CN102176117A - Intelligent processing method for dynamic weighing signal of fruit high-speed sorting system - Google Patents
Intelligent processing method for dynamic weighing signal of fruit high-speed sorting system Download PDFInfo
- Publication number
- CN102176117A CN102176117A CN 201110022875 CN201110022875A CN102176117A CN 102176117 A CN102176117 A CN 102176117A CN 201110022875 CN201110022875 CN 201110022875 CN 201110022875 A CN201110022875 A CN 201110022875A CN 102176117 A CN102176117 A CN 102176117A
- Authority
- CN
- China
- Prior art keywords
- model
- sampling
- data
- speed
- fruit
- 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.)
- Granted
Links
- 235000013399 edible fruits Nutrition 0.000 title claims abstract description 131
- 238000005303 weighing Methods 0.000 title claims abstract description 71
- 238000003672 processing method Methods 0.000 title claims abstract description 18
- 238000005070 sampling Methods 0.000 claims abstract description 104
- 238000001914 filtration Methods 0.000 claims abstract description 31
- 238000012549 training Methods 0.000 claims abstract description 31
- 238000000034 method Methods 0.000 claims abstract description 12
- 239000011159 matrix material Substances 0.000 claims description 16
- 230000003068 static effect Effects 0.000 claims description 15
- 230000000630 rising effect Effects 0.000 claims description 12
- 230000001360 synchronised effect Effects 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000003860 storage Methods 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 238000012886 linear function Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 8
- 238000004458 analytical method Methods 0.000 abstract 1
- 230000002349 favourable effect Effects 0.000 abstract 1
- 238000007781 pre-processing Methods 0.000 description 5
- 101100190621 Arabidopsis thaliana PLC4 gene Proteins 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 244000141359 Malus pumila Species 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 235000021016 apples Nutrition 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 244000144730 Amygdalus persica Species 0.000 description 1
- 240000000560 Citrus x paradisi Species 0.000 description 1
- 235000006040 Prunus persica var persica Nutrition 0.000 description 1
- 241000220324 Pyrus Species 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000011022 operating instruction Methods 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 235000021017 pears Nutrition 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Landscapes
- Feedback Control In General (AREA)
Abstract
The invention discloses an intelligent processing method for a dynamic weighing signal of a fruit high-speed sorting system. By using the method, the high-precision acquisition and prediction of fruit weight can be realized on the basis of an improved AR (Autoregressive) model. The method comprises the following steps of: acquiring the grading speed in real time; sampling and storing a group of data of standard fruits; filtering and eliminating low-frequency noise in a signal sampling value; reading initial data, modeling by using the improved AR model and calculating orders and parameters of the model by an operation station system; storing the parameters and the orders of the model, which are obtained by training at different sorting speeds, in a file; predicting fruit weight information by using the trained model parameters according to the sorting speed through a control station system; and carrying out precision analysis on the predication result by the operation station system. According to the method, at the grading speed of 20 fruits/second, the weighing precision can reach within 1.5+/-0.5 percent and the optimal orders and parameters of the model can be acquired again after the grading speed is changed. The intelligent processing method has simple structure, low requirement on hardware and favorable extensibility and can be conveniently transplanted for processing high-speed weighing signals in other fields.
Description
Technical Field
The invention relates to an intelligent processing method of a dynamic weighing signal of a high-speed fruit sorting system.
Background
According to statistics, the output of apples, pears, peaches and pomelos in China is the first in the world. In order to improve the economic value of the fruits, the harvested fruits need to be subjected to post-production treatment such as cleaning, grading, packaging and the like. The dynamic weighing technology is one of the key technologies for postpartum treatment, and the accuracy of procedures such as grading, boxing and the like is directly influenced by the precision of dynamic weighing. The level of accuracy of dynamic weighing depends mainly on two aspects: firstly, the mechanical structure design of the fruit cup and the mechanical structure design of the weighing area are carried out; secondly, the design of a weighing signal processing system. The weighing signal processing system generally includes an operator station system and a control station system. The operation station system provides an operation interface for user interaction and consists of a server and a client; the control station system collects field data, receives an operating instruction of the operating station system, calculates and outputs a control instruction of the field equipment, and transmits an important control state to the operating station system. In the high-speed case, the output signal starts to fall before reaching a stable value due to the delay of the system itself, and how the signal processing system obtains high-precision weight information from the existing signal becomes a bottleneck that restricts the system classification speed.
The prior high-speed weighing signal processing method has the following documents:
[1] J.Calpe, E.Soria, M.Martinez, J.V.France, A.Rosado, L.Gomez-Chova, J.Vila High-speed weighing system based DSP [ J ] IEEE.2002.0-7803-.
[2] A new filtering method for improving weighing precision is disclosed, which features high precision but not meeting the requirement of high speed.
[3] The research of a vehicle dynamic weighing system based on neural network adaptive filtering of the Zhang Rui, Lvwenhong, Zhang Ruizi provides a processing method of a vehicle dynamic load signal, the method adopts the adaptive filtering variable step length LMS algorithm based on the neural network to filter the noise of the weighing signal in each frequency band, the speed is high, but the algorithm is complex, and the precision can not meet the requirement of a fruit weighing system.
