CN103234610B - Weighing method applicable to truck scale - Google Patents

Weighing method applicable to truck scale Download PDF

Info

Publication number
CN103234610B
CN103234610B CN201310177182.5A CN201310177182A CN103234610B CN 103234610 B CN103234610 B CN 103234610B CN 201310177182 A CN201310177182 A CN 201310177182A CN 103234610 B CN103234610 B CN 103234610B
Authority
CN
China
Prior art keywords
weighing
layer
training
formula
weighting model
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.)
Expired - Fee Related
Application number
CN201310177182.5A
Other languages
Chinese (zh)
Other versions
CN103234610A (en
Inventor
林海军
滕召胜
汪鲁才
杨进宝
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.)
Hunan Normal University
Original Assignee
Hunan Normal 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 Hunan Normal University filed Critical Hunan Normal University
Priority to CN201310177182.5A priority Critical patent/CN103234610B/en
Publication of CN103234610A publication Critical patent/CN103234610A/en
Application granted granted Critical
Publication of CN103234610B publication Critical patent/CN103234610B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Control Of Transmission Device (AREA)
  • Traffic Control Systems (AREA)

Abstract

The invention discloses a weighing method applicable to a truck scale, and the method comprises a weighing sensor, a data acquisition device, a microprocessor and a display which are used, and the following steps that a weighing mathematical model is established, a weighing signal is acquired and on-line weighing is carried out; the step of establishing the weighing mathematical model comprises an ideal weighing model, an actual weighing model and the training methods thereof; the ideal weighing model is a linear function; the actual weighing model is a three-layer BP neural network, wherein a first layer is an input layer, a second layer is a hidden layer and a third layer is an output layer; before on-line weighing, the training of setting sample sizes must be carried out on the ideal weighing model and the actual weighing model; the training is carried out when the microprocessor is connected with an external computer; sample information of set quantity is acquired; the ideal weighing model and a derivative thereof serve as constraint conditions; finally the parameters W, b (1), V and b (2) of the actual weighing model are acquired and saved in the microprocessor; and then the external computer is removed.

