CN107942658B - Method and system for predicting rotation period of circular knitting machine by adopting adaptive filter - Google Patents

Method and system for predicting rotation period of circular knitting machine by adopting adaptive filter Download PDF

Info

Publication number
CN107942658B
CN107942658B CN201711103444.8A CN201711103444A CN107942658B CN 107942658 B CN107942658 B CN 107942658B CN 201711103444 A CN201711103444 A CN 201711103444A CN 107942658 B CN107942658 B CN 107942658B
Authority
CN
China
Prior art keywords
period
iteration
knitting machine
circular knitting
prediction
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201711103444.8A
Other languages
Chinese (zh)
Other versions
CN107942658A (en
Inventor
谢维波
刘涛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huaqiao University
Original Assignee
Huaqiao University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Huaqiao University filed Critical Huaqiao University
Priority to CN201711103444.8A priority Critical patent/CN107942658B/en
Publication of CN107942658A publication Critical patent/CN107942658A/en
Application granted granted Critical
Publication of CN107942658B publication Critical patent/CN107942658B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
    • G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Filters That Use Time-Delay Elements (AREA)
  • Image Analysis (AREA)

Abstract

The invention relates to a method and a system for predicting the rotation period of a circular knitting machine by adopting a self-adaptive filter, aiming at the phenomenon of unstable rotation period in the actual production of the circular knitting machine in the textile industry, the filter combines the advantages of an RLS period prediction algorithm and a single-step period prediction algorithm, and can effectively and accurately predict the rotation period in the actual production of the circular knitting machine. The method can quickly and accurately predict the period of the large circular knitting machine, reduces the accurate snapshot difficulty of a defect detection vision system, and improves the positioning accuracy of the defects of the gray cloth.