Therefore, at present, the processing scheme for the high-speed fruit weighing signals at home and abroad is less researched, and the scheme for improving the weighing precision is mostly only aimed at the condition of low grading speed.
Disclosure of Invention
The invention aims to solve the problems that the existing fruit weighing system is low in measurement precision and cannot meet the real-time requirement under the high-speed condition, and provides an intelligent processing method for dynamic weighing signals of a high-speed fruit sorting system.
The intelligent processing method of the dynamic weighing signal of the fruit high-speed sorting system comprises the following steps:
1) when the fruit conveying device is static, sampling to obtain a group of stable values of the standard fruits, then starting to rotate the conveying device, determining a sampling period according to the grading speed, and sampling to obtain a group of data when each standard fruit passes through the weighing sensor;
2) eliminating low-frequency noise in the sampling signal by adopting first-order lag filtering to obtain a preprocessed signal;
3) the operating station system reads the preprocessed data, analyzes the characteristics of the data and obtains an improved AR model according to the real-time requirement;
4) determining parameters and orders of an improved AR model, and storing the model parameters and the orders obtained by training at different grading speeds in a file;
5) the control station system predicts the fruit weight information by using the trained model parameters;
6) and evaluating the precision of the prediction result.
The step 1) is as follows: requires n standard fruits with different weights, under static conditions, forAveraging values obtained after multiple sampling of each standard fruit to serve as final static sampling values of the fruit, and recording the values as follows: b1，b2，…，bn(ii) a In the high speed case, according to the classification speed, the sampling period is determined: let the length of the load-bearing foot of the weighing sensor be L, the grading speed be V, and the time interval from the beginning of rising to the beginning of falling of the signal
The sampling can only be at
The sampling is completed within a time span, and if the sampling frequency of each fruit is m, the sampling period is
(ii) a Finally, sampling according to a set sampling period, and storing the sampling data y of each standard fruit1(1)，y1(2)，…，y1(m) ；y2(1)，y2(2)，…，y2(m) ；…… ；yn（1)，yn(2)，…，yn(m)。
The step 2) is as follows: eliminating low-frequency noise in the sampled data by using first-order lag filtering, defining y (k) as a k-th sampled value, defining x (k-1) as a k-1-th filtered output value, substituting alpha as a filtering coefficient with the value less than 1 into a first-order lag filtering formula
Obtaining a filtering output x (k) corresponding to the kth sampling data; filtering each group of sampling data to obtain a group of data x1(1)，x1(2)，…，x1(m) ；x2(1)，x2(2)，…，x2(m) ；…… ；xn（1)，xn(2)，…，xn(m), as data for model training.
The step 3) is as follows: the output signal of the load cell is approximated as a stationary random time series, and an AR model of the time series is described as: the current value x (n) of the sequence is expressed as a linear function of the past values plus an error term e (n), expressed by equation (1) as follows:
a in equation (1)1，a2，…，apIs the model coefficient; x (n-1), x (n-2), …, x (n-p) are p past values of the signal, and a p-order AR model is formed; assuming that p initial sample values x (1), x (2), …, x (p) are known, and the signal reaches a stable value after N time points, the stable value x (p + N) of the signal is predicted, and the following equation (2) is obtained by recursion using equation (1):
coefficient A in equation (2)1，A2，…，ApThe signal stability value x (p + N) is linearly represented by initial sampling values x (1), x (2), …, x (p); for fruits with different weights, the linear relation between the signal stable value and the initial sampling value is consistent after the fruits pass through the same weighing sensor.
The step 4) is as follows: computing improved AR model orders and parameters using an operator station system
First, the operating station system reads the model training data x1(1)，x1(2)，…，x1(m) ；x2(1)，x2(2)，…，
x2(m) ；…… ；xn（1)，xn(2)，…，xn(m), and a static stability value b1，b2，…，bn，
Stored in a matrix variable, noted:
wherein m represents the number of samples per fruit; n represents the number of standard fruits used for model parameter prediction, each row of the matrix A represents a group of initial data for model training, and a signal stable value corresponding to the ith row of the matrix A is the ith row of the matrix B; assuming that the order of the model is p, another set of variables is defined, which are respectively:
， ，
wherein,
training data for determining the parameters of the model,
as training data
The corresponding stable value of the value,
for estimating the model error in the determination of the model order,
is composed of
The data corresponds to a stable value of the data,is the model parameter for the order P,
for storing
The prediction data obtained by the calculation is used,
the standard deviation storage module is used for storing the standard deviation between the predicted value and the actual stable value when the order is i as a model error;
the improved AR model obtained from equation (2) is used for model parameter calculation as follows:
calculating linear coefficients
Then using the coefficient pairThe medium data is predicted, and
obtaining a set of predicted values
And then compared
And
the standard deviation when the p-order model is calculated is as follows:
calculating to obtain model errors corresponding to a group of model orders, and taking
The minimum corresponding order i is taken as the order of the model, and the model parameter when the order is i is calculated
(ii) a Finally, the order and the parameters of the model are used as the optimal coefficients under the current sampling speed and are transmitted to a control station system for predicting the sampling stable value of the fruit weight signal; when the system is initialized, the steps are repeated for training for multiple times, and model parameters and orders corresponding to different grading speeds are obtained and stored by files; when the grading speed needs to be changed according to the field condition, the operation station system directly transmits the model parameters and the orders under the corresponding grading speed in the file to the control station system, or recalculates the parameters and the orders of the model and transmits the parameters and the orders to the control station system; if the field conditions change, the recalculated value is inconsistent with the value stored in the file, and a selection can be made as to whether to update the file.