Description

A kind of Weighing method being applicable to truck scale
Technical field
The present invention relates to a kind of Weighing method being applicable to truck scale.
Technical background
Truck scale is the important branch of weighing apparatus, is mainly used in bulk supply tariff metering.Current analog electronic vehicle weighing apparatus occupies the leading position in truck scale market, and it is primarily of load-bearing force transmission mechanism (scale body), simulation LOAD CELLS, the large master unit composition of weighting display instrument three.Truck scale is different according to range, generally has 4 ~ 12 tunnel LOAD CELLS.These sensors, according to certain topological structure, are distributed in below scale body symmetrically, constitute a multisensor syste.There is coupling in this multisensor syste, each road sensor exports interrelated, and relevant with load loading position.Truck scale is concentrated cumulative in the output signal of simulation connection He Zhongjiangge road LOAD CELLS, obtain a voltage signal proportional with tested quality of loads, after signal condition, A/D conversion, obtain weighing results by single-chip microcomputer process, send display, communication, complete weighing of tested load.Uneven loading error and linearity error are two principal elements affecting truck scale weighing results accuracy.Uneven loading error is because truck scale is by the impact of various non-linear factor, and when tested load is in diverse location on truck scale loading end, weighing results is inconsistent and the error that produces; Linearity error is because the characteristic of each road LOAD CELLS is inconsistent, causes the constrained input of truck scale and nonideal linear relationship, thus the weighting error produced.The uneven loading error of existing truck scale is what to separate with linearity error compensation process, traditional uneven loading error compensation method is by resistor in the artificial weighing apparatus of vehicles repeatedly terminal box, change the sensor passage gain of every road, compensate uneven loading error, this method manual operation is loaded down with trivial details, inefficiency, compensation effect is poor, for this reason, scholar is had to propose multiple linear regression analysis method (" research of large-scale weighing machine system partial load digitized compensation method ", Chen Chang, Wang Xiaoliang, Qin Zijun, Journal of Dalian University of Technology Total, 1994, 1), utilize the method for Solving Linear angular difference correction factor (" studying based on the intelligent weighing sensor of advance data treatment technology ", Zhu Zijian, Nanjing Aero-Space University's PhD dissertation, 2005), but these methods can not solve each sensor output relevance problem because topological structure brings, the various non-linear factor impact of truck scale is not considered yet, therefore compensation effect is poor, have scholar to adopt neural net method to carry out truck scale uneven loading error and linearity error compensate (" the truck scale error compensation based on multi-sensor information fusion ", vast stretch of wooded country army, Teng Zhaosheng, Chi Hai, etc., Chinese journal of scientific instrument, 2009,6, " the truck scale error compensation based on multiple RBF neural network ", vast stretch of wooded country army, Teng Zhaosheng, Chi Hai, etc., Hunan University's journal, 2010,5), although can greatly reduce weighting error, neural network needs a large amount of training samples, and workload is large, the main cause that workload is large is that truck scale range is large, and the standard test weight needed during test is many, load(ing) point is many, and weighing information obtains not easily.It is after uneven loading error has compensated that existing truck scale linearity error compensates, profit completes with the following method: first utilize the standard test weight of Different Weight to be carried in successively on truck scale body, obtain target weighing results and actual weighing results, then actual weighing results is doubly taken advantage of correction factor, make it to equal target weighing results.This method is based on truck scale input-output on the basis being linear relationship, and differ comparatively large with reality, therefore compensation effect is poor.
Summary of the invention
The object of the invention is to overcome deficiency of the prior art, provide a kind of Weighing method being applicable to truck scale newly: namely utilize the Nonlinear Function Approximation ability that neural network is good, the actual weighting model of structure truck scale; Utilize the constraint condition of the desirable weighting model constructing neural network of truck scale simultaneously, to reduce the sample needed for neural metwork training, reduce workload, complete the optimization of the actual weighting model of truck scale, finally realize truck scale correct amount and error compensation.
Object of the present invention is achieved by following technical proposals:
Described Weighing method comprises use LOAD CELLS, data collector, microprocessor and display; Described LOAD CELLS is connected with microprocessor by data collector; Described display is connected with microprocessor; The step of described Weighing method comprises foundation and to weigh mathematical model, weighing-up wave collection, online weighing; The weigh mathematical model of mathematical model of described foundation comprises desirable weighting model, actual weighting model and their training method, its step:
1) described desirable weighting model is linear function; Being input as of described linear function nthe data that road LOAD CELLS exports , export as A (X); Its Input output Relationship expression formula is formula (1):
(1);
In formula, p i for the gain coefficient of LOAD CELLS, its value obtains by carrying out training to desirable weighting model;
2) described actual weighting model is three layers of BP neural network, and ground floor is input layer, and the second layer is hidden layer, and third layer is output layer, and their network structure is as follows respectively:
The neuronal quantity of input layer nfor the number of LOAD CELLS;
The neuronal quantity of hidden layer m= , in formula: k=1 ~ 10 is correction factor; lfor the neuronal quantity of output layer; Hidden layer excitation function adopts Log-Sigmoid function, namely exports for formula (2):
(2);
The neuronal quantity of output layer lit is 1; Output layer excitation function adopts linear function; The neural network of output layer exports for formula (3):
(3);