Description

Method and system for predicting rotation period of circular knitting machine by adopting adaptive filter
Technical Field
The invention relates to the field of production measurement and control of circular knitting machines in the textile industry, in particular to a circular knitting machine rotation period prediction method and a circular knitting machine rotation period prediction system adopting an adaptive filter, which are applied to a defect detection system with complex production environment and unstable circular knitting machine rotation state.
Background
In the traditional production of a circular knitting machine, the defect cloth in production is often monitored by adopting a manual detection mode, but the traditional method is easily interfered by the subjectiveness of detection workers and has low efficiency. Along with the popularization of industrial video monitoring and the development of video image processing technology, more and more researches on intelligent defect detection based on machine vision appear in the visual field of people. The conventional intelligent defect detection research is performed around improving the accuracy of a defect detection algorithm and reducing the time complexity of the algorithm, but the application of the algorithm to the actual industrial production is still limited by the production environment. In a common defect detection system, video image information is acquired in real time by adopting frame-by-frame analysis and timing snapshot, and the image information is rapidly analyzed to obtain an analysis result. The defect information of the gray cloth has low frequency, a large amount of image information can be wasted by adopting a frame-by-frame analysis method, and due to the influence of the actual production environment of the circular knitting machine and the limitation of the long-term production life condition of the circular knitting machine, the setting of proper time by adopting a timing snapshot method becomes extremely difficult.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a method and a system for predicting the rotation period of a circular knitting machine by adopting a self-adaptive filter, which have the advantages of high prediction speed, high prediction accuracy, strong field adaptability and self-learning capability; the actual period is obtained by analyzing the specific serial port signal, and the actual period is used as a learning object of the self-adaptive filter, so that the filter parameters which accord with the current operation rule of the circular knitting machine are self-learned, and the rotation period of the circular knitting machine can be effectively predicted according to different working environments of the circular knitting machine.
The technical scheme adopted by the invention for solving the technical problems is as follows:
a method for predicting the rotation period of a circular knitting machine by adopting an adaptive filter comprises the following steps:
receiving serial port signals sent by the proximity switch in real time, and obtaining the actual rotation period RC of the large circular knitting machine by calculating the time difference of the two latest effective serial port signalsi(ii) a Wherein i represents the rotation times of the circular knitting machine, and the iteration cycle times are represented in a filter;
filter parameter learning is carried out by adopting RLS period prediction algorithm, and the prediction period PC of the next turn is calculatedi；
Judging whether the current filter parameter reaches a learning standard condition;
if the current filter parameter does not reach the learning standard condition, the single step period prediction algorithm is adopted to obtain the prediction period PC of the next turniAnd jumping out of the filter iteration process, and waiting for the (i + 1) th iteration;
if the current filter parameter reaches the learning criteria barThe PC is obtained by adopting the RLS period prediction algorithmiPredicted period PC as next revolutioniAnd jumping out of the filter iteration process, and waiting for the (i + 1) th iteration.
Preferably, the RLS periodic prediction algorithm is used for filter parameter learning, and the learning process is as follows:
a. calculating the period error eiThe following are:
ei＝RCi-PCi-1(1)
wherein e isiThe periodic error of the ith iteration is represented by an error feedback signal of an RLS filter, and the periodic error of the ith iteration is represented by two parameters RCiAnd PCi-1Subtracting to obtain the result; RC (resistor-capacitor) capacitoriExpressed as the actual period of the ith iteration, for calculating the period error e as the ith iterationiA first parameter of (a); PC (personal computer)i-1Outputting the obtained prediction period for the i-1 th iteration to calculate the period error e as the i-th iterationiA second parameter of (1);
b. computing a gain vector ki(m), as follows:
Figure BDA0001463806820000021
wherein k isi(m) represents the gain vector of the i-th iteration of size m x 1; λ is a constant, representing a forgetting factor; m is a constant and represents the RLS adaptive filter order; pi-1(m) an inverse matrix of the i-1 st iteration of size m x m; u. ofi-1(m) an input matrix representing the i-1 th iteration of size m x 1;
c. computing an inverse matrix P of an iterationi(m) is as follows
Figure BDA0001463806820000022
Wherein, Pi(m) an inverse matrix of the i-th iteration of size m x m; t represents matrix transposition;
d. calculating weight vector wi(m), as follows:
wi(m)＝wi-1(m)+ki(m)ei(4)
wherein, wi(m) represents a weight vector representing the ith iteration of size m x 1; w is ai-1(m) represents vector values of the weight vectors representing the i-1 th iteration of size m x 1;
e. updating an input matrix ui(m), as follows:
ui(m)＝[RCi；ui-1(1:m-1)](5)
wherein u isi(m) represents the input matrix of the ith iteration of size m x 1, which is for the RCiPerforming m-order sampling composition, and updating u at each iterationi(m), the update mode is an in-out queue mode, i.e. a new RC is input at the head of the queue for each iterationiAnd the u vector values from 1 to m-1 of the i-1 th iteration are pushed to the tail of the queue, and the tail of the queue uiThe value of (m) is represented by ui(m-1) covering;
f. calculating the predicted period PC of the next revolutioniThe following are:
PCi＝wi T(m)ui(m) (6)
wherein, PCiThe prediction period obtained for the output of the ith iteration, i.e. the output of the RLS filter, is used as the period error e of the (i + 1) th iterationi+1The second parameter of (2) is the weight vector w of the ith iterationiTransposed vector of (m) and input vector u of order miAnd (m) is obtained by taking the cross product of the two.
Preferably, in the single-step cycle prediction algorithm, the actual rotation cycle RC is usediPredicted period PC set to the next revolutioni。
Preferably, the judging whether the current filter parameter meets the learning criterion condition includes:
if the error value | e calculated by the RLS period prediction algorithm is adopted in the continuous t periodsiI and the single step period prediction error value eiWhen the proportion theta of the difference η obtained by subtracting the' | which is a negative value is more than or equal to a preset value, the learning standard condition is reached, wherein ei'＝RCi-RCi-1。