The step 5) is as follows: predicting a weighing signal stable value, firstly obtaining a grading speed, and determining that a model parameter and an order corresponding to the grading speed are stored in a control station system; when the fruit passes through the weighing sensor, the rising edge of an external synchronous signal triggers the interruption of a control station system, and the sampling times are determined by the order of the model; and after a group of data obtained by sampling is filtered, the control station system predicts by using the trained AR model to obtain a signal stability value, and the signal stability value is converted into weight information after calibration.
The step 6) is as follows: randomly taking K fruits, determining a sampling period, sampling to obtain K groups of high-speed dynamic data, predicting a stable value by using a trained model to obtain K predicted data, calibrating and converting the K predicted data into K predicted weights, and calculating to obtain the deviation, the relative error and the prediction precision of the predicted weights and the actual weights.
The invention adopts the improved AR model to predict the weight information of the fruits, so that the fruit weight prediction method can realize the high speed of 20 fruits per second
The grading precision can reach within (1.5 +/-0.5)% in a grading state, the grading speed is allowed to be changed according to the field requirement, and the optimal model order and parameters are obtained again, so that the requirement on weighing precision is met, the requirement on hardware is low, the structure is simple, the expansibility is good, and the portability is good.
Drawings
FIG. 1 is a schematic structural diagram of a high-speed fruit weighing and grading system adopting an improved AR model, wherein a fruit conveying device 1, a variable frequency motor 2, a PC (personal computer) 3, a PLC4, a fruit outlet 5, an electromagnet 6, a fruit cup 7, a signal preprocessing module 8, a bearing piece 9, a weighing sensor 10, a weighing area 11 and a synchronous signal generator 12 are arranged in the fruit weighing and grading system;
FIG. 2 is a general flow diagram of the signal processing module of the present invention;
FIG. 3 is a flow chart of the improved AR model training algorithm of the present invention;
FIG. 4 is a high speed weighing data processing flow diagram of the present invention;
FIG. 5 is a graph of the order and maximum obtained from model training data obtained at a classification rate of 20 fruits/second
According to the corresponding table of the relative errors, when the order is 9, the maximum relative error is 0.81 percent, namely the model parameter corresponding to the order is optimal;
FIG. 6 shows the maximum of 10 sets of data predicted by using the model parameters for the order of 9
The relative error is less than 1 percent, and the precision requirement is met.
Detailed Description
The intelligent processing method of the dynamic weighing signal of the fruit high-speed sorting system comprises the following steps:
1) when the fruit conveying device is static, sampling to obtain a group of stable values of the standard fruits, then starting to rotate the conveying device, determining a sampling period according to the grading speed, and sampling to obtain a group of data when each standard fruit passes through the weighing sensor;
2) eliminating low-frequency noise in the sampling signal by adopting first-order lag filtering to obtain a preprocessed signal;
3) the operating station system reads the preprocessed data, analyzes the characteristics of the data and obtains an improved AR model according to the real-time requirement;
4) determining parameters and orders of an improved AR model, and storing the model parameters and the orders obtained by training at different grading speeds in a file;
5) the control station system predicts the fruit weight information by using the trained model parameters;
6) and evaluating the precision of the prediction result.
The step 1) is as follows: n standard fruits with different weights are needed, under a static condition, the value obtained by sampling each standard fruit for multiple times is averaged to be used as the final static sampling value of the fruit and is respectively recorded as: b1，b2，…，bn(ii) a In the high speed case, according to the classification speed, the sampling period is determined: let the length of the load-bearing foot of the weighing sensor be L, the grading speed be V, and the time interval from the beginning of rising to the beginning of falling of the signal
The sampling can only be at
The sampling is completed within a time span, and if the sampling frequency of each fruit is m, the sampling period is
(ii) a The weighing sensor can be equivalent to an inertia link, the output signal of the fruit passing by is equivalent to the step response and the rise time of the inertia link
A time constant T dependent on the inertia element, which is satisfied at high speed
Therefore, the sampling value of the stable part cannot be directly obtained by sampling, and only the rising part data is used for prediction; when standard fruits pass through the weighing sensor, the rising edge of an external synchronous signal triggers the interruption of a control station system, after a period of time is delayed, sampling is carried out according to a set sampling period, and the sampling data y of each standard fruit is stored1(1)，y1(2)，…，y1(m) ；y2(1)，y2(2)，…，y2(m) ；…… ；yn（1)，yn(2)，…，yn(m)。
The step 2) is as follows: eliminating low-frequency noise in the sampled data by using first-order lag filtering, defining y (k) as a k-th sampled value, defining x (k-1) as a k-1-th filtered output value, substituting alpha as a filtering coefficient with the value less than 1 into a first-order lag filtering formula
Obtaining a filtering output x (k) corresponding to the kth sampling data; filtering each group of sampling data to obtain a group of data x1(1)，x1(2)，…，x1(m) ；x2(1)，x2(2)，…，x2(m) ；…… ；xn（1)，xn(2)，…，xn(m), as data for model training.