In formula, wfor neural network input layer is to the weight matrix of hidden layer, b (1)for hidden layer bias vector, vfor hidden layer is to the weight vector of output layer, b (2)for output layer bias, xfor neural network input vector, w mi for input layer iroad is input to of hidden layer mindividual neuronic connection weights, for hidden layer mindividual neuronic bias, v m for hidden layer mindividual neuron to the connection weights of output layer, x( i) be input layer iroad inputs;
3) before truck scale drops into online weighing, the training of set point number must be carried out to desirable weighting model and actual weighting model, training process carries out when microprocessor is connected with outer computer, with desirable weighting model and derivative thereof for constraint condition, finally obtain actual weighting model parameter w, b (1), vand b (2)preserve in the microprocessor, then withdraw outer computer; Special training software is installed in outer computer;
The step of training actual weighting model is as follows:
I) gather training sample: the standard test weight preparing some, each standard test weight quality is different, by the standard test weight random loading of different quality on truck scale body, nroad sensor just has nindividual output data, nindividual output data and corresponding standard test weight quality form one group of training sample and are kept in outer computer;
II) construct training objective function, its relational expression is formula (4):
(4);
III) ask the derivative of the desirable weighting model of truck scale, its relational expression is formula (5):
(5);
IV) ask the weight coefficient of desirable weighting model in training objective function, its relational expression is formula (6):
(6);
V) ask third layer output layer derivative, its relational expression is formula (7):
(7);
VI) ask respectively w, b (1) , Vand b (2)increment , , , , and right w, b (1) , Vand b (2)upgrade, their relational expression is respectively formula (8), (9):
(8)
(9)
In formula (9), , , , be respectively , , , value after renewal, , , , be respectively , , , value before renewal;
Vii) arrange training starting condition, carry out according to formula (6) ~ (9) training setting quantity, the error amount that training is produced, in setting range, obtains the weight matrix of input layer to hidden layer respectively w, hidden layer bias vector b (1), hidden layer is to the weight vector of output layer v, output layer bias b (2)end value, and be kept in the storage element of microprocessor, for called during online weighing.
In the training process to desirable weighting model, first utilize the training sample gathered in actual weighting model training process, then utilize least square method to train, obtain final model coefficient p i , its relational expression is formula (10):
(10)
In formula (10), pserve as reasons p i vector, namely p=[ p 1, p 2..., p n ] t, nfor the number of LOAD CELLS; x=[ x 1, x 2..., x j..., x k ] be input amendment matrix, y=[ y 1, y 2, y j ..., y k ].
Described online weighing should carry out after actual weighting model training is qualified, and its step is as follows:
1) by gather nthe weighing-up wave of road LOAD CELLS is as the input vector of BP neural network ground floor x;
2) by input vector xwith the actual weighting model parameter of preserving in the memory unit w, b (1), vand b (2)substitute into together in formula (3), try to achieve BP neural network and export for final weighing results;
3) display shows final weighing results.
The method that described weighing-up wave gathers, the data obtained after the output signal of each road LOAD CELLS is carried out signal amplification, filtering and analog-to-digital conversion process, train the input vector with online weighing as desirable weighting model and actual weighting model x.
Described microprocessor is single-chip microcomputer, dsp processor or other embedded system devices, and with storage unit.
Compared with the prior art the present invention has following advantage: the present invention can realize the automatic Weighing of truck scale, and carries out the auto-compensation of uneven loading error and linearity error simultaneously, substantially increases the accuracy of weighing results; Decrease the sample size needed for the training of truck scale weighting model simultaneously, improve work efficiency.
Accompanying drawing explanation
Fig. 1 is the FB(flow block) that the present invention trains embodiment.
Fig. 2 is the FB(flow block) of online weighing embodiment of the present invention.
Fig. 3 is neural network embodiment of the present invention, wherein, f1 is hidden layer excitation function, f2 is output layer excitation function.
Fig. 4 is the signal acquisition circuit theory diagram of one embodiment of the invention.
Fig. 5 is embodiments of the invention 5 truck scale online weighing and error compensation simulation result figure, and wherein (a) compensates forward and backward weighing results comparison diagram, and (b) compensates forward and backward weighing results graph of errors comparison diagram.
In the diagram: 1-modulate circuit, 2-analog to digital conversion circuit, 3-microprocessor, 4-power module, 5-outer computer, 6-keyboard, 7-display, 8-communication interface.
Embodiment
Below in conjunction with drawings and Examples, the invention will be further described:
See Fig. 1-5, described Weighing method comprises use LOAD CELLS, data collector, microprocessor 3 and display 7; Described LOAD CELLS is connected with microprocessor by data collector; Described display is connected with microprocessor; The step of described Weighing method comprises foundation and to weigh mathematical model, weighing-up wave collection, online weighing; The weigh mathematical model of mathematical model of described foundation comprises desirable weighting model, actual weighting model and their training method, its step;
1) described desirable weighting model is linear function; Being input as of described linear function nthe data that road LOAD CELLS exports , export as A (X); Its Input output Relationship expression formula is formula (1):
(1);
In formula, p i for the gain coefficient of LOAD CELLS, its value obtains by carrying out training to desirable weighting model;
2) described actual weighting model is three layers of BP neural network, and ground floor is input layer, and the second layer is hidden layer, and third layer is output layer, and their network structure is as follows respectively:
The neuronal quantity of input layer nfor the number of LOAD CELLS;
The neuronal quantity of hidden layer m= , in formula: k=1 ~ 10 is correction factor; lfor the neuronal quantity of output layer; Hidden layer excitation function adopts Log-Sigmoid function, namely exports for formula (2):
(2);