Preferably, t is equal to 10, and the preset value is equal to 60%. that is, in 10 consecutive cycles, 6 times of measured η values are negative values, it is determined that the current filter parameter meets the learning standard condition.
Preferably, the method for judging the validity of the serial port signal comprises the following steps: and judging whether the value of the specified field of the serial port signal is equal to a set value or not, and if so, judging that the received serial port signal is valid.
The invention also provides a system for predicting the rotation period of the circular knitting machine by adopting the self-adaptive filter, and the method for predicting the rotation period of the circular knitting machine based on the self-adaptive filter comprises the following steps: the device comprises a lower computer, an upper computer and a serial port transmission line connected between the lower computer and the upper computer; the lower computer comprises a circular knitting machine and a proximity switch; the proximity switch comprises an infrared probe fixed on the inner wall of the machine table of the circular knitting machine and a circumferential outer wall nut capable of performing circumferential motion along with the circular knitting machine; and when the circular knitting machine works for each circle, the proximity switch generates a serial port signal and sends the serial port signal to the upper computer through the serial port transmission line, and the upper computer performs rotation period prediction according to the circular knitting machine rotation period prediction method.
The invention has the following beneficial effects:
(1) in order to accurately snapshot images by adopting a timing snapshot method in the actual production environment of a circular knitting machine, not only the environmental influence of complex actual production is overcome, but also the influence of the service life of the circular knitting machine is overcome, therefore, the invention provides a circular knitting machine rotation period prediction method adopting an adaptive filter, and the adaptive filter suitable for the rotation period prediction of the industrial production of the circular knitting machine is realized by combining an RLS (least square method) period prediction algorithm and a single-step period prediction algorithm; experiments prove that the method for predicting the rotation period of the circular knitting machine by using the adaptive filter can quickly and accurately analyze and process the characteristic serial port signals of the circular knitting machine, predict the rotation period of the circular knitting machine, and is suitable for different working environments of the circular knitting machine;
(2) the RLS period prediction algorithm has certain self-learning capacity, can achieve certain accuracy after the actual rotation period of the circular knitting machine is learned and trained for a period of time, and can keep trained parameters to predict the subsequent rotation period after the actual rotation period of the circular knitting machine is learned and trained, but the RLS period prediction algorithm has certain learning buffer period, and the period predicted by the RLS period prediction algorithm is far away from the actual period during the period, so that the RLS period prediction algorithm cannot be directly used for the timing snapshot operation of a defect detection vision system; the single-step period prediction algorithm assumes that the predicted rotation period of the circular knitting machine is approximately equal to the actual rotation period of the previous rotation, and because the number of times of fluctuation changes of the previous rotation and the next rotation in the actual work of the circular knitting machine is small, the assumed conditions are basically met, the single-step period prediction algorithm can be directly used for the timing snapshot operation of the defect detection visual system, but the algorithm has certain assumed conditions, has no self-adaptive capacity, has no strong period prediction accuracy, and is not suitable for the complex production environment of the circular knitting machine because the RLS period prediction algorithm is not strong. Based on the method, the method and the system for predicting the rotation period of the circular knitting machine by adopting the self-adaptive filter are provided by combining the RLS period prediction algorithm and the single-step period prediction algorithm, have self-adaptive capacity and are suitable for the complex working environment of the circular knitting machine;
(3) the invention is closely combined with a defect detection visual system, can quickly and accurately acquire a prediction period through the real-time analysis of serial port signals and the processing of self-adaptive filtering, and the prediction period can be used for calculating the rotation speed of a circular knitting machine, positioning the image position of the defect detection visual system needing to be snapshoted and positioning the position of the defects in the whole batch of cloth, thereby laying a foundation for realizing an efficient gray cloth detection system.
The present invention will be described in further detail with reference to the drawings and embodiments, but the method and system for predicting the rotation period of a circular knitting machine using an adaptive filter according to the present invention are not limited to the embodiments.
Drawings
FIG. 1 is a schematic diagram of the operation of the adaptive filter RLS period prediction algorithm of the present invention;
FIG. 2 is a block diagram of an update input vector u of the present inventioni(m) a queue pattern diagram;
FIG. 3 is a core diagram of the RLS cycle prediction algorithm of the present invention;
FIG. 4 is a block diagram of a circular machine spin cycle prediction system employing an adaptive filter in accordance with the present invention;
FIG. 5 is a data sample graph of an actual cycle of the present invention;
FIG. 6 is a diagram of the adaptive filter operating mechanism of the present invention;
FIG. 7 is a diagram of the filter error weight (Power) distribution of the present invention;
FIG. 8 is a single step PC error plot of the filter of the present invention;
FIG. 9 is a 16 th order PC error plot for the filter of the present invention;
FIG. 10 is a difference η between the absolute value of the 16 th order error of the cycle (70-79) and the absolute value of the single step error of the present invention;
FIG. 11 is a difference η between the absolute value of the 16 th order error of the present invention and the absolute value of the single step error;
FIG. 12 is a positive-negative statistic of the difference η between the absolute value of the 16 th order error and the absolute value of the single step error of the present invention.
Detailed Description
Specific implementations of the present invention are further described in detail below with reference to the accompanying drawings.
Referring to fig. 1, 2 and 3, the working principle of the RLS period prediction algorithm in the adaptive filter is described as follows.
As shown with reference to figure 1 of the drawings,
Figure BDA0001463806820000051