The step 3) is as follows: the output signal of the load cell is approximated as a stationary random time series, and an AR (autoregressive) model of the time series is described as: the current value x (n) of the sequence is expressed as a linear function of the past values plus an error term e (n), expressed by equation (1) as follows:
a in equation (1)1，a2，…，apIs the model coefficient; x (n-1), x (n-2), …, x (n-p) are p past values of the signal, and a p-order AR model is formed; due to the requirement of real-time performance, complex calculation cannot be performed after each sampling is finished, the model order and the parameters are respectively determined for the sampling signals of each fruit, and a stable value is obtained through multiple iterations, so that a long time is consumed; assuming that p initial sample values x (1), x (2), …, x (p) are known, and the signal reaches a stable value after N time points, the stable value x (p + N) of the signal is predicted, and the following equation (2) is obtained by recursion using equation (1):
……
（2）
coefficient A in equation (2)1，A2，…，ApThe signal stability value x (p + N) is linearly represented by initial sampling values x (1), x (2), …, x (p); the experiment proves that the method has the advantages that,
the prediction precision within 2 percent can be achieved as long as the amplitude of the signal reaches about 50 percent of the stable value; assuming that the length of the weighing foot is L, T1 is defined as the time taken by the signal amplitude from 0 to half the stable amplitude, and the speed V of the extension set is only required to be less than L/T1, compared with the weighing method of directly reading the stable value, the grading speed of the system can be doubled. And by analyzing the structural characteristics of the weighing sensor and the characteristics of the weighing output signal at low speed, the system is equivalent to an inertia link, the output signal of the fruit passing by is equivalent to the step response of the inertia link, and for the fruits with different weights, the inertia link passes by the same weighing sensor
The values of T are the same, and the linear relation between the stable value of the signal and the initial sampling value is consistent.
The step 4) is as follows: calculating and improving the order and the parameters of the AR model by using the operating station system, firstly, reading model training data x by using the operating station system1(1)，x1(2)，…，x1(m) ；x2(1)，x2(2)，…，x2(m) ；…… ；xn（1)，xn(2)，…，xn(m), and a static stability value b1，b2，…，bnStored in matrix variables, noted as:
wherein m represents the number of samples per fruit; n represents the number of standard fruits used for model parameter prediction, each row of the matrix A represents a group of initial data for model training, and a signal stable value corresponding to the ith row of the matrix A is the ith row of the matrix B; assuming that the order of the model is p, another set of variables is defined, which are respectively:
wherein,
training data for determining the parameters of the model,
as training data
The corresponding stable value of the value,
for estimating the model error in the determination of the model order,
is composed ofThe data corresponds to a stable value of the data,
is the model parameter for the order P,
for storing
The prediction data obtained by the calculation is used,
the standard deviation storage module is used for storing the standard deviation between the predicted value and the actual stable value when the order is i as a model error;
the improved AR model obtained from equation (2) is used for model parameter calculation as follows:
calculating linear coefficients
Then using the coefficient pairThe medium data is predicted, and
obtaining a set of predicted values
And then compared
And
the standard deviation when the p-order model is calculated is as follows:
calculating to obtain model errors corresponding to a group of model orders, and taking
The minimum corresponding order i is taken as the order of the model, and the model parameter when the order is i is calculated
(ii) a Finally, the order and the parameters of the model are used as the optimal coefficients under the current sampling speed and are transmitted to a control station system for predicting the sampling stable value of the fruit weight signal; when the system is initialized, the steps are repeated for training for multiple times to obtain the corresponding model parameters and orders at different grading speedsStoring the data by using a file; when the grading speed needs to be changed according to the field condition, the operation station system directly transmits the model parameters and the orders under the corresponding grading speed in the file to the control station system, or recalculates the parameters and the orders of the model and transmits the parameters and the orders to the control station system; if the field conditions change, the recalculated value is inconsistent with the value stored in the file, and a selection can be made as to whether to update the file.
The step 5) is as follows: predicting a weighing signal stable value, firstly obtaining a grading speed, and determining that a model parameter and an order corresponding to the grading speed are stored in a control station system; when the fruit passes through the weighing sensor, the rising edge of an external synchronous signal triggers the interruption of a control station system, sampling is started after a period of time delay, and the sampling frequency is determined by the order number of the model; and after a group of data obtained by sampling is filtered, the control station system predicts by using the trained AR model to obtain a signal stability value, and the signal stability value is converted into weight information after calibration.