The neuronal quantity of output layer lit is 1; Output layer excitation function adopts linear function; The neural network of output layer exports for formula (3):
(3);
In formula, wfor neural network input layer is to the weight matrix of hidden layer, b (1)for hidden layer bias vector, vfor hidden layer is to the weight vector of output layer, b (2)for output layer bias, xfor neural network input vector, w mi for input layer iroad is input to of hidden layer mindividual neuronic connection weights, for hidden layer mindividual neuronic bias, v m for hidden layer mindividual neuron to the connection weights of output layer, x( i) be input layer iroad inputs;
3) before truck scale drops into online weighing, the training of set point number must be carried out to desirable weighting model and actual weighting model, training process carries out when microprocessor is connected with outer computer, with desirable weighting model and derivative thereof for constraint condition, finally obtain actual weighting model parameter w, b (1), vand b (2)preserve in the microprocessor, then withdraw outer computer; Special training software is installed in outer computer;
The step of training actual weighting model is as follows:
I) gather training sample: the standard test weight preparing some, each standard test weight quality is different, by the standard test weight random loading of different quality on truck scale body, nroad sensor just has nindividual output data, nindividual output data and corresponding standard test weight quality form one group of training sample and are kept in outer computer;
II) construct training objective function, its relational expression is formula (4):
(4);
III) ask the derivative of the desirable weighting model of truck scale, its relational expression is formula (5):
(5);
IV) ask the weight coefficient of desirable weighting model in training objective function, its relational expression is formula (6):
(6);
V) ask third layer output layer derivative, its relational expression is formula (7):
(7);
VI) ask respectively w, b (1) , Vand b (2)increment , , , , and right w, b (1) , Vand b (2)upgrade, their relational expression is respectively formula (8), (9):
(8)
(9)
In formula (9), , , , be respectively , , , value after renewal, , , , be respectively , , , value before renewal;
Vii) arrange training starting condition, carry out the training of set point number according to formula (6) ~ (9), the error amount that training is produced, in setting range, obtains the weight matrix of input layer to hidden layer respectively w, hidden layer bias vector b (1), hidden layer is to the weight vector of output layer v, output layer bias b (2)end value, and be kept in the storage element of microprocessor, for called during online weighing.
In the training process to desirable weighting model, first utilize the training sample gathered in actual weighting model training process, then utilize least square method to train, obtain final model coefficient p i , its relational expression is formula (10):
(10)
In formula (10), pserve as reasons p i vector, namely p=[ p 1, p 2..., p n ] t, nfor the number of LOAD CELLS; x=[ x 1, x 2..., x j..., x k ] be input amendment matrix, y=[ y 1, y 2, y j ..., y k ].
Described online weighing should carry out after actual weighting model training is qualified, and its step is as follows:
1) by gather nthe weighing-up wave of road LOAD CELLS is as the input vector of BP neural network ground floor x;
2) by input vector xwith the actual weighting model parameter of preserving in the memory unit w, b (1), vand b (2)substitute into together in formula (3), try to achieve BP neural network and export for final weighing results;
3) display 7 shows final weighing results.
The method that described weighing-up wave gathers, the data obtained after the output signal of each road LOAD CELLS is carried out signal amplification, filtering and analog-to-digital conversion process, train the input vector with online weighing as desirable weighting model and actual weighting model x.
Described microprocessor is single-chip microcomputer, dsp processor or other embedded system devices, and with storage unit.
Described data collector comprises modulate circuit 1 and analog to digital conversion circuit 2, adopts known technology.
Embodiment 1:
Figure 4 shows that the major constituents being applicable to truck scale weighing section of the present invention: the output terminal connection data harvester of each LOAD CELLS, data collector comprises modulate circuit 1 and analog to digital conversion circuit 2; Weighing-up wave carries out through amplifying in modulate circuit 1, filtering process, and carry out being converted to digital signal through analog to digital conversion circuit 2, digital data transmission is to microprocessor 3; Microprocessor also configures power module 4, keyboard 6, display 7 and communication interface 8; Power module 4 is the power supplies such as modulate circuit 1, analog to digital conversion circuit 2, microprocessor 3.During training, microprocessor 3 connects outer computer 5 by communication interface 8; During online weighing, outer computer 5 is withdrawn.
Embodiment 2: the method for actual weighting model training and step.
In the present embodiment, truck scale have 8 tunnel LOAD CELLS ( n=8), range is 40 tons, and the max cap. of every road LOAD CELLS is 20 tons, and the number of divisions is 4000, verification scale interval eand actual graduation value dbe 10kg; Microprocessor 3 adopts the High Performance SCM MSP430F449 of TI company.See Fig. 1, the step of training actual weighting model is as follows:
(1) desirable weighting model is constructed, the constraint condition as True k Scale Weighing System: with nthe weighing-up wave that road LOAD CELLS exports for input, A (X) are for exporting, the desirable weighting model of structure truck scale, as shown in Equation (1);
(2) structure is based on the actual weighting model of truck scale of neural network: with No. 8 sensor weighing-up waves be input, with truck scale weighing results for output, construct three layers of BP neural network that one 8 input 1 exports, as the actual weighting model of truck scale, its structure as shown in Figure 3.The hidden layer neuron number of this neural network mmeet , kget 1 ~ 10. kcan be determined by following methods: first make k=1, calculate the error of neural network, if error meets the demands or minimum, then determine knumerical value; Otherwise, kadd 1, recalculate the error of neural network, until meet the demands or error minimum.By test of many times, the present embodiment is determined k=2, i.e. hidden layer neuron number m=5.In Fig. 3, for of input layer iindividual neuron is to of hidden layer mneuronic weights; , , be respectively the 1st of hidden layer the, 2, mindividual neuronic bias; for output layer is to hidden layer mindividual neuronic weights; b (2)for output layer bias; , , be respectively the 1st of hidden layer the, 2, mindividual neuronic output; f 1for hidden layer excitation function, it adopts Log-Sigmoid function; Output layer excitation function adopts linear function, and therefore network exports represent with formula (3);