to ⑤, forming an RLS cycle prediction iterative process according to
Figure BDA0001463806820000052
The sequence of the process proceeds to ⑤, with the filter input being the actual period RCiThe output is preCycle measurement PCi。
Figure BDA0001463806820000053
- ① - ② process for solving cycle error e of i-th iterationiTask of (2), input signal RC of process ①iExpressed as the actual period and process of the large circular knitting machine
Figure BDA0001463806820000054
Input signal PCiExpressed as the prediction period, RC, of the previous iteration outputiAnd PCiSubtracting to obtain a periodic error ei。
④ the process completes the input vector u of the iterationi(m) update, the update mode adopts a queue mode, as shown in fig. 2, and a new RC is input into the filter every time the filter is iteratediRC at the ith iterationiWill be uiThe head (front) element of the (m) enters the queue, the m-1 time queue entering elements before the ith iteration move to the tail (rear) of the queue in sequence according to the queue order, and ui(m) will be updated continuously as the iteration progresses.
③ and ⑤ processes construct the core content of RLS operation, namely solving weight vector wi(m) and calculating the prediction period PC i③ the process completes the weight vector w of the current iterationi(m) update, wi(m) is the key to achieving periodic prediction of the RLS adaptive filter, which can be based on the input matrix ui(m) and actual period data RCiContinuously correcting w by self-learningi(m) weight vectors up to wi(m) tends to be stable, ⑤ process completes the prediction period PC of this iterationiOutput, its value is weight vector wi(m) transposed vector and input vector uiAnd (m) accumulating the vector products to obtain the product.
Referring to fig. 3, "1", "2" and "3" in the figure respectively represent gain vectors ki(m), inverse matrix Pi(m) and an input vector ui(m) an update procedure. K according to RLS period prediction algorithm formula (2)i(m) the update needs to be applied to this iterationInput vector u before updatei-1(m) and inverse matrix Pi-1(m), and a forgetting factor λ; p according to RLS period prediction algorithm formula (3)i(m) the update requires the inverse matrix P to be used before the current iterative updatei-1(m) and an input vector ui-1(m), updated gain vector ki(m), and a forgetting factor λ; according to RLS period prediction algorithm formula (4), wi(m) updating requires the weight vector w before the current iteration updatingi-1(m) updated gain vector ki(m) and
Figure BDA0001463806820000062
periodic error e calculated by the process from- ① - ②i；ui(m) the update occurs at wi(m) after updating, finally, according to RLS period prediction algorithm formula (6), ⑤ process calculation is completed to obtain the prediction period PC of the current iterationi。
The detailed operation of the system for predicting the rotation period of the circular knitting machine by adopting the self-adaptive filter is described as follows.
Referring to fig. 4, the circle machine period prediction system is composed of an upper computer, a lower computer and a connecting line between the upper computer and the lower computer. Specifically, the lower computer comprises a circular knitting machine and a proximity switch, namely a hardware model of the circular knitting machine; the upper computer comprises a proximity switch serial port signal analysis module and an adaptive filter module, namely a software model of the circular knitting machine. In the lower computer, "1" and "2" form a proximity switch, "1" is the infrared probe, is fixed at the big circular knitting machine board inner wall, "2" is the circumference outer wall nut that can block infrared light and be fixed and can carry out circular motion along with big circular knitting machine.
Specifically, the proximity switch signal is a 9-byte serial port signal, the protocol of the proximity switch signal is shown in table 1, and byte information adopts a 16-system format, namely a 00-FF pattern. The 'header' adopts uniform '005A' information, so that the uniformity and the safety of the equipment are ensured; the user of the 'equipment address' can customize information according to the number of the circular knitting machines, so that unified equipment management is facilitated; the 'command' uniformly adopts '01' information, mainly for distinguishing command information except serial port signals; the "reserved" 4 bytes are temporarily left untreated; the 'checksum' is the sum of the previous 8 bytes of information and can be used for verifying the accuracy of the proximity switch signal by the upper computer.
TABLE 1
Figure BDA0001463806820000061
Figure BDA0001463806820000071
Every work circle of big circular knitting machine, proximity switch all can produce a serial signals to upload to the host computer through RS485/RS232 agreement, the host computer passes through USB commentaries on classics serial ports and receives serial signals, and then accomplishes serial signals analysis on the host computer, and acquires actual cycle RC data set process.
The detailed data acquisition process of the system is as follows.
Step one, generating and uploading a proximity switch serial port signal. When the '2' approaches to the '1', an interrupt pulse signal can be generated, the signal can generate a 9-byte proximity switch signal in the lower computer, and the serial port signal can be uploaded to the upper computer through a serial port transmission line (RS485/RS232) and a USB (universal serial bus) transfer serial port;
and step two, analyzing the proximity switch serial port signal. After the upper computer receives the serial port signal, the serial port signal is checked and analyzed, and whether the current serial port signal is correct or not is judged, so that the upper computer is used for eliminating wrong serial port signals uploaded by the lower computer due to interference of environmental factors. When the proximity switches can work normally, when the upper computer receives a second proximity switch signal, the circular knitting machine is considered to rotate for a circle, at the moment, the upper computer can calculate the rotation period of one rotation of the circular knitting machine according to the time of the received two proximity switch signals, the rotation period is marked as RC, the current second proximity switch signal is used as a first proximity switch signal of the next rotation, and the RC of each rotation can be calculated in sequence according to the process;
and step three, collecting and manufacturing an RC data set. According to the first step and the second step, the RC collection is carried out on the textile production operation in the real circular knitting machine environment of the textile factory by the design software tool, because many sudden factors exist in the real textile environment, such as broken needles, broken threads, manual shutdown of workers and the like, unreliable influence is caused on the RC data set, and absolute influence is caused on the size of the RC data set, the experimental selection is carried out on the experimental circular knitting machine for multiple times, and then the RC data set with the size of 4578 is selected through comparative analysis, and the comparison and analysis are shown in figure 5.
