any component signal, the absolute value of the difference between the two time points is obtained as the basic time deviation, thus obtaining the basic time deviation corresponding to each pair of matched time points. Using each basic time deviation as the independent variable of the cross-correlation function, a cross-correlation function is constructed for the key point curves of the two time-segmented signals. The construction of the cross-correlation function is a well-known technique in signal processing, and will not be elaborated upon in this embodiment. The constructed cross-correlation function... The expression is:

[0063] ;

[0064] The domain of the cross-correlation function integral is... The value range is the base time deviation corresponding to each matching time point. Indicates the number of matching time point pairs. Indicates the first The time point in the corresponding time period of the previous segment signal in the adjacent segment signal of the matching time point pair. Indicates the first The time point in the corresponding time period of the next segment signal in the adjacent segment signal of the matching time point pair. This represents the signal value in the corresponding time period of the preceding segment of the signal in an adjacent segment. This represents the signal value in the corresponding time period of the next segment of the adjacent segmented signal. It is an exponential function with the natural constant as its base. As the attenuation coefficient, this embodiment uses To narrate.

[0065] Furthermore, the basic time deviation corresponding to the maximum value of the cross-correlation function output is obtained, and this basic time deviation is used as the final time deviation of the two time period signals. The final time deviation of the two time period signals in each order of each adjacent segment signal in the component signal is obtained according to the above method. All final time deviations are linearly normalized, and the result is used as the relative deviation of the corresponding two time period signals.

[0066] So far, the component signals are obtained by EMD decomposition of the steel wire rope vibration signal, and the segmented signal division and time period signal acquisition are performed according to the cycle length and the running length; the stress release moment is reflected according to the extreme value point distribution in the segmented signal, the matching similarity analysis between the extreme value points in the time period signals of the same order in the adjacent segmented signals is performed, the matching result of the time points corresponding to the extreme value points is analyzed in consideration of the impact load attenuation influence, and the relative deviation of the time period signals of the same order in the adjacent segmented signals of the same component signal is quantitatively obtained.

[0067] In step S003, a plurality of stable time period signals of each component signal are screened based on the relative deviations between the time period signals of different segmented signals of the same component signal; the weight is constructed based on the distribution performance of the stable time period signals in each component signal, the original vibration signal corresponding to the time interval of each stable time period signal in each segmented signal is weighted and fused, the vibration inherent information of the running cycle is obtained, and then the vibration random information of each running cycle is obtained.

[0068] Preferably, in an embodiment of the present application, a plurality of stable time period signals of each component signal are screened based on the relative deviations between the time period signals of different segmented signals of the same component signal, and the specific method comprises:

[0069] A preset deviation threshold is adopted, and the deviation threshold is described as 0.3 in this embodiment. For any time period signal in any segmented signal in any component signal, if at least one of the two relative deviations corresponding to the time period signal and the time period signals of the same order in the adjacent previous segmented signal and the adjacent subsequent segmented signal of the segmented signal is less than or equal to the deviation threshold, the time period signal is taken as a stable time period signal of the component signal. The time sequence signals of each segmented signal in the component signal are threshold judged according to the above method, and a plurality of stable time period signals of the component signal are obtained. It is particularly pointed out that if the segmented signal is the first or the last segmented signal in the component signal, each time period signal only corresponds to one relative deviation, and the threshold judgment is performed on the corresponding one relative deviation. After obtaining a plurality of stable time period signals, if there is a time period signal of any order which is not judged as a stable time period signal in each segmented signal, the mean value of the relative deviations corresponding to each time period signal in the plurality of time period signals of the order in each segmented signal of the component signal is obtained (each time period signal corresponds to two relative deviations, and the mean value is calculated, and if it is a time period signal in the first segmented signal or the last segmented signal, the mean value is directly obtained from the corresponding one relative deviation), and the time period signal corresponding to the minimum value of all the obtained mean values of the relative deviations is taken as a stable time period signal.

[0070] It is required to be explained that the period signal with smaller relative deviation has less lag performance caused by impact load, and it has more obvious periodic performance for the steel wire rope vibration signal, contains more fixed information and less random information. Since the inherent information needs to be obtained from the segmented signal in the component signal, it is necessary to ensure that at least one period signal in each order is a stable period signal for subsequent fixed information acquisition. Then, the order without stable period signal is obtained after threshold judgment.