The step 6) is as follows: randomly taking K fruits, determining a sampling period, sampling to obtain K groups of high-speed dynamic data, predicting a stable value by using a trained model to obtain K predicted data, calibrating and converting the K predicted data into K predicted weights, and calculating to obtain the deviation, the relative error and the prediction precision of the predicted weights and the actual weights.
Example (b):
as shown in fig. 1, the fruit weighing and grading system comprises a fruit conveying device 1, a variable frequency motor 2, a PC machine 3, a PLC4, a fruit outlet 5, an electromagnet 6, a fruit cup 7, a signal preprocessing module 8, a bearing sheet 9, a weighing sensor 10, a weighing area 11 and a synchronous signal generator 12; fruit conveyor 1 links to each other with inverter motor 2, synchronous signal generator 13 installs on fruit conveyor 1's the follow driving wheel, be equipped with fruit cup 7 on fruit conveyor 1's the conveyer belt, fruit has been put in the fruit cup 7, be weighing area 11 in the one end below of fruit conveyor 1's conveyer belt, be equipped with bearing piece 9 and weighing sensor 10 in weighing area 11, be equipped with electro-magnet 6 in the other end below of fruit conveyor 1's conveyer belt, weighing sensor 10 and signal preprocessing module 8, PLC4, electro-magnet 6 connects gradually, the below of electro-magnet 6 is equipped with down fruit mouth 5, PLC4 respectively with inverter motor 2, PC 3 links to each other. The signal preprocessing module 8 is responsible for preprocessing the output signal of the real-time symmetrical retransmission sensor 11, such as amplifying and filtering. When the fruit cup 7 passes through the weighing area 11, the PLC processing module 8 calculates the accurate weight information borne by the fruit cup in real time, judges the grade of the fruit, tracks the fruit in real time, controls the action of unloading the fruit by the electromagnet 6 and controls the grading speed of the variable frequency motor 2. The PC 2 obtains an accurate algorithm model according to different grading speeds, and monitors the whole fruit grading process in real time. The control station system of the invention adopts S7-300 series PLC of Siemens company, and the operation station system adopts VB monitoring software to realize the intelligent processing method of high-speed weighing.
Assume at this point that the speed of the fruit conveyor is 20 fruits/second, i.e. the time interval for each fruit to pass the load cell is 50 milliseconds. When the fruit passes through the weighing sensor, because the stay time is very short, the output waveform of the transmitter does not reach a stable value, the fruit leaves the sensor, and the output signal begins to decline, so that the real sampling value of the fruit cannot be directly obtained from the sampling value of the transmitter.
As shown in fig. 2, the total flow of the system is: after initialization is completed, if model training is carried out, data of standard fruits are obtained; and if the model training is finished, the sampling data is the sampling data of the fruit to be classified, and then the corresponding model is used for prediction. The data preparation process of model training is as follows: taking 15 standard fruits of different weights, the weight of the standard fruits is required to cover a large range, for example, the weight of the apples is used for weighing, and the weight can be varied from 50g to 500 g. Under the static condition, sequentially placing 15 standard fruits on a fruit cup, sampling each standard fruit for multiple times, then taking an average value as a static stable value of the weight of the fruit, and respectively recording the value as: b1, b2, …, b 15. Then the fruit conveying device starts to rotate at a high speed, when the rising edge of the synchronous signal comes, interruption is generated, the grading speed is obtained according to the time interval of the rising edge, and the sampling period is calculated: if the length of the load-bearing leg of the load cell is L and the classification speed is V, the signal rises from the beginning to the beginningTime interval of descent
The sampling can only be at
The sampling is completed within time, and if the sampling times of each fruit are m, the sampling speed is high
And according to the limit condition of the sampling interval and the limit of the minimum time of the PLC interruption interval, taking a proper sampling time. When the rising edge of the external synchronizing signal arrives, delaying a sampling period again, the PLC begins to sample each standard effect for N times according to the set sampling period, and the sampling data is stored in a DB block of the PLC in a Dint type, and occupies a storage space of 60 bytes.