In order to construct a working mechanism of a circular knitting machine adaptive filter, the following two part of experiments are designed in the embodiment, namely an experiment for finding the optimal order m of an RLS period prediction algorithm; and experiment II, comparing the RLS period prediction algorithm for confirming the order with the single-step period prediction algorithm. The experimental data were actual period RC data samples of size 4578.
The RLS period prediction algorithm needs to be applied to P0(m) giving an initial value, wherein the value range of the RLS forgetting factor is 0.9 according to the literature<λ<1, repeated experiments show that lambda is 0.98, and P is set for any one group of RC data0(m) (1/λ) × eye, eye represents an identity matrix of size m × m.
In the first experiment, in order to find the optimal order of the RLS period prediction algorithm, an experiment that the RLS error weight (Power) changes along with the order N is designed. Error weighting (Power) using the periodic error eiThe square number of (i.e. Power ═ e)i*ei. The order N ranges from 2 to 30. The experimental contents are that 4578 iterations of RLS period prediction algorithm are designed, and along with the iteration, eiAnd (4) the error weight tends to be in a stable state, then the Power value is calculated, the iteration process is repeated according to different values of the order N, and the relation change of the error weight Power and the order N can be obtained.
In experiment two, the single-step cycle prediction is to predict the actual cycle RC of the ith iterationiPrediction period PC as the i-1 st iterationiFinally, 4578 sets of prediction periods are obtained; in order to confirm RLS period prediction of the order, a program can be designed on a Matlab platform according to the working principle of an RLS period prediction algorithm in an adaptive filter, an RC data set is tested, and 4578 groups of prediction periods are obtained. The experimental results shown in fig. 8, 9, 10, 11 and 12 can be plotted according to the above experimental contents.
The experimental results and analyses were as follows:
according to experiment one, see fig. 7, a distribution diagram of error weights (powers) of the RLS filter according to the order N is given, and since the 1 st order PC can be approximately regarded as a single step PC, the 1 st order PC is not given in the distribution diagram. Along with the increase of the N order, the Power value presents a distribution situation that the Power value firstly drops rapidly and then rises slowly, a minimum value N-16 can be found as the minimum value of the Power, namely the optimal order of the RLS period prediction algorithm is considered to be 16 orders, namely m-MaxN16. Experiment two was performed on a 16 th order RLS filter.
According to the second experiment, the advantages and the disadvantages of the RLS period N-order prediction algorithm and the single-step period prediction algorithm are analyzed from different angles respectively in the steps of FIG. 8, FIG. 9, FIG. 10, FIG. 11 and FIG. 12, and the conclusion that the adaptive filter can be used for industrial field production of the circular knitting machine is given.
Fig. 8 and 9 show the experimental results of filter single step PC error and filter 16 th order PC error, respectively. From fig. 8 it can be observed that: the filter single step PC error value can exhibit irregular fluctuations around the 0 axis, ranging from-170 to 120; the larger the fluctuation amplitude is, the farther the fluctuation amplitude is from the axis 0, the worse the single-step period prediction performance is, 107 periods are distributed in different period regions in the graph, and 7 times are distributed in different period regions in the graph, wherein the number of the periods exceeds 50; from fig. 9 it can be observed that: the filter 16-order PC error value also shows irregular fluctuation around the 0 axis, and the fluctuation range is-290-230; there are 20 cycles in excess of 50 in the figure, and 12/20 before cycle 80, 4 in excess of 100, distributed at cycle numbers 2, 6, 7, 8. From the above phenomena, the distance between the point C and the point B is greater than that between the point C and the point A, the prediction result of the 16-order period is generally more accurate than that of the single-step period, the abnormal prediction value can be controlled in the stage of acquiring the standard learning period in the 80 th period, and the prediction effect of the 16-order period is better than that of the single-step period.
See FIG. 10 for the difference between the absolute value of the 16 th order error and the absolute value of the single step error for the period (70-79) as observed in FIG. 10FIG. 11 shows that the absolute value of the 16 th order error is η, the positive value shows that the absolute value of the 16 th order error is larger than the absolute value of the single step error, the negative value shows that the absolute value of the 16 th order error is smaller than the absolute value of the single step error, the positive value shows that the absolute value of the 16 th order error is smaller than the absolute value of the single step error, the figure 11 shows that the positive value of η is far smaller than the negative value, the positive value is more than the negative value before the 80 th order, the 80-300 th order positive value is equal to the negative value statistical number, and the error of the 16 th order prediction period is similar to the error of the single step prediction period, the figure 12 shows that the experimental result of the positive and negative statistics of the absolute value of the 16 th order error is equal to the absolute value of the single step error, the 1 in the figure shows that the difference is positive, the 2 shows that the difference is negative, the figure 12 shows that the positive and negative values (1) η 1 and 352 have the positive and negative proportions of the group 328, and the negative value η 2 in the total 4575, the group of the group
Figure BDA0001463806820000091
The prediction accuracy of the 16 th order cycle is improved by 40.46% compared with the prediction accuracy of the single step cycle.
Through the experimental analysis, the 16-order RLS period prediction algorithm can effectively and more accurately predict the rotation period of the circular knitting machine, but a self-learning process of the vector weight needs to be carried out through a buffer period of about 80 before the stable PC is predicted. The single-step period prediction algorithm is adopted, although self-learning cannot be achieved and adaptability is not achieved, the single-step period prediction algorithm can be used as a temporary prediction period in the RLS buffer period, and normal operation of timing snapshot work is guaranteed. Referring to the working mechanism of the circular knitting machine adaptive filter shown in fig. 6, the working mechanism can be used in a complex production environment of the circular knitting machine, namely, an RLS buffer period and an RLS prediction period are set, wherein the RLS buffer period is formed by a single step prediction period, and the RLS prediction period is formed by a prediction period obtained by an RLS period prediction algorithm with a determined order.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (4)