(3) objective function of actual weighting model training is constructed: with the desirable weighting model of truck scale and derivative thereof for constraint condition, structure training sample objective function, represents with formula (4);
(4) standard test weight is loaded, gather weighing-up wave, form training and testing sample: the standard test weight utilizing the different tonnages such as 0.5 ton, 1 ton, 3 tons, 6 tons, 12 tons, 18 tons, 24 tons, 36 tons, be carried in the diverse location of truck scale body respectively, system is by LOAD CELLS, modulate circuit 1, analog to digital conversion circuit 2 and microprocessor 3, gather 50 group of 8 road load cell signal, obtain 50 groups of samples , as the formula (11), wherein 30 groups as train samples, and 20 groups are used for neural network test sample book.These samples are sent to outer computer 5 by communication interface 8 and preserve by microprocessor 3, and for neural network off-line training is prepared, computing machine here refers to outer computer;
(11)
(5) coefficient of desirable weighting model is asked: utilize formula (10) to ask the coefficient of desirable weighting model p i ;
(6) setting training initial parameter: target square error MSE is 0.0000000001, learning rate ηbe 0.8, coefficient μ j determine online, hidden neuron number mbe 5, frequency of training is 10000.
(7) get training sample, neural network training, regulate neural parameter w, b (1), vand b (2): from the sample set be kept at outer computer, take out training sample, and utilize special training software to train, utilize formula (9) to adjust parameter simultaneously w, b (1), vand b (2); Special training software adopts the training method shown in formula (4) ~ (10), utilizes MATLAB and the exploitation of Virtual Instrument LabVIEW Platform Designing, other programming languages also can be utilized to realize.
(8) enter determining program " training completes ", if completed, obtain the qualification parameters of actual weighting model w, b (1), vand b (2); Otherwise, return step and " get training sample " and restart next round training;
(9) outer computer by communication interface by actual weighting model parameter downloads in microprocessor, and preserve in the memory unit, for online weighing is prepared; Withdraw outer computer, i.e. the connection of disconnecting external computing machine and microprocessor simultaneously, terminate training.
According to neural network design theory (" neural network design ", the work such as Martin T. Hagan, Dai Kuiyi, China Machine Press, 2005,8), according to traditional neural network training method (namely not utilizing the desirable weighting model of truck scale), at least need training sample number Num=( m+ 1) * n+ ( m+ 1) * l, in formula, mfor hidden neuron number, ninput layer number, lfor output layer number.In the present embodiment, m=5, n=8, l=1, therefore Num=54, namely at least need 54 groups of training samples, otherwise larger extensive error can be produced, thus cause weighting model unavailable.But because the desirable weighting model that present invention utilizes truck scale is as priori, only make use of 30 groups of training samples can meet the demands, and training sample number far fewer than 54 groups, thus decreases workload.
Embodiment 3: online weighing.
In the present embodiment, truck scale have 8 tunnel LOAD CELLS ( n=8), range is 40 tons, and the max cap. of every road LOAD CELLS is 20 tons, and the number of divisions is 4000, verification scale interval eand actual graduation value dbe 10kg; Microprocessor 3 adopts the High Performance SCM MSP430F449 of TI company, utilizes in embodiment 2 and has trained qualified actual weighting model to carry out online weighing.See Fig. 2, online weighing step is as follows:
(1) tested load loads: namely truck is carried on truck scale body optional position, carries out weighing-up wave collection after stable;
(2) microcontroller acquires nthe weighing-up wave of road LOAD CELLS: system, by LOAD CELLS, modulate circuit 1, analog to digital conversion circuit 2 and microprocessor 3, gathers 8 tunnel LOAD CELLS output signals x i ( i=1,2 ..., 8), as the input vector X of actual weighting model, namely X=[ x 1, x 2..., x 8];
(3) call w, b (1), vand b (2): the parameter calling actual weighting model from storage unit w, b (1), vand b (2);
(4) output of actual weighting model is calculated: calculate the output obtaining neural network according to formula (3), this output is the final weighing results of truck scale after error compensation;
(5) show weighing results over the display, and terminate this online weighing.
The truck scale adopting the method for the invention after tested after, obtain weighing and error compensation effect as shown in Figure 5.Fig. 5 (a) carries out the forward and backward output weighing results of online weighing and error compensation for adopting the method, and Fig. 5 (b) be that the error adopting the method to compensate front and back contrasts.As can be seen from Fig., adopt the truck scale weighing results of the method for the invention accurate, error compensation is effective.
The concerned countries standard utilizing truck scale to examine and determine, to adopting the truck scale of the method for the invention to carry out on-the-spot test, table 1 is uneven loading error verification result, and table 2 is linearity error verification result.The parameter of truck scale: 8 tunnel LOAD CELLS ( n=8), range is 40 tons, the max cap. of LOAD CELLS is 20 tons, the number of divisions is 4000, verification scale interval eand actual graduation value dbe 10kg.
table 1 uneven loading error verification result
Position 1# 2# 3# 4# 5# 6# 7# 8#
Error E(kg) -7 2 -5 -6 -8 7 4 -5
Permissible error (kg) ±10 ±10 ±10 ±10 ±10 ±10 ±10 ±10
table 2 linearity error verification result
Weighing range (t) 0~0.2 0.2~5 5~10 10~20 20~40
Measurement error (kg) +3 -5 -8 -9 +12
Permissible error (kg) ±5 ±5 ±10 ±10 ±15
In table 1,1# represents No. 1 loading area (i.e. region, LOAD CELLS position), other have same meaning.As can be seen from table 1,2, the truck scale uneven loading error adopted in this way is less than the permissible error that concerned countries standard specifies, linearity error is less than the permissible error that concerned countries standard specifies equally, and error compensation is respond well.
Above embodiment is not as a limitation of the invention.