As shown in fig. 3, the training of the AR model order and parameters is improved. Firstly, using first-order lag filtering to eliminate low-frequency noise in sampling data, defining y (k) as k sampling value, defining x (k-1) as k-1 filtering output value, making alpha be filter coefficient, its value be less than 1, substituting into first-order lag filtering formula
Obtaining a filtering output x (k) corresponding to the kth sampling data; filtering each group of sampling data to obtain a group of data x1(1)，x1(2)，…，x1(m) ；x2(1)，x2(2)，…，x2(m) ；…… ；xn（1)，xn(2)，…，xn(m), as data for model training. Experiments show that the value of a is between 0.4 and 0.6, the filtering effect is good, and the filtered data is continuously stored in a DB block of the PLC in a Dint type. After filtering is finished, a button for reading initial data on an operation station system interface becomes visible, and after the button is clicked, filtering output data and static stable values of 15 fruits are stored in a matrix variable and recorded as:
wherein m =10, representing a number of subsamples per fruit of 10; n =15, representing the number of standard fruits used for model parameter prediction; each row of the matrix A represents a group of sampling values, and the stable value corresponding to the ith row corresponds to the ith row of the matrix B. The order of the first hypothesis model is
Defining another set of matrix variables, respectively
， 
，
Wherein,
training data for determining the parameters of the model,
for a stable value corresponding to the training data,
by using
The model error at the time of determination of the model order,
is composed of
The data corresponds to a stable value of the data,
is the model parameter for the order P,
for storing
The prediction data obtained by the calculation is used,
for storing the model error at order i. Formula for obtaining improved AR model from step 3)
Calculating linear coefficientsThen using the coefficient pair
The medium data is predicted, and
=
*
obtaining a set of predicted values
And then compared
And
calculating the standard deviation of the p-order model, finally obtaining the standard deviation corresponding to a group of model orders, and taking the standard deviation
The minimum order i is taken as the order of the final model, and the model parameter when the order is i is calculated
And obtaining the order and the parameters of the model. And changing the grading speed to obtain the grading speed from 5 fruits/second to 20 fruits/second, and storing the corresponding model order and parameters in a file. When the grading speed is changed, the grading speed can be selected through the operation station system, if the 'retraining model' is clicked, the steps are repeated, new model parameters are obtained and transmitted to the PLC, and if the 'writing order and parameters' are clicked, the model order and the parameter values corresponding to the grading speed in the file are directly transmitted to the PLC. At a classification rate of 20 fruits per second, the relationship between the model standard deviation and the order is shown in fig. 5, and it is known that the model parameter is optimal when the order of the model is 9.
As shown in FIG. 4, the weighing signal prediction in the high speed condition is performed by first obtaining the classification speedThe model parameters and the order corresponding to the classification speed are already stored in the control station system. When the fruit passes through the weighing sensor, triggering the interruption of the control station system by the rising edge of the synchronous signal, sampling at the sampling speed determined in the step 1), wherein the sampling times are determined by the order number of the model; and after a group of data obtained by sampling is filtered, the control station system predicts by using the trained AR model to obtain a stable value of the signal. Then, calibration is carried out, and the linear relation between the magnitude of the signal stable value and the actual weight of the fruit is
Wherein
is the coefficient of the first-order term,
is a coefficient of a constant term and is,
and
and determining through a calibration experiment, and calculating to obtain the actual weight of the fruit.
The method for evaluating the result precision comprises the following steps: randomly selecting 10 fruits, sampling to obtain 10 groups of high-speed dynamic data, and predicting a stable value by using a trained model to obtain 10 predicted data. And VB reads the prediction data, calibrates and converts the prediction data into the weight of the fruit, outputs the weight of the fruit into a text box, inputs the corresponding real weight of the fruit from a keyboard, and calculates to obtain the deviation, relative error and prediction precision of the predicted value and the actual value. During the experiment, 10 randomly sampled sets of sample data were predicted using the obtained model at a classification speed of 20 fruit/sec, and the obtained relative error is shown in fig. 6.
Claims (7)
1. An intelligent processing method for dynamic weighing signals of a high-speed fruit sorting system is characterized by comprising the following steps:
1) when the fruit conveying device is static, sampling to obtain a group of stable values of the standard fruits, then starting to rotate the conveying device, determining a sampling period according to the grading speed, and sampling to obtain a group of data when each standard fruit passes through the weighing sensor;
2) eliminating low-frequency noise in the sampling signal by adopting first-order lag filtering to obtain a preprocessed signal;
3) the operating station system reads the preprocessed data, analyzes the characteristics of the data and obtains an improved AR model according to the real-time requirement;
4) determining parameters and orders of an improved AR model, and storing the model parameters and the orders obtained by training at different grading speeds in a file;
5) the control station system predicts the fruit weight information by using the trained model parameters;
6) and evaluating the precision of the prediction result.
2. The intelligent processing method for the dynamic weighing signal of the fruit high-speed sorting system according to claim 1
The method is characterized in that the step 1) is as follows: n standard fruits with different weights are needed, under a static condition, the value obtained by sampling each standard fruit for multiple times is averaged to be used as the final static sampling value of the fruit and is respectively recorded as: b1，b2，…，bn(ii) a In the high speed case, according to the classification speed, the sampling period is determined: let the length of the load-bearing foot of the weighing sensor be L, the grading speed be V, and the time interval from the beginning of rising to the beginning of falling of the signal
The sampling can only be at
The sampling is completed within a time span, and if the sampling frequency of each fruit is m, the sampling period is
(ii) a Finally, sampling according to a set sampling period, and storing the sampling data y of each standard fruit1(1)，y1(2)，…，y1(m) ；y2(1)，y2(2)，…，y2(m) ；…… ；yn（1)，yn(2)，…，yn(m)。
3. The intelligent processing method of dynamic weighing signals of a high-speed fruit sorting system according to claim 1, wherein the step 2) is: eliminating low-frequency noise in the sampled data by using first-order lag filtering, defining y (k) as a k-th sampled value, defining x (k-1) as a k-1-th filtered output value, substituting alpha as a filtering coefficient with the value less than 1 into a first-order lag filtering formula
Obtaining a filtering output x (k) corresponding to the kth sampling data; filtering each group of sampling data to obtain a group of data x1(1)，x1(2)，…，x1(m) ；x2(1)，x2(2)，…，x2(m) ；…… ；xn（1)，xn(2)，…，xn(m), as data for model training.