1. A method for predicting the rotation period of a circular knitting machine by adopting an adaptive filter is characterized by comprising the following steps:
receiving serial port signals sent by the proximity switch in real time, and obtaining the actual rotation period RC of the large circular knitting machine by calculating the time difference of the two latest effective serial port signalsi(ii) a Wherein i represents the rotation times of the circular knitting machine, and the iteration cycle times are represented in a filter;
filter parameter learning is carried out by adopting RLS period prediction algorithm, and the prediction period PC of the next turn is calculatedi；
Judging whether the current filter parameter reaches a learning standard condition;
if the current filter parameter does not reach the learning standard condition, the single step period prediction algorithm is adopted to obtain the prediction period PC of the next turniAnd jumping out of the filter iteration process, and waiting for the (i + 1) th iteration;
if the current filter parameter reaches the learning standard condition, the RLS periodic prediction algorithm is adopted to calculate to obtain the PCiPredicted period PC as next revolutioniAnd jumping out of the filter iteration process, and waiting for the (i + 1) th iteration;
the RLS period prediction algorithm is adopted to learn filter parameters and calculate the prediction period PC of the next turniThe method comprises the following steps:
a. calculating the period error eiThe following are:
ei＝RCi-PCi-1(1)
wherein e isiThe periodic error of the ith iteration is represented by an error feedback signal of an RLS filter, and the periodic error of the ith iteration is represented by two parameters RCiAnd PCi-1Subtracting to obtain the result; RC (resistor-capacitor) capacitoriExpressed as the actual period of the ith iteration, for calculating the period error e as the ith iterationiA first parameter of (a); PC (personal computer)i-1Outputting the obtained prediction period for the i-1 th iteration to calculate the period error e as the i-th iterationiA second parameter of (1);
b. computing a gain vector ki(m), as follows:
Figure FDA0002451613630000011
wherein k isi(m) denotes the gain vector of the i-th iteration of size m x 1, typically kiThe larger the value (m) is, the stronger the correction capability of the RLS period prediction algorithm is; λ is a constant, is an empirical coefficient, and represents a forgetting factor; m is a constant, is an empirical coefficient, represents the RLS adaptive filter order, and can be obtained by calculation according to the collected actual rotation period of the current circular knitting machine; pi-1(m) an inverse matrix of the i-1 st iteration of size m x m; u. ofi-1(m) an input matrix representing the i-1 th iteration of size m x 1;
c. computing an inverse matrix P of an iterationi(m) is as follows
Figure 2
Wherein, Pi(m) an inverse matrix of the i-th iteration of size m x m; t represents matrix transposition;
d. calculating weight vector wi(m), as follows:
wi(m)＝wi-1(m)+ki(m)ei(4)
wherein, wi(m) represents a weight vector representing the ith iteration of size m x 1; w is ai-1(m) represents vector values of the weight vectors representing the i-1 th iteration of size m x 1;
e. updating an input matrix ui(m), as follows:
ui(m)＝[RCi；ui-1(1:m-1)](5)
wherein u isi(m) represents the input matrix of the ith iteration of size m x 1, which is for the RCiPerforming m-order sampling composition, and updating u at each iterationi(m), the update mode is an in-out queue mode, i.e. a new RC is input at the head of the queue for each iterationiAnd the u vector values from 1 to m-1 of the i-1 th iteration are pushed to the tail of the queue, and the tail of the queue uiThe value of (m) is represented by ui(m-1) covering;
f. calculating the predicted period PC of the next revolutioniThe following are:
PCi＝wi T(m)ui(m) (6)
wherein, PCiThe prediction period obtained for the output of the ith iteration, i.e. the output of the RLS filter, is used as the period error e of the (i + 1) th iterationi+1The second parameter of (2) is the weight vector w of the ith iterationiTransposed vector of (m) and input vector u of order mi(m) obtaining a vector product;
in the single-step period prediction algorithm, the actual rotation period RC is usediPredicted period PC set to the next revolutioni；
The judging whether the current filter parameter meets the learning standard condition includes:
if the error value | e calculated by the RLS period prediction algorithm is adopted in the continuous t periodsiI and the single step period prediction error value eiWhen the proportion theta of the negative value of the difference obtained by subtracting the' | is more than or equal to a preset value, the learning standard condition is met; wherein e isi'＝RCi-RCi-1。
2. The method of claim 1, wherein t is equal to 10, and the predetermined value is equal to 60%.
3. The method for predicting the rotation period of the circular knitting machine by using the adaptive filter as claimed in claim 1, wherein the method for judging the effectiveness of the serial port signal comprises the following steps: and judging whether the value of the specified field of the serial port signal is equal to a set value or not, and if so, judging that the received serial port signal is valid.
4. A circular knitting machine rotation period prediction system using an adaptive filter, based on the circular knitting machine rotation period prediction method using an adaptive filter of any one of claims 1 to 3, comprising: the device comprises a lower computer, an upper computer and a serial port transmission line connected between the lower computer and the upper computer; the lower computer comprises a circular knitting machine and a proximity switch; the proximity switch comprises an infrared probe fixed on the inner wall of the machine table of the circular knitting machine and a circumferential outer wall nut capable of performing circumferential motion along with the circular knitting machine; and when the circular knitting machine works for each circle, the proximity switch generates a serial port signal and sends the serial port signal to the upper computer through the serial port transmission line, and the upper computer performs rotation period prediction according to the circular knitting machine rotation period prediction method.
CN201711103444.8A 2017-11-10 2017-11-10 Method and system for predicting rotation period of circular knitting machine by adopting adaptive filter Active CN107942658B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711103444.8A CN107942658B (en) 2017-11-10 2017-11-10 Method and system for predicting rotation period of circular knitting machine by adopting adaptive filter