[0071] Preferably, in an embodiment of the present application, the weight is constructed based on the distribution performance of the stable period signal in each component signal, and the original vibration signal corresponding to the time interval of each stable period signal in each segmented signal is weighted and fused to obtain the vibration inherent information of the running cycle, and then the vibration random information of each running cycle is obtained. The specific method includes:

[0072] For any stable period signal in any component signal, the number of component signals in which the time interval corresponding to the stable period signal is a stable period signal (not considering the same order value, directly considering the same time interval) is obtained. Add 1 to the sum value to obtain the occurrence frequency of the stable period signal. The ratio of the occurrence frequency to the number of component signals is taken as the occurrence frequency of the stable period signal.

[0073] It is required to be explained that the occurrence frequency is the number of period signals in each component signal that are stable period signals at the same time interval. The higher the occurrence frequency is, the less likely it is that the corresponding time interval in the steel wire rope vibration signal is affected by impact load, and the higher the credibility of the stable period signal in the corresponding component signal is.

[0074] Further, the occurrence frequency of all stable period signals in the component signal of any order is weighted and normalized to obtain the confidence weight of each stable period signal. The original vibration signal under the time interval of each stable period signal is weighted and summed according to the confidence weight, that is, the original vibration signal corresponding to the time interval of each stable period signal is weighted and fused according to the confidence weight, and the result is taken as the fixed period information of the order. According to the above method, the fixed period information of each order is obtained, and then the fixed period information is combined according to the order sequence to obtain a combined signal corresponding to a time range of the segmented signal. The time range corresponding to each segmented signal is taken as each running cycle, and the combined signal is taken as the vibration inherent information of the running cycle.

[0075] Further, according to the time range of each segmented signal in any component signal, i.e. each operating cycle, the corresponding segment of the vibration signal in the original vibration signal is obtained as the segmented vibration signal of each operating cycle; for the segmented vibration signal of any operating cycle, the difference signal obtained by subtracting the vibration inherent information of the operating cycle from the segmented vibration signal is taken as the vibration random information of the operating cycle.

[0076] At this point, the stable period signal is screened on the basis of the relative deviation to reflect that the impact load has less hysteresis effect on it under the corresponding component signal, and the periodic time performance is more regular. By quantifying the frequency of the stable period signal in different component signals in the same time interval, the overall stability of the corresponding time interval in the original vibration signal is reflected, and then this is used as a weight to weight and fuse the vibration signals of each segment of the original vibration signal in the time range corresponding to the different segmented signals, i.e. the operating cycle, to extract the vibration inherent information in the original vibration signal that is stable and not affected by the impact load, and to obtain the vibration random information of each operating cycle.

[0077] Step S004, according to the correlation between the vibration inherent information of each operating cycle and the original vibration signal, adjust the loss function of the LSTM model and train it, real-time predict the load value and perform PID feedforward control.

[0078] Specifically, for any operating cycle, the Pearson correlation coefficient is calculated between the vibration inherent information of the operating cycle and the segmented vibration signal of the operating cycle, and is taken as the information proportion of the operating cycle. The difference obtained by subtracting the information proportion from 1 is taken as the random proportion of the operating cycle.

[0079] Further, an LSTM model is constructed, a training data set is constructed based on the segmented vibration signal of each operating cycle in the steel wire rope vibration signal, and is taken as the input of the LSTM model. The hidden layer of the LSTM model is 2 layers of LSTM units, each layer has 128 neurons, the activation function is ReLU linear activation function, and the output of the training process is the signal prediction value. For the segmented vibration signal of any operating cycle, the output is the signal prediction value of the operating cycle when the segmented vibration signal is taken as the input of the LSTM model. The L2 norm of the vibration inherent information and the inherent information in the signal prediction value of the operating cycle is obtained, as well as the L2 norm of the vibration random information and the random information in the signal prediction value of the operating cycle. The information proportion is taken as the weight of the L2 norm corresponding to the inherent information, and the random proportion is taken as the weight of the L2 norm corresponding to the random information. The loss function is constructed by weighted summation (the segmented vibration signal of each operating cycle corresponds to a loss function when taken as input), and the LSTM model is trained by the segmented vibration signal of each operating cycle to obtain the trained LSTM model.

[0080] Further, asFigure 4 Fig. 1 shows the flow of the adaptive PID control algorithm architecture, Figure 4 , denotes the error, denotes the error rate of change, , and denote the adjustment of different PID parameters, i.e., the adjustment of the proportion, the derivative and the integral, , and denote different PID parameters, , and denote updated different PID parameters, denotes the output value of the PID, denotes the rate or trend of error change, denotes the total control amount, denotes the feedforward gain coefficient, denotes the load prediction value; data acquisition by sensors and data preprocessing, wavelet denoising and temperature compensation in data preprocessing are all known technologies in the corresponding field, and will not be described in this embodiment; wherein the data acquisition using sensors includes high-precision strain gauges (precision ±0.1% FS) and vibration sensors (frequency response 0~1kHz), the sampling frequency is ≥1kHz, and real-time signal processing is realized by FPGA to realize millisecond-level closed-loop feedback; wavelet denoising in data and processing uses Daubechies 5 wavelet basis, and temperature compensation is processed based on a three-variable coupling model; the closed-loop feedback process of the entire adaptive PID control algorithm is as follows: the sensor data is transmitted to the PLC after being preprocessed by the edge node; the PLC calculates the current error and the error rate of change ; the fuzzy logic module outputs the PID parameter adjustment amount; the LSTM prediction module provides the feedforward compensation signal; the control signal is output to the motor driver to adjust the loading force.

[0081] Further, the input layer of the trained LSTM model is historical load data, and the embodiment obtains load data of nearly 200ms, and the collection time interval is set to 1ms; the output layer is the load prediction value of the future 150ms; at the same time, the specific implementation of the fuzzy logic adjusting the PID parameters is as follows:

[0082] 1) Fuzzy rule design:

[0083] Input variables:

[0084] Error ( ): the difference between the set value and the actual load;

[0085] Error rate of change ( First derivative of error.

[0086] Output variable: PID parameter adjustment amount , , ).

[0087] Fuzzy set partitioning:

[0088] Error ( ): {negative large, negative medium, zero, positive medium, positive large};

[0089] Error rate of change ( ): {negative fast, negative slow, zero, positive slow, positive fast};

[0090] Output adjustment amount: {large decrease, small decrease, unchanged, small increase, large increase}.

[0091] 2) Fuzzy rule base:

[0092]

[0093] 3) Defuzzification method:

[0094] The fuzzy output is converted into an accurate value using the Centroid Method, and the formula is as follows:

[0095]

[0096] wherein, is the membership degree, is the center value of the corresponding fuzzy set; it is the existing technology in the PID control algorithm, and this embodiment will not be repeated.

[0097] In the cooperative control of the LSTM prediction module and the PID, when the obtained load threshold value is close to the threshold value, which is set to ≥90% threshold value in this embodiment, the feedforward compensation signal is triggered, which is the feedforward gain coefficient multiplied by the load prediction value , the feedforward compensation signal is superimposed with the PID output to form the total control amount , which is the sum of the PID output value and the feedforward compensation signal , so as to realize PID feedforward adaptive control; at the same time, a safety interlocking mechanism is set in the algorithm architecture, including mechanical jamming monitoring, communication interruption monitoring and temperature overrun monitoring, and is linked to an emergency shutdown response module, with a response time of less than 10 ms.

[0098] So far, by monitoring the vibration signal of the cableway steel wire rope, the impact load hysteresis effect is analyzed, the LSTM model is trained, the problem that the LSTM model cannot predict the sudden impact load is solved, so that the PID feedforward control can be based on the hysteresis effect of the impact load in the historical load data, can respond to the possible impact load in time in the load prediction process, and can feedforward compensation, thereby improving the accuracy of the PID adaptive control of the cableway detaching and holding gripper.

[0099] Referring to Figure 2 , which shows a cableway detaching and holding gripper adaptive PID control system structure diagram provided by another embodiment of the application, the system includes a stud sleeve, a rod end joint bearing, a detaching and holding gripper, a platform frame, a proximity switch, a switch frame, and a detaching and holding gripper test bench electrical control system, the detaching and holding gripper test bench electrical control system includes a memory, a processor, and a computer program stored in the memory and running on the processor, and when the processor executes the computer program, the above method steps S001 to S004 are implemented.

[0100] Referring to Figure 3 , which shows the internal structure of the detaching and holding gripper test control box, the detaching and holding gripper test control box is arranged at a suitable position near the cableway control test bench, the sensor signal conversion unit 308 is used to collect the load data collected by the load sensor and the weighing sensor, and convert it into an electrical signal; the data processing unit 307 is used to receive the 4~20mA signal transmitted by the sensor signal conversion unit 308 and perform related data conversion, operation processing and storage; the logic relationship processing circuit 302 completes the test device early warning and shutdown determination, and sends a control signal to the cableway gripper control system, i.e. the adaptive PID control system of the cableway detaching and holding gripper, through the alarm relay 305; the data processing unit 307 transmits data to the logic relationship processing circuit 302, and then displays the weighing and counting of the fatigue test bench through the man-machine interface 301, and views the historical curve and queries the historical data; the parameters are set and modified through the buttons of the man-machine interface 301, and operation instructions are issued.