Claims (5)

1. be applicable to a Weighing method for truck scale, described Weighing method uses LOAD CELLS, data collector, microprocessor and display; Described LOAD CELLS is connected with microprocessor by data collector; Described display is connected with microprocessor; Described Weighing method comprises foundation and to weigh mathematical model, weighing-up wave collection, online weighing; It is characterized in that: the weigh mathematical model of mathematical model of described foundation comprises desirable weighting model, actual weighting model, and its step is as follows;
1) described desirable weighting model is linear function; Being input as of described linear function nthe data that road LOAD CELLS exports x ( i) , export as A (X); Its Input output Relationship expression formula is formula (1):
(1);
In formula, p i for the gain coefficient of LOAD CELLS, its value obtains by carrying out training to desirable weighting model;
2) described actual weighting model is three layers of BP neural network, and ground floor is input layer, and the second layer is hidden layer, and third layer is output layer, and their network structure is as follows respectively:
The neuronal quantity of input layer nfor the number of LOAD CELLS;
The neuronal quantity of hidden layer , in formula: k=1 ~ 10 is correction factor; lfor the neuronal quantity of output layer; Hidden layer excitation function adopts Log-Sigmoid function, namely exports a m (1)for formula (2):
(2);
The neuronal quantity of output layer lit is 1; Output layer excitation function adopts linear function; The neural network of output layer exports for formula (3):
(3);
In formula, wfor neural network input layer is to the weight matrix of hidden layer, b (1)for hidden layer bias vector, vfor hidden layer is to the weight vector of output layer, b (2)for output layer bias, xfor neural network input vector, w mi for input layer iroad is input to of hidden layer mindividual neuronic connection weights, b m (1)for hidden layer mindividual neuronic bias, v m for hidden layer mindividual neuron to the connection weights of output layer, x( i) be input layer iroad inputs;
3) before truck scale drops into online weighing, the training of set point number must be carried out to desirable weighting model and actual weighting model, training process carries out when microprocessor is connected with outer computer, with desirable weighting model and derivative thereof for constraint condition, finally obtain actual weighting model parameter w, b (1), vwith b (2)preserve in the microprocessor, then withdraw outer computer; Special training software is installed in outer computer;
The step of training actual weighting model is as follows:
I) gather training sample: the standard test weight preparing some, each standard test weight quality is different, by the standard test weight random loading of different quality on truck scale body, nroad sensor just has nindividual output data, nindividual output data and corresponding standard test weight quality form one group of training sample and are kept in outer computer;
II) construct training objective function, its relational expression is formula (4):
(4);
III) ask the derivative of the desirable weighting model of truck scale, its relational expression is formula (5):
(5);
IV) ask the weight coefficient of desirable weighting model in training objective function μ j , its relational expression is formula (6):
(6);
V) ask third layer output layer derivative, its relational expression is formula (7):
(7);
VI) ask respectively w, b (1) , Vwith b (2)increment Delta v m , Δ b (2), Δ w mi , Δ b m (1), and right w, b (1) , Vwith b (2)upgrade, their relational expression is respectively formula (8), (9):
(8)
(9)
In formula (9), v new m , b (2) new , w new mi , b m (1) new be respectively v m , b (2), w mi , b m (1)value after renewal, v old m , b (2) old , w old mi , b m (1) old be respectively v m , b (2), w mi , b m (1)value before renewal;
Vii) arrange training starting condition, carry out setting training according to formula (7) ~ (9), the error amount that training is produced, in setting range, obtains the weight matrix of input layer to hidden layer respectively w, hidden layer bias vector b (1), hidden layer is to the weight vector of output layer v, output layer bias b (2)end value, and be kept in the storage element of microprocessor, for called during truck scale online weighing;
In the training process to desirable weighting model, first utilize the training sample gathered in actual weighting model training process, then utilize least square method to train, obtain final coefficient p i , its relational expression is formula (10):
(10)
In formula (10), pfor p i vector, namely p=[ p 1, p 2..., p n ] t, nfor the number of LOAD CELLS; x=[ x 1, x 2..., x j..., x k ] be input amendment matrix, y=[ y 1, y 2, y j ..., y k ] be output sample matrix.
2. the Weighing method being applicable to truck scale according to claim 1, is characterized in that: described online weighing should carry out after actual weighting model training is qualified, and its step is as follows:
1) will collect nthe weighing-up wave of road LOAD CELLS is as the input vector of BP neural network ground floor x;
2) by input vector xwith the actual weighting model parameter of preserving in the memory unit w, b (1), vand b (2)substitute into together in formula (3), try to achieve BP neural network and export for final weighing results;
3) display shows final weighing results.
3. the Weighing method being applicable to truck scale according to claim 1 and 2, it is characterized in that: the method that described weighing-up wave gathers is as follows, the data obtained after the output signal of each road LOAD CELLS is carried out signal amplification, filtering and analog-to-digital conversion process, train the input vector with online weighing as desirable weighting model and actual weighting model x.
4. the Weighing method being applicable to truck scale according to claim 1 and 2, is characterized in that: described microprocessor is single-chip microcomputer, dsp processor or other embedded system devices, and with storage unit.
5. the Weighing method being applicable to truck scale according to claim 3, is characterized in that: described microprocessor is single-chip microcomputer, dsp processor or other embedded system devices, and with storage unit.
CN201310177182.5A 2013-05-14 2013-05-14 Weighing method applicable to truck scale Expired - Fee Related CN103234610B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310177182.5A CN103234610B (en) 2013-05-14 2013-05-14 Weighing method applicable to truck scale

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310177182.5A CN103234610B (en) 2013-05-14 2013-05-14 Weighing method applicable to truck scale

Publications (2)

Publication Number Publication Date
CN103234610A CN103234610A (en) 2013-08-07
CN103234610B true CN103234610B (en) 2015-05-06

Family

ID=48882661

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310177182.5A Expired - Fee Related CN103234610B (en) 2013-05-14 2013-05-14 Weighing method applicable to truck scale

Country Status (1)

Country Link
CN (1) CN103234610B (en)

Families Citing this family (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105973443B (en) * 2016-05-17 2018-09-04 大连理工大学 Digital automobile weighing apparatus uneven loading error modification method based on least square method
CN105973444B (en) * 2016-06-25 2018-09-28 湖南师范大学 A kind of improved truck scale weighing method
CN107607182B (en) * 2017-08-04 2019-12-13 广西大学 truck weighing system and weighing method
CN108491404B (en) * 2018-01-22 2022-07-01 国电南瑞科技股份有限公司 State estimation bad data identification method based on BP neural network
CN109377046A (en) * 2018-10-18 2019-02-22 上海经达信息科技股份有限公司 Overload of vehicle method of discrimination, system and device based on BP neural network
CN109579967A (en) * 2018-11-27 2019-04-05 上海交通大学 Intelligent Dynamic weighing method and system
CN109668610A (en) * 2019-01-11 2019-04-23 东南大学 The system of vehicle dynamically weighting method and its use based on neural net regression
CN112129386A (en) * 2019-06-24 2020-12-25 梅特勒-托利多(常州)测量技术有限公司 Weighing apparatus and weighing method
CN112393794B (en) * 2019-08-18 2024-01-26 华东理工大学 Diagnosis and reading correction method for single sensor fault or unbalanced load of platform scale
CN112747813A (en) * 2019-10-29 2021-05-04 杭州壬辰科技有限公司 Automatic part weighing and counting system and method
CN111664927B (en) * 2020-05-28 2022-04-26 首钢京唐钢铁联合有限责任公司 Method and device for judging metering state of rail weigher
DE102020007453A1 (en) 2020-12-07 2022-06-09 Daimler Truck AG Method for determining a vehicle mass average and its output in the vehicle
CN112964345A (en) * 2021-02-07 2021-06-15 广东电子工业研究院有限公司 Freight car weighing system and weighing method thereof
CN114543952A (en) * 2022-01-06 2022-05-27 红云红河烟草(集团)有限责任公司 Error self-compensation method of electronic belt scale for cigarette production
CN115790804B (en) * 2023-02-08 2023-04-21 福建省智能交通信息工程有限公司 Dynamic truck scale state monitoring method, medium, equipment and device
CN116481626B (en) * 2023-06-28 2023-08-29 深圳市汉德网络科技有限公司 Vehicle-mounted weighing self-adaptive high-precision calibration method and system

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5228113A (en) * 1991-06-17 1993-07-13 The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration Accelerated training apparatus for back propagation networks
CN102506983A (en) * 2011-10-31 2012-06-20 湖南师范大学 Weighing error automatic compensation method of vehicle scale

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050187643A1 (en) * 2004-02-19 2005-08-25 Pavilion Technologies, Inc. Parametric universal nonlinear dynamics approximator and use
US20040148144A1 (en) * 2003-01-24 2004-07-29 Martin Gregory D. Parameterizing a steady-state model using derivative constraints

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5228113A (en) * 1991-06-17 1993-07-13 The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration Accelerated training apparatus for back propagation networks
CN102506983A (en) * 2011-10-31 2012-06-20 湖南师范大学 Weighing error automatic compensation method of vehicle scale

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
基于先验知识和神经网络的非线性建模与预测控制;薛福珍;《系统仿真学报》;20040531;第16卷(第5期);1057-1063 *
如何增加人工神经元网络的透明度?;胡包钢等;《模式识别与人工智能》;20070228;第20卷(第1期);72-84 *
采用导数约束关系的前向神经网络学习优化方法研究;杨仁付;《中国优秀硕士学位论文全文数据库 信息科技辑》;20070316(第04期);36-39 *

Also Published As

Publication number Publication date
CN103234610A (en) 2013-08-07

Similar Documents

Publication Publication Date Title
CN103234610B (en) Weighing method applicable to truck scale
CN105973444B (en) A kind of improved truck scale weighing method
Wang et al. A novel safety assurance method based on the compound equivalent modeling and iterate reduce particle‐adaptive Kalman filtering for the unmanned aerial vehicle lithium ion batteries
CN102506983A (en) Weighing error automatic compensation method of vehicle scale
CN104134999B (en) Distribution network based on multi-data source measures the practical method of calculation of efficiency analysis
CN105116343B (en) The electrokinetic cell state of charge method of estimation and system of least square method supporting vector machine
CN101738331B (en) Tunnel construction simulation plane strain model test device
CN103457279A (en) Determination method for partition D-STATCOM integrated optimization configuration scheme of large-scale power grid
CN101651343B (en) Method and system for checking electric power system model based on hybrid dynamic simulation
CN100533311C (en) Embedded type weighing system based on intelligent fuzzy control
CN106682369A (en) Heating pipe network hydraulic simulation model identification correction method and system, method of operation
CN106569164A (en) Method and system for synchronization testing of electric quantity of double-core electric energy meter
CN108197725B (en) Water demand prior information-based water supply network node water demand checking method
CN102620939B (en) Engine torque predicting method and engine torque predicting device
CN104724638B (en) The method that forklift weighing device and fork truck are weighed automatically
CN107066712A (en) Hydraulic turbine model parameter identification method and device based on guide vane opening-power characteristic
CN112446091A (en) Artificial neural network-based pulsating pressure prediction method
CN103106305A (en) Space grid structure model step-by-step correction method based on actual measurement mode
CN103884931A (en) Testing and recording device for load characteristics of transformer substation bus
CN107462299A (en) A kind of data processing method that oil tank volume list demarcation is carried out with integration
CN106772204B (en) The emulation mode and device of electric energy meter metering circuit
CN106610478A (en) Energy storage battery characteristic estimation method and system based on mass data
CN105352571A (en) Granary weight detection method and device based on index relation estimation
CN108988347A (en) A kind of adjusting method and system that power grid Transient Voltage Stability sample set classification is unbalance
CN107861082A (en) A kind of calibration intervals of electronic metering equipment determine method and device

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

Granted publication date: 20150506

Termination date: 20170514

CF01 Termination of patent right due to non-payment of annual fee