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711103444.8A CN107942658B (en) 2017-11-10 2017-11-10 Method and system for predicting rotation period of circular knitting machine by adopting adaptive filter

Publications (2)

Publication Number Publication Date
CN107942658A CN107942658A (en) 2018-04-20
CN107942658B true CN107942658B (en) 2020-06-26

Family

ID=61933766

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711103444.8A Active CN107942658B (en) 2017-11-10 2017-11-10 Method and system for predicting rotation period of circular knitting machine by adopting adaptive filter

Country Status (1)

Country Link
CN (1) CN107942658B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111220092B (en) * 2019-12-10 2021-04-23 东南大学 Method for constructing fringe filter in optical measurement

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002013064A (en) * 2000-06-30 2002-01-18 Fukushin Kogyo Kk Method and apparatus for caustic reduction of polyester fiber
TW554102B (en) * 2002-10-24 2003-09-21 Kuen-Shie Tsai Manufacturing method for double knitting jacquard weave fabric appearing color image
CN101515166A (en) * 2009-03-19 2009-08-26 杭州嘉拓科技有限公司 Device for monitoring yarn moving state and monitoring method for same
EP2149630A1 (en) * 2008-07-30 2010-02-03 Deimo S.p.A. Circular knitting machine
CN102877153A (en) * 2011-07-14 2013-01-16 广东柏堡龙股份有限公司 Negative ion cool lining and method for preparing same
CN102879401A (en) * 2012-09-07 2013-01-16 西安工程大学 Method for automatically detecting and classifying textile flaws based on pattern recognition and image processing
CN104007423A (en) * 2014-05-27 2014-08-27 电子科技大学 Sky wave radar sea clutter suppression method based on chaos sequence prediction
CN104021395A (en) * 2014-06-20 2014-09-03 华侨大学 Target tracing algorithm based on high-order partial least square method
CN106592084A (en) * 2017-02-22 2017-04-26 武汉纺织大学 Yarn conveying device and method for electromagnetic driving circular knitting machine
CN107256545A (en) * 2017-05-09 2017-10-17 华侨大学 A kind of broken hole flaw detection method of large circle machine