4. The intelligent processing method for the dynamic weighing signal of the high-speed fruit sorting system according to claim 1, wherein the step 3) is as follows: the output signal of the load cell is approximated as a stationary random time series, and an AR model of the time series is described as: the current value x (n) of the sequence is expressed as a linear function of the past values plus an error term e (n), expressed by equation (1) as follows:
a in equation (1)1，a2，…，apIs the model coefficient; x (n-1), x (n-2), …, x (n-p) are p past values of the signal, and a p-order AR model is formed; assuming that p initial sample values x (1), x (2), …, x (p) are known, and the signal reaches a stable value after N time points, the stable value x (p + N) of the signal is predicted, and the following equation (2) is obtained by recursion using equation (1):
coefficient A in equation (2)1，A2，…，ApThe signal stability value x (p + N) is linearly represented by initial sampling values x (1), x (2), …, x (p); for fruits with different weights, the linear relation between the signal stable value and the initial sampling value is consistent after the fruits pass through the same weighing sensor.
5. The intelligent processing method for the dynamic weighing signal of the high-speed fruit sorting system according to claim 1, wherein the step 4) is as follows: computing improved AR model orders and parameters using an operator station system
First, the operating station system reads the model training data x1(1)，x1(2)，…，x1(m) ；x2(1)，x2(2)，…，
x2(m) ；…… ；xn（1)，xn(2)，…，xn(m), and a static stability value b1，b2，…，bn，
Stored in a matrix variable, noted:
wherein m represents the number of samples per fruit; n represents the number of standard fruits used for model parameter prediction, each row of the matrix A represents a group of initial data for model training, and a signal stable value corresponding to the ith row of the matrix A is the ith row of the matrix B; assuming that the order of the model is p, another set of variables is defined, which are respectively:
wherein,
training data for determining the parameters of the model,
as training data
The corresponding stable value of the value,
for estimating the model error in the determination of the model order,
is composed of
The data corresponds to a stable value of the data,
is the model parameter for the order P,
for storing
The prediction data obtained by the calculation is used,
the standard deviation storage module is used for storing the standard deviation between the predicted value and the actual stable value when the order is i as a model error;
the improved AR model obtained from equation (2) is used for model parameter calculation as follows:
(3)
calculating linear coefficients
Then using the coefficient pair
The data in the data table are predicted,order to
(4)
Obtaining a set of predicted values
And then compared
And
the standard deviation when the p-order model is calculated is as follows:
calculating to obtain model errors corresponding to a group of model orders, and taking
The minimum corresponding order i is taken as the order of the model, and the model parameter when the order is i is calculated
(ii) a Finally, the order and the parameters of the model are used as the optimal coefficients under the current sampling speed and are transmitted to a control station system for predicting the sampling stable value of the fruit weight signal; when the system is initialized, the steps are repeated for training for multiple times, and model parameters and orders corresponding to different grading speeds are obtained and stored by files; when the grading speed is required to be changed according to the field conditions, the operation station system directly transmits the model parameters and the grades under the corresponding grading speed in the file to the control station system,or recalculating the parameters and the orders of the model and transmitting the recalculated parameters and the orders to the control station system; if the field conditions change, the recalculated value is inconsistent with the value stored in the file, and a selection can be made as to whether to update the file.
6. The intelligent processing method for the dynamic weighing signal of the high-speed fruit sorting system according to claim 1, wherein the step 5) is as follows: predicting a weighing signal stable value, firstly obtaining a grading speed, and determining that a model parameter and an order corresponding to the grading speed are stored in a control station system; when the fruit passes through the weighing sensor, the rising edge of an external synchronous signal triggers the interruption of a control station system, and the sampling times are determined by the order of the model; and after a group of data obtained by sampling is filtered, the control station system predicts by using the trained AR model to obtain a signal stability value, and the signal stability value is converted into weight information after calibration.
7. The intelligent processing method for the dynamic weighing signal of the high-speed fruit sorting system according to claim 1, wherein the step 6) is as follows: randomly taking K fruits, determining a sampling period, sampling to obtain K groups of high-speed dynamic data, predicting a stable value by using a trained model to obtain K predicted data, calibrating and converting the K predicted data into K predicted weights, and calculating to obtain the deviation, the relative error and the prediction precision of the predicted weights and the actual weights.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110022875 CN102176117B (en) | 2011-01-20 | 2011-01-20 | Intelligent processing method for dynamic weighing signal of fruit high-speed sorting system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110022875 CN102176117B (en) | 2011-01-20 | 2011-01-20 | Intelligent processing method for dynamic weighing signal of fruit high-speed sorting system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102176117A true CN102176117A (en) | 2011-09-07 |
| CN102176117B CN102176117B (en) | 2012-12-05 |
Family
ID=44519308
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201110022875 Expired - Fee Related CN102176117B (en) | 2011-01-20 | 2011-01-20 | Intelligent processing method for dynamic weighing signal of fruit high-speed sorting system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102176117B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103840970A (en) * | 2014-01-24 | 2014-06-04 | 珠海多玩信息技术有限公司 | Method and device for obtaining running status of service |
| CN104386419A (en) * | 2014-11-17 | 2015-03-04 | 浙江大学 | Ellipsoidal fruit conveying and rotating device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101085442A (en) * | 2006-06-06 | 2007-12-12 | 中国农业大学 | Method for treating and classifying orange image based on RGB composite model |
| US20100139989A1 (en) * | 2008-12-09 | 2010-06-10 | Datalogic Scanning, Inc. | Systems and methods for reducing weighing errors associated with partially off-scale items |
| CN101898191A (en) * | 2010-07-16 | 2010-12-01 | 浙江大学 | Weighting method and fruit weighting grading system adopting DSP (Digital Signal Processor) processing module and intelligent algorithm |
-
2011
- 2011-01-20 CN CN 201110022875 patent/CN102176117B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101085442A (en) * | 2006-06-06 | 2007-12-12 | 中国农业大学 | Method for treating and classifying orange image based on RGB composite model |
| US20100139989A1 (en) * | 2008-12-09 | 2010-06-10 | Datalogic Scanning, Inc. | Systems and methods for reducing weighing errors associated with partially off-scale items |
| CN101898191A (en) * | 2010-07-16 | 2010-12-01 | 浙江大学 | Weighting method and fruit weighting grading system adopting DSP (Digital Signal Processor) processing module and intelligent algorithm |
Non-Patent Citations (1)
| Title |
|---|
| 《农机化研究》 20101031 毛华先 水果品质智能化实时检测和分级系统研究 95-97 1-7 , 第10期 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103840970A (en) * | 2014-01-24 | 2014-06-04 | 珠海多玩信息技术有限公司 | Method and device for obtaining running status of service |
| CN103840970B (en) * | 2014-01-24 | 2017-09-15 | 珠海多玩信息技术有限公司 | A kind of method and device for obtaining service operation state |
| CN104386419A (en) * | 2014-11-17 | 2015-03-04 | 浙江大学 | Ellipsoidal fruit conveying and rotating device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102176117B (en) | 2012-12-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110678816B (en) | Method and control device for controlling a technical system | |
| CN108875710A (en) | Elevator door speed of service estimation method based on energy threshold algorithm | |
| CN103494316B (en) | A kind of method dividing redried leaf tobacco batch by weight | |
| Delwiche | Theory of fruit firmness sorting by impact forces | |
| CN101898191B (en) | Weighting method and fruit weighting grading system adopting DSP (Digital Signal Processor) processing module and intelligent algorithm | |
| CN102176117B (en) | Intelligent processing method for dynamic weighing signal of fruit high-speed sorting system | |
| CN111389753A (en) | P L C fruit vegetables intelligence sorting system that can long-range wisdom control | |
| CN106603037A (en) | Smooth data processing method and system | |
| CN114199364A (en) | Vibration monitoring system of aircraft engine | |
| CN106768243B (en) | A kind of quick lock in accurate weight method | |
| JPH06222885A (en) | Method and apparatus for digital processing and filtration of signal in industrial control application | |
| CN113225047A (en) | TVLP-MF-based dynamic checkweigher rapid filtering method and system | |
| CN109271889B (en) | Action recognition method based on double-layer LSTM neural network | |
| CN102435676B (en) | Method for inspecting acceptability of amorphous alloy products by means of sound | |
| CN106706957A (en) | Acceleration estimation method and apparatus thereof, and locomotive motion control method and locomotive | |
| Kelemençe et al. | Dynamic weighing using a time-variant low pass filter | |
| Dethe et al. | On the prediction of packet process in network traffic using FARIMA time-series model | |
| CN111693125A (en) | Method and system for calculating length of weighing platform of high-precision dynamic weighing equipment | |
| CN104942810A (en) | Intelligent robot capable of denoising | |
| Meller et al. | Adaptive filtering approach to dynamic weighing: A checkweigher case study | |
| Umemoto et al. | Improvement of accuracy for continuous mass measurement in checkweighers with an adaptive notch filter | |
| CN118673448B (en) | Electric chain saw chain equipment and work safety detection method | |
| Kim et al. | Adaptable noise reduction of ECG signals for feature extraction | |
| Iribas-Latour et al. | Closed loop identification of wind turbines models for pitch control | |
| CN108627357A (en) | A kind of coalcutter cutting load flexible measurement method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20121205 Termination date: 20130120 |