[0101] The main programming configuration of the detaching and holding gripper test control box is as follows:

[0102] 1) The man-machine interface is the opening and closing test bench interface;

[0103] 2) The opening and closing test bench includes display of the counting sensor value, and includes starting and stopping of each motor;

[0104] 3) Each group of sensors is allocated an independent counting and weighing data storage area for querying historical data;

[0105] 4) Assign a counting, weighing real-time data display area for each group of sensors on the human-machine interface 301, so as to read the counting, weighing real-time data of each group in real time;

[0106] 5) According to the change of 4~20mA of each group of sensors, and the conversion relationship between 4~20mA and actual counting, weighing, through the adaptive PID control algorithm designed by the above trained LSTM model, the real-time value of each group of counting, weighing is calculated.

[0107] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. An adaptive PID control method for a cableway detachment grip, characterized in that, The method includes the following steps: Collect the vibration signal of the steel wire rope during the operation of the cable car, and record the cycle length of one operation cycle of the cable car and the running time between adjacent stations of the cable car in each operation cycle. The vibration signal is decomposed into several component signals by EMD, and each component signal is divided into several segment signals and several time period signals according to the period duration and running time. Based on the distribution of extreme points in each segment signal of each component signal, the key point curves of each time period signal are obtained. Then, the matching results of the key point curves between the time period signals of the same order of adjacent segment signals are analyzed to obtain the relative deviation between the time period signals of adjacent segment signals of the same component signal. Based on the relative deviation between time period signals of different segments of the same component signal, several stable time period signals of each component signal are obtained by screening; based on the distribution of stable time period signals in each component signal, weights are constructed, and the original vibration signals corresponding to the time intervals of stable time period signals in each segment signal are weighted and fused to obtain the inherent vibration information of the running cycle, and then obtain the random vibration information of each running cycle. Based on the inherent vibration information of each operating cycle and the correlation with the original vibration signal, the loss function of the LSTM model is adjusted and trained to predict the load value in real time and perform PID feedforward control.

2. The adaptive PID control method for a cableway detachment gripper according to claim 1, characterized in that, The specific method for obtaining several component signals from the vibration signal through EMD decomposition includes: The vibration signal of the wire rope is decomposed by EMD to obtain several IMF components, and each IMF component is used as a component signal.

3. The adaptive PID control method for a cableway detachment gripper according to claim 1, characterized in that, The specific method for dividing each component signal into several segment signals and several time-period signals based on the period duration and runtime is as follows: The average duration of each cycle of all cable car operations is taken as the first segment duration. For the duration of the same order in each cycle, the average duration of that order in each cycle is taken as the second segment duration. Several second segment durations are obtained in each cycle and arranged in order. For any component signal, the component signal is divided into several segment signals by the first segment duration; for any segment signal, the segment signal is segmented in the order of the second segment duration, starting from the first data point, and then segmented one second segment duration at a time. After acquiring a segment signal for each second segment duration, the signal is segmented in the order of the next second segment duration. By segmenting the signal in the order of each second segment duration, several time period signals of the component signal and the segment signal are obtained.

4. The adaptive PID control method for a cableway detachment gripper according to claim 1, characterized in that, The key point curves of the signals in each time period are obtained using the following method: For any component signal and any segment of the signal, obtain several extreme points in the segment, connect the extreme points that are adjacent in time sequence to obtain the extreme point curve of the segment, and truncate the extreme point curve of each time period in the segment to obtain the key point curve of each time period.

5. The adaptive PID control method for a cableway detachment gripper according to claim 4, characterized in that, The specific method for obtaining the relative deviation between time-segment signals of adjacent segments with the same component signal is as follows: For any two time period signals in any order of any two adjacent segment signals in the component signal, DTW matching is performed on the key point curves of the two time period signals, and the corresponding extreme points in the key point curves of the two time period signals are matched to obtain several pairs of matched extreme points. The time corresponding to the first data point in each segment signal is taken as the zero time of each segment signal. The time points corresponding to each extreme point in the corresponding segment signal are obtained, and the corresponding matching time point pairs are obtained for each matching extreme point pair. For any pair of matching time points between any two adjacent segment signals in any component signal, the absolute value of the difference between the two time points is obtained as the basic time deviation, and the basic time deviation corresponding to each pair of matching time points is obtained. Using the baseline time deviations as the independent variables of the cross-correlation function, a cross-correlation function is constructed for the key point curves of the signals from the two time periods. The expression is: ； The domain of the cross-correlation function integral is... The value range is the base time deviation corresponding to each matching time point. Indicates the number of matching time point pairs. Indicates the first The time point in the corresponding time period of the previous segment signal in the adjacent segment signal of the matching time point pair. Indicates the first The time point in the corresponding time period of the next segment signal in the adjacent segment signal of the matching time point pair. This represents the signal value in the corresponding time period of the preceding segment of the signal in an adjacent segment. This represents the signal value in the corresponding time period of the next segment of the adjacent segmented signal. It is an exponential function with the natural constant as its base. The attenuation coefficient; Obtain the base time deviation corresponding to the maximum value of the cross-correlation function output, and use this base time deviation as the final time deviation between the two time periods; obtain the final time deviation between the two time periods of each order of adjacent segment signals in the component signal, perform linear normalization on all final time deviations, and use the result as the relative deviation between the corresponding two time periods.

6. The adaptive PID control method for a cableway detachment gripper according to claim 1, characterized in that, The specific method for obtaining several stable time periods of each component signal through the filtering process is as follows: For any segment signal in any component signal, if at least one relative deviation between the time segment signal and the time segment signal of the same order in the preceding and following segments is less than or equal to a deviation threshold, the time segment signal is regarded as a stable time segment signal of the component signal; threshold judgment is performed on the time sequence signal of each segment signal in the component signal to obtain several stable time segment signals of the component signal. If there exists a time period signal in any sequence that is not identified as a stable time period signal in any of the segment signals, obtain the mean of the relative deviations of the respective time period signals in each segment signal of that sequence in the component signal, and take the time period signal corresponding to the minimum value among all the obtained mean values ​​of relative deviations as a stable time period signal.

7. The adaptive PID control method for a cableway detachment gripper according to claim 1, characterized in that, The specific method for constructing weights based on the distribution of signals in each component signal during stable periods includes: For any stable time period signal in any component signal, obtain the number of component signals whose corresponding time period signals are stable time period signals in other component signals besides the stable time period signal. Add 1 to the number to obtain the sum as the occurrence frequency of the stable time period signal. Take the ratio of the occurrence frequency to the number of component signals as the occurrence frequency of the stable time period signal. The frequency of occurrence of all stable time period signals in any sequence of the component signal is weighted and normalized, and the result is used as the confidence weight of each stable time period signal.

8. The adaptive PID control method for a cableway detachment gripper according to claim 7, characterized in that, The specific method for obtaining the inherent vibration information of the operating cycle and then acquiring the random vibration information of each operating cycle includes: The original vibration signals under the time interval of each stable period signal are weighted and summed according to the confidence weight, and the result is used as the fixed period information of that order; For each sequence, information of a fixed time period is obtained. The information of the fixed time period is combined in sequence to obtain a combined signal of a time range corresponding to the segmented signal. The time range corresponding to each segmented signal is taken as each running cycle. The combined signal of this segment is recorded as the vibration inherent information of the running cycle. Obtain a segment of vibration signal corresponding to the original vibration signal for each operating cycle, and use it as the segmented vibration signal for each operating cycle; for any segmented vibration signal of an operating cycle, use the difference signal obtained by subtracting the segmented vibration signal from the inherent vibration information of the operating cycle as the random vibration information of that operating cycle.

9. The adaptive PID control method for a cableway detachment gripper according to claim 8, characterized in that, The method for adjusting the loss function of the LSTM model and training it based on the correlation between the inherent vibration information of each operating cycle and the original vibration signal includes the following specific methods: For any given operating cycle, the Pearson correlation coefficient is calculated between the inherent vibration information of that operating cycle and the segmented vibration signal of that operating cycle, and is used as the information weight of that operating cycle. The difference obtained by subtracting the information weight from 1 is used as the random weight of that operating cycle. An LSTM model is constructed by building a training dataset based on segmented vibration signals from each running cycle of the wire rope vibration signal. This dataset is used as the input to the LSTM model, and the output during training is the signal prediction value. For any segmented vibration signal in a given motion cycle, the output is the signal prediction value for that motion cycle when used as the input to the LSTM model. The L2 norms of the intrinsic vibration information and the intrinsic information in the signal prediction value for that motion cycle, as well as the L2 norms of the random vibration information and the random information in the signal prediction value for that motion cycle, are obtained respectively. The information proportion is used as the weight of the L2 norm corresponding to the intrinsic information, and the random proportion is used as the weight of the L2 norm corresponding to the random information. A loss function is constructed by weighted summation. The LSTM model is then trained using segmented vibration signals from each motion cycle to obtain a trained LSTM model.

10. An adaptive PID control system for a cableway detachment grip, the system comprising a stud sleeve, a rod end bearing, a detachment grip, a platform frame, a proximity switch, a switch frame, and an electrical control system for a detachment grip test bench, wherein the electrical control system for the detachment grip test bench includes a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the adaptive PID control method for cableway detachment grippers as described in any one of claims 1-9.

Images