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
ES2220898T3 (en) * 1999-08-11 2004-12-16 Toyo Boseki Kabushiki Kaisha BALISTIC UNMATERIAL THAT INCLUDES HIGH RESISTANCE POLYETHYLENE FIBERS.
JP5898499B2 (en) * 2011-01-21 2016-04-06 日本バイリーン株式会社 Nonwoven fabric and method for producing the same

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002013064A (en) * 2000-06-30 2002-01-18 Fukushin Kogyo Kk Method and apparatus for caustic reduction of polyester fiber
TW554102B (en) * 2002-10-24 2003-09-21 Kuen-Shie Tsai Manufacturing method for double knitting jacquard weave fabric appearing color image
EP2149630A1 (en) * 2008-07-30 2010-02-03 Deimo S.p.A. Circular knitting machine
CN101515166A (en) * 2009-03-19 2009-08-26 杭州嘉拓科技有限公司 Device for monitoring yarn moving state and monitoring method for same
CN102877153A (en) * 2011-07-14 2013-01-16 广东柏堡龙股份有限公司 Negative ion cool lining and method for preparing same
CN102879401A (en) * 2012-09-07 2013-01-16 西安工程大学 Method for automatically detecting and classifying textile flaws based on pattern recognition and image processing
CN104007423A (en) * 2014-05-27 2014-08-27 电子科技大学 Sky wave radar sea clutter suppression method based on chaos sequence prediction
CN104021395A (en) * 2014-06-20 2014-09-03 华侨大学 Target tracing algorithm based on high-order partial least square method
CN106592084A (en) * 2017-02-22 2017-04-26 武汉纺织大学 Yarn conveying device and method for electromagnetic driving circular knitting machine
CN107256545A (en) * 2017-05-09 2017-10-17 华侨大学 A kind of broken hole flaw detection method of large circle machine

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
BEST LEAST SQUARES SOLUTION FOR PRONY MODEL;Weibo Xie等;《Proceedings of 2007 International Symposium on Intelligent Signal Processing and Communication Systems》;20071201;第292-295页 *
基于ARM9提花控制系统研究与实现;刘玲;《中国优秀硕士学位论文全文数据库 工程科技I辑》;20111215(第S2期);第B024-41页 *
基于机器视觉的白胚布疵点检测方法综述;郑一露等;《信息技术与信息化》;20151231;第163-165页 *

Also Published As

Publication number Publication date
CN107942658A (en) 2018-04-20

Similar Documents

Publication Publication Date Title
EP3382485B1 (en) Controller, control program, control system, and control method
CN110971677B (en) Electric power internet of things terminal equipment side channel safety monitoring method based on countermeasure reinforcement learning
CN104615856B (en) Gas consumption amount forecast model method for building up and device based on hot-blast stove group
CN110363355B (en) Cloud-edge collaborative forecasting system and method for alumina production index
CN107942658B (en) Method and system for predicting rotation period of circular knitting machine by adopting adaptive filter
CN105675320B (en) A kind of mechanical system running status method for real-time monitoring based on acoustic signal analysis
CN113568960A (en) Real-time analysis method and system for data of Internet of things and computer readable storage medium
CN110769453B (en) Multi-modal monitoring data dynamic compression control method under unstable network environment
CN112950908B (en) Data monitoring and early warning method, system, medium and electronic terminal
Yu et al. A novel modified PID controller applied to temperature control with self-tuning ability
CN116400639B (en) PLC (programmable logic controller) collected data intelligent cleaning method and system
CN112732817A (en) Big data processing method of Internet of things system
CN109656229B (en) Construction method of robot end performance prediction model based on GA-RBF network
CN112123333A (en) Multi-joint robot collision detection method and device
CN111108738B (en) Data processing device, data analysis device, data processing system, and method for processing data
CN115042191B (en) Pre-training model fine-tuning training method and device, electronic equipment and storage medium
CN116704729A (en) Industrial kiln early warning system and method based on big data analysis
CN111510109A (en) Signal filtering method, device, equipment and medium
Zhaosheng The optimization design of PID controller parameters based on particle swarm optimization
CN115407736A (en) Method, device, equipment and medium for improving product yield
CN114378835B (en) Robot control system based on image recognition and control method thereof
Norgren et al. Industrial evaluation of process control using non-periodic sampling
CN111522224A (en) Parameter self-tuning PLC control method based on prediction PI and bias relay feedback
CN110597203A (en) Fault diagnosis method based on multi-GPU parallel CRPF
CN116561598B (en) CPS intelligent manufacturing management platform big data processing method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant