0% found this document useful (0 votes)
30 views8 pages

Measurement: Sciencedirect

Download as pdf or txt
Download as pdf or txt
Download as pdf or txt
You are on page 1/ 8

Measurement 129 (2018) 625–632

Contents lists available at ScienceDirect

Measurement
journal homepage: www.elsevier.com/locate/measurement

Compensation of automatic weighing error of belt weigher based on BP T


neural network

Bingying Lia,b, , Yongxin Lia, Haitao Wangb, Yuming Mab, Qiang Hub, Fangli Gea
a
School of Mechanical Engineering, Nanjing University of Science and Technology, Nanjing, China
b
Jiangsu Institute of Metrology, Nanjing, China

A R T I C LE I N FO A B S T R A C T

Keywords: A belt weigher is widely used in industrial production and trade settlement; however, it is difficult to maintain its
Error compensation nominal measuring accuracy in service. With the belt weigher indication, average speed of belt, variation in belt
BP neural network sag, running deviation of belt, environmental temperature, and humidity as inputs, and the control instrument
Automatic weighing indication as output, a BP neural network model to compensate the automatic weighing error of the belt weigher
Dynamic material experiment
is built. We obtained the sample data depending on the test system for the type evaluation of the belt weigher in
Belt weigher
the Jiangsu Institute of Metrology, as well as supplemental relevant parameters for the monitoring devices. The
BP model is trained and validated using MATLAB. The validation results show that the absolute value of the
maximum relative automatic weighing error output by the BP model is less than 0.5%, significantly lower than
the error before using the BP model for compensation. The BP model is effective, feasible, and practical for
compensating the automatic weighing error of the belt weigher.

1. Introduction attracted much attention. In Germany (1962), A. Malting and P.


Vierling first added the mechanics characteristic of a material in the
An electronic belt weigher (hereinafter referred to as belt weigher) dynamics analysis of belt and changed the model of the belt from an
can dynamically weigh bulk materials without interrupting the move- elastic rod to a viscoelastic rod [4]. In the former Soviet Union (1967),
ment of the conveyor or quantitative segmenting material and is critical experts calculated the complete dynamic equation of stress wave in belt
in industrial production and trade settlement [1]. The International transmission [4]. In Netherlands (1994), LODEWIJKS first proposed to
Recommendations (OIML R) 50 2014 (E) state that a belt weigher can simplify the belt as a unit model of beam and subsequently introduced
be divided into four accuracy classes as follows: 0.2, 0.5, 1, and 2 [2]. the dynamic analysis of the belt conveyor into a two-dimensional model
Typically, owing to the material flow rate, speed of belt, environment, [5]. In Australia (2016), Paul Munzenberger and Craig Wheeler built a
accumulated belt creep over time, and other influencing factors, it is laboratory test facility to measure the indentation rolling resistance of
difficult for a belt weigher to maintain its nominal measuring accuracy conveyor belts and improved the accuracy of the conveyor belt tension
during actual use. For example, for a belt weigher of accuracy class 0.5, [6]. In China, Bingao Hong and Yiping Wei established a mechanics
experienced technical personnel have found that its absolute value of model of the belt weigher with a single roller and compensated the
relative automatic weighing error in-service between two periodic error in its operation process based on mechanics analysis [7]. Bin Dan
verifications is typically more than 0.5% ruled in OIML R 50. However, of Chongqing University analysed all aspects of the influencing factors
the error is higher than 5% when it is used in a port, warehouse, and of the belt weigher in detail, built the mechanics model, and compen-
power plant with bad environmental conditions [3]. The lower mea- sated for errors in its operation process [8].
suring accuracy of the belt weigher negatively impacts the trade set- However, both in China and abroad, studies regarding the auto-
tlement of bulk materials, energy metering, energy conservation, and matic weighing error of a belt weigher primarily focus on analysing the
emission reduction. Therefore, the method to maintain a belt weigher mechanical properties of the belt and establishing a mechanical model
within the nominal level of accuracy becomes an urgent industry pro- to calculate and compensate for errors [4–11]. In fact, the influencing
blem and must be overcome before scientific and technological workers factors of a belt weigher in service are numerous, nonlinear, and in-
can be measured. teractive; therefore, it is impractical to determine their regularity and
Since the invention of the belt weigher, its measuring accuracy has build an accurate compensation model of error based only on


Corresponding author at: Jiangsu Institute of Metrology, Nanjing, China.
E-mail address: lby_8441@sina.com (B. Li).

https://doi.org/10.1016/j.measurement.2018.07.080
Received 11 August 2017; Received in revised form 9 July 2018; Accepted 25 July 2018
Available online 29 July 2018
0263-2241/ © 2018 Published by Elsevier Ltd.
B. Li et al. Measurement 129 (2018) 625–632

difficult to analyse. Ideally, the load cell bears the real load value of the
material on the weighing segment. In fact, owing to the conveyor belt
with properties of a viscoelastic material, it will respond to changes
when affected by external forces, and subsequently causes the load cell
to receive unexpected signals and output a signal of weighing gravity
WL̂ (t ) that is different from WL (t ), thus causing the error of weighing
gravity. The sources of error of the weighing gravity contain the ‘belt
effect’, vibration of belt, running deviation of belt, sag resistance,
centrifugal force, and running resistance of belt. Although the sources
of error of the weighing gravity are complicated, most of them can be
reduced, eliminated, or consolidated by appropriate methods. Finally,
they can be approximately summarised as the influence of tension and
running deviation of the belt [9–12].
Fig. 1. Assembly diagram of belt weigher.

(2) Speed-measuring error


theoretical analysis. The following method is more feasible and prac-
tical in comparison: first, analyse the sources of automatic weighing The speed measuring point is inconsistent with the stress point be-
error of the belt weigher, excavate their relatedness, and design a cause of the improper installation location, the skid phenomenon that
multiparameter monitoring program; subsequently, obtain sample data occurs randomly between the measuring wheel and conveyor belt, or
through proper calibration and, finally, determine the available em- the uneven speeds of the up and down surfaces of the belt. These may
pirical formula by supervised learning such as a neural network. cause the speed sensor to output a signal of belt speedthatisdifferentwith
that is different with, thus , thus leading to errors in speed measuring.
2. Analysis of the sources of automatic weighing error However, with the improvement in signal processing technology, the
error of speed measuring will not be the major source of automatic
2.1. Weighing principle of belt weigher weighing error provided that the installation location of the speed
sensor is reasonably selected.
A belt weigher dynamically, continuously, and automatically
weighs materials in a conveying process. Its assembly diagram is shown (3) Signal processing error
in Fig. 1.
Currently, a belt weigher obtains the totalised value of a material by In theory, the hardware and software designs of the functional units
integration: When the belt conveys a material to the weighing segment, such as the multiplier, integrator, or display may affect the weighing
the load cell obtains the instantaneous weight of the materialWL (t ) (kg), result. However, with the popularity of high-precision integrated chips
and the speed sensor obtains the instantaneous velocity of the belt and display modules as well as algorithm optimisations in signal pro-
(m / s ). Assumingthat (m/s). Assuming that the weighing length is (m cessing, signal processing errors will be negligible.
(m), and the duration of the whole weighing process is T(s), the tota-
lised value (kg ) of (kg) of the material conveyed can be calculated as [4] (4) Calibration error
T WL (t ) v (t )
I= ∫0 L
dt
(1)
A belt weigher requires regular calibrations to ensure its normal
operating state. Occasionally, owing to the limitations of conditions, a
The transmission chain of the corresponding measuring signal is belt weigher is calibrated under the static condition that is significantly
shown in Fig. 2 [11]. different from the actual operating condition, and may cause errors in
By observing Fig. 2, and one can intuitively infer that the process of its subsequent automatic weighing. In principle, calibration errors also
testing and transferring the signals of and and WL (t ) may cause an affect the accuracy of the output signal of the load cell. However, ca-
automatic weighing error. libration errors can be effectively avoided provided that the actual
material and the control instruments meet the regulation requirements
2.2. Sources of automatic weighing error during calibration.

The sources of automatic weighing error can be summarised in five (5) Environmental error
categories as follows [4]:
The belt weigher is often used in relatively poor conditions, e.g.
(1) Error of weighing gravity unstable temperature and humidity . The volatile temperature and
humidity will change the belt status and will significantly affect the
According to literature research, the error of weighing gravity stability and linearity of the weighing result of the belt weigher [13].
constitutes the largest source of automatic weighing error, i.e. ap- Temperature may also affect the electrical performance of the belt
proximately 85% of the total automatic weighing error [11]. The error weigher body and the performance of the load cell. Therefore, the
of weighing gravity has a complicated composition and is the most temperature and humidity should be monitored.

Fig. 2. Transmission chain of measuring signal.

626
B. Li et al. Measurement 129 (2018) 625–632

3. BP neural network model

A neural network is a mathematical algorithm model that processes


distributed parallel information using a physical device to simulate the
structure and function of a biological neural network; it is composed of
multiple inputs and single output neurons connected according to cer-
tain topological structures; further, it studies through sample training,
changes the weight value and threshold value of the internal connection
such that the error between the output value and target value is
minimal, and obtains a nonlinear mapping that can describe the re-
lationship between the input and output of a system [14]. A neural
network has massive parallel ability, distributed storage and processing Fig. 4. Schematic diagram of variation in belt sag.
power, powerful self-organising, self-adaptive and self-learning abil-
ities, as well as fault tolerant and generalisation ability without prior 3.1. Selection of model parameters
knowledge. It is suitable for processing problems that have inaccurate
and fuzzy information while considering many factors and conditions. Belt tension is the major source of automatic weighing error and the
Currently, as a typical neural network model, the multilayer feed- basic attributes of a belt weigher, and its influence on the weighing
forward neural network based on error back propagation (hereinafter result cannot be eliminated. In addition, it varies in real time with the
referred to as BP neural network) has been widely used because of its variation in material on the weighing section; therefore, its direct
excellent ability in arbitrary complex pattern classification and multi- measurement is difficult. Experts and scholars have performed sig-
dimensional function mapping, as well as its mature training method. In nificant investigations into the measurement methods of belt tension,
theory, a BP neural network with a single hidden layer can approximate and found that the variations in belt sag and belt tension are strongly
any nonlinear mapping provided that the sample is sufficient and the correlated. It is more practical to establish a relationship between
parameter settings of the network are reasonable. With the develop- tension and sag, and monitor the variation in belt sag between two
ment of artificial intelligence, an increasing number of researchers in weighing rollers indirectly to reflect the variation in belt tension [4,18].
China and abroad are using the BP neural network to compensate the The variation in belt sag is defined as the belt drop at midpoint of two
dynamic weighing error. For example, Bahar introduced a BP neural weighing rollers during automatic weighing relative to no-load, as
network with three layers into the dynamic weighing platform to si- shown in Fig. 4, where the variation in belt sag can be calculated as
mulate the dynamic weighing action and achieved precise dynamic Δh = |h−h 0 |.
weighing [15]. Jie Yan, Yan Zhao, and Junli Zhang introduced a BP The running deviation the belt is defined as the offset between the
neural network into the weighing system of concrete mixing to com- belt longitudinal centre line and the theory of the roller centre line. As
pensate for the nonlinear errors of its weighing sensor [16]. Yufeng shown in Fig. 5, the belt deviates to side B. The running deviation of the
Zhuang, Xiaojin Hu, and Yu Zhai introduced a BP neural network into belt can be calculated as Δl = |l1−l 0 |, and vice versa to side A.
the dynamic weighing system of a microscale capsule to compensate for According to the weighing principle above of the belt weigher, and
the nonlinear errors of its weighing sensor [17]. However, the research the analysis of sources of automatic weighing error, with the dynamic
based on BP neural network to compensate for the automatic weighing totalised value of the material indicated by the belt weigher (belt
error of a belt weigher has not been found. weigher indication), average speed of belt, average variation in belt sag,
A BP neural network contains an input layer, a hidden layer, and an average running deviation of belt, the temperature and humidity of the
output layer. A typical model of neurons is shown in Fig. 3. environment in one process of automatic weighing as inputs, and the
The mathematical model of the neurons in Fig. 3 can be described as true value of the totalised material indicated by a control instrument
follows: (control instrument indication) in one process of automatic weighing as
n n
output, a BP neural network model (hereinafter referred to as BP
sj = ∑ ωji xi−θj bj = ∑ ωji xi (x 0 = θj , ωj0 = −1) yj = f (sj ) model) containing six inputs, one output, and a hidden layer is built, as
i=1 i=0 (2) shown in Fig. 6.
Building a BP model requires the selection of activation functions in
the hidden layer and output layer. The activation functions add non-
sj: Weighted sum of inputs
linear factors to the neural network model, and preserve and map the
:Thresholdvalue : Threshold value
features of the input data. The primary function of a neural network
:Coefficientofconnectionweights : Coefficient of connection weights
model is to solve nonlinear problems. The method to select the acti-
:Transformationfunctionofoutput : Transformation function of output
vation functions based on different needs of data processing is im-
portant; In fact, researchers preferentially choose the tansig function as
In Fig. 3, the output of the neurons, yjis affected by the input signal,
the activation function of the hidden layer because the tansig function
xj(j = 1,2,…,n), as well as other internal factors of the neurons; there-
fore, an additional input signal θj is often added while building a neural
network model.

Fig. 3. Typical model of neurons. Fig. 5. Schematic diagram of running deviation of belt.

627
B. Li et al. Measurement 129 (2018) 625–632

f (x ) = x (4)

From the empirical formula in [17], the number of neuron nodes


() of ) of the hidden layer is calculated as

j= i+k +α (5)

:: Number of neuron nodes of input layer


:: Number of neuron nodes of output layer
:Oneconstantbetween1and10 : One constant between 1 and 10

After several simulation attempts, is is determined as eight.

3.2. Implementation of model algorithm

Fig. 6. Block diagram of BP model. The self-built BP model is trained by the built-in toolbox of the
forward neural network (feedforwardnet) under the operation en-
with nonlinearity, differentiability, and monotonicity can meet the vironment of MATLAB R2011b. The maximum number of training
basic requirements of the BP model [14]; in addition, it can compress (net.trainParam.epochs) is set as 1000, the learning rate
the continuous input real value between -1 and 1 and output 0 mean (net.trainParam.lr) is set as 0.01, and the minimum error of training
values to effectively guarantee that the update of the weight value is a target (net.trainParam.goal) is set as 0.0001. The feedforwardnet
gentle iterative change rather than a sawtooth mutation. The training toolbox will automatically divide the input data into training data,
speed and calculation precision of the tansig function are outstanding. validation data, and testing data with the ratios of 70%, 15%, and 15%,
The prototypes of the tansig function are as follows: respectively. To guarantee the number of training samples, the ratios
are changed to 85%, 8%, and 7%, respectively.
2
f (x ) = −1
1 + e−2x (3)
4. Acquisition of sample data
The training process of the BP neural network is as follows: the
input layer values spread separately to the hidden layer after network We obtained the experimental sample data depending on the test
computing, and subsequently spread to the output layer similarly. system for the type evaluation of the belt weigher in the National Type
Finally, the network calculates the error between the output value and Evaluation Laboratory of Automatic Weighing Instrument of the
sample value, and the network continues training until the error Jiangsu Institute of Metrology, as well as supplemental relevant para-
reaches a small set value. The purelin function can maintain the am- meters for the monitoring devices.
plification and reduction of arbitrary previous range values, and eases
the comparison between the output value and sample value. Therefore, 4.1. Test system for type evaluation of belt weigher
we selected the purelin function as the activation function of the output
layer. The system shown in Fig. 7 is equipped with eight conveyor belts
The prototypes of the purelin function are as follows: with different bandwidths. The maximum flow rate is up to 7200 t/h,

Device to Circulate Five Control Instruments


Material (Hopper weighers)

Eight Conveyor Belts With


Different Bandwidth

Fig. 7. The test system for type evaluation of belt weigher.

628
B. Li et al. Measurement 129 (2018) 625–632

the highest belt speed is 4 m/s. Its bandwidth, material flow rate, belt
speed, etc., can accommodate the belt weighers typically used in the
market. The system can real-time control the parameters such as uni-
formity of material delivery, speed of belt and feed flow, and meet the
experimental requirements of a belt weigher of each accuracy class. It
uses wear-resistant, high-hardness, and non-friable ceramic granules as
the experimental
material, with a total capacity of 150 tons. It is also equipped with
five hopper weighers as the control instruments. The system fully
complies with the requirements of material experiment stated in the
International Recommendations and Chinese National Procedures of
Belt Weigher.

4.2. Speed monitoring of belt

The speed monitoring of the belt involves a photoelectric speed


sensor that converts the length displacement of the belt into pulses by
photoelectric conversion to measure the speed, and is the most widely Fig. 9. Monitoring device of belt sag.
used currently. The disc with holes of photoelectric speed sensors is
installed in the upper location and measures the belt speed of the
weighing segment directly; therefore, the speed error minute, as shown
in Fig. 8.

4.3. Monitoring of belt sag

The monitoring of belt sag involves the German industrial laser


rangefinder sensor LEAZE LOD2, which is installed at the midpoint of
two weighing rollers, as shown in Fig. 9. After setting the sampling
frequency, the sensor can output the distance between the belt and the
reference point in real time. Assuming that the belt sag is zero during
no-load, the average variation in belt sag during loading can be cal-
culated.

4.4. Monitoring of running deviation of belt

The monitoring of the belt’s running deviation involves the dis-


placement sensor of the grating line; Fig. 10 shows the installation. Fig. 10. Monitoring device of running deviation of belt.
The sensor contains aluminium extrusions, a grating ruler, a reading
head, a signal cable (with plug), etc. As a reaction device, the grating
ruler reflects the belt displacement by outputting a photoelectric signal is zero during no-load, the average running deviation of the belt during
when the belt deviates during testing. To facilitate the collection and loading can be calculated.
processing of displacement data, the sensor processes the photoelectric The material experiment mimics the actual use condition of the belt
signal by reshaping and matching, and subsequently transmits it to the weigher. The experiment uses a data acquisition card (NI usb-6281) to
display box by a signal cable. The display box changes the photoelectric complete the data acquisition of relevant parameters. The temperature
signal into a displacement digital signal to complete the measurement and humidity are directly obtained with the temperature and humidity
of linear displacement. Assuming that the running deviation of the belt metre, respectively, considering that the temperature and humidity of
the natural environment vary, in addition to altering the belt speed,
feed flow, etc. to obtain the experimental sample data. The material
experiment was primarily performed on the belt with 0.8-m bandwidth
and around the totalised values of 10 t, 18 t, 27 t, 35 t, and 45 t; 13
groups of sample data each, and a total of 65 groups of sample data are
obtained.

5. Simulation and analysis

According to the collection order, 65 groups of sample data are


assigned with sequence numbers #1–#65. We selected 13#, 26#, 39#,
52#, and 65# as the validation data and other 60 groups of sample data
as the input data of the BP model. The training process of the BP model
is as follows:

Step 1: the maximum number of training is set as 1000, the learning


rate is set as u= 0.01, and the minimum error of the training target is
set as 0.0001.
Step 2: input the sample data, initialise the weight
Fig. 8. Monitoring device of belt speed.

629
B. Li et al. Measurement 129 (2018) 625–632

Fig. 11. Performance curve of model error.

matrixandthethresholdvalue , andsubsequentlynormalisethedata. and The trained BP model uses the five retained groups of sample data to
the threshold value, and subsequently normalise the data. validate its effect. Table 1 shows the validation results and the com-
Step 3: calculate the weighted sums of the input of the eight neurons pensation results of the automatic weighing error using the BP model.
in the hidden layer and the output value of the model. In Table 1, the output value of the BP model is closer to the true
Step 4: the mean square error (MSE) of the output value of the value of the material. The absolute value of the maximum relative
model and the sample value are calculated. If the MSE does not automatic weighing error output by the BP model is less than 0.5%, and
reach the training target, the error needs to be propagated backward is significantly lower than the error before using the BP model for
by differentiating the MSE with respect to the weight to obtain the compensation. The validation results show that the BP model exhibits
gradient.. effectiveness, high precision, and better generalisation ability to var-
Step 5: update the weight matrix as follows: iations in belt tension, belt speed, running deviation, temperature, and
humidity. In addition to the belt weigher with the same weighing
(1) W= (0)W - u × D(1) (6)
principle but different bandwidth, new experimental sample data can
be obtained and the BP model can be subsequently retrained, such that
Step 6: the BP model recalculates using the updated weights and the BP model has universal applicability.
iterates continuously until the MSE is less than the training target. However, limited by the cost of the dynamic material experiment on
the test system for the type evaluation of the belt weigher, we obtained
Under the MATLAB software platform and according to the training only 65 groups of sample data experimentally. Therefore, the re-
process above, the BP model was simulated and trained several times. presentativeness of the sample data is deficient. Moreover, it cannot
Subsequently, the ideal performance curve of the model error (Fig. 11), artificially change the temperature and humidity in the laboratory;
and the regression curve of the model error (Fig. 12) of the BP model therefore, the training sample is unbalanced relative to the temperature
are obtained. Figs. 11 and 12 show that the training results achieve the and humidity. For future work, we will continually collect sample data
training target after the seventh iteration, as well as the small number with different environment temperatures and humidities during dif-
of iterations and short training time. ferent seasons to improve the accuracy and generalisation of the BP
In Fig. 11, the abscissa represents the number of iterations and the model, to achieve the goal of maintaining the error in-service within its
ordinate represents the MSE between the output value of the BP model nominal range for a long time.
and the sample value. The smaller the MSE, the closer is the output
value of the BP model to the sample value. The ‘train’ blue line shows 6. Conclusions
the MSE performance in each iteration in the BP training process; the
‘validation’ green line shows the MSE performance in each iteration in (1) The automatic weighing process of a belt weigher can be modelled
the BP cross validating process; the ‘test’ red line shows the MSE per- using the BP neural network.
formance in each iteration in the BP testing process. The ‘best’ dotted (2) A simple and practical dynamic totalised method of belt weigher is
line shows that the training results are the most ideal after the seventh proposed based on the BP neural network, while complex me-
iteration. The ‘goal’ dotted line represents one of the set targets for the chanical analysis and cross influence of parameters are avoided.
BP model to stop training. (3) The average belt speed, average variation in belt sag, average
The BP model synchronously predicts the training sample, valida- running deviation of the belt, and the environment temperature and
tion sample, testing sample, and all samples while training by the re- humidity during automatic weighing are influencing factors of the
gression equation. Fig. 12 shows the regression results. The abscissa weighing results. The material experiment primarily depended on
represents the target value; the ordinate represents the output value of the test system for the type evaluation of the belt weigher in the
the BP model; ‘R’ is the regression coefficient. The closer is the value of National Type Evaluation Laboratory of Automatic Weighing
‘R’ to one, the better is the BP model and the higher is the prediction Instrument of the Jiangsu Institute of Metrology. We used the
accuracy.

630
B. Li et al. Measurement 129 (2018) 625–632

Fig. 12. Regression curve of model error.

Table 1 model. Establishing an experimental database is of great sig-


Compensation results using BP model. nificance.
Sequence number 13# 26# 39# 52# 65#
Acknowledgement
Dynamic totalised value of material 9969 18,624 27,288 34,823 44,889
(kg) The authors would like to acknowledge the financial support pro-
Average speed of belt (m/s) 0.99 1.78 1.48 1.98 2.00
vided by the General Administration of Quality Supervision, Inspection,
Average variation in belt sag (mm) 13.92 15.75 16.49 17.34 16.66
Average running deviation of belt 0 4 4 5 7 and Quarantine of the People’s Republic of China (AQSIQ,
(mm) 2010QK171), and the Quality and Technical Supervision Bureau of
Temperature (°C) 30.1 32.0 27.7 28.4 26.5 Jiangsu Province (JSQTS, KJ112511).
Relative humidity (%) 60 49 48 68 60
True value of material (kg) 9859 18,419 27,152 34,405 44,305
Output value of BP model (kg) 9908 18,433 27,235 34,241 44,413 References
Error before compensation (%) 1.12 1.11 0.50 1.21 1.32
Error after compensation (%) 0.50 0.08 0.31 -0.48 0.24 [1] Changyan Shi, State of arts and trends for the dynamic weighing and force mea-
suring technology, ACTA Metrologica Sinica 22 (3) (2001) 201–205 (In Chinese).
[2] OIML R 50-1:2014 (E), Continuous totalizing automatic weighing instruments (belt
weighers)-Part 1: Metrological and technical requirements, OIML.
experimental sample data to train and validate the self-built BP [3] Wei Liu, Yongliang Zhang, Improvement of the detection ability of automatic
model. The validation results showed that the absolute value of the weighing to promote the development of industry and energy conservation and
maximum relative automatic weighing error output by the BP emissions reduction, in: 10th Weighing Technology Seminar, Shanghai, China,
2011.
model is less than 0.5%, and is significantly lower than the error [4] Qi Chu, The Method and Experimental Research on Dynamic Weighing of Belt
before using the BP model. Conveyor, China University of Mining and Technology, 2014 (Ph.D Thesis).
(4) The BP model has better generalisation ability for belt weighers [5] G. Lodewijks, On the application of beam elements in finite element models of belt
conveyors part I, Bulk Solids Handling 14 (4) (1994) 729–737.
with different belt speeds, different belt tensions, different running
[6] Paul Munzenberger, Craig Wheeler, Laboratory measurement of the indentation
deviations, and different environment temperatures and humidities. rolling resistance of conveyor belts, Measurement 94 (2016) 909–918.
The belt weigher with the same weighing principle but different [7] Bingao Hong, Yiping Wei, A high-accuracy dynamic weighing system based on
bandwidths can refer to the BP model. The BP model has universal sigle-idle conveyor belt, in: Proceeding of the Eighth International Conference on
Machine Learning and Cybernetic, Baoding, China, 2009, pp. 2483–2487.
applicability. [8] Bin Dan, The Study on Dynamic Measurement Technology and System of Material
(5) The BP model requires a large amount of automatic weighing ex- Flow on Belt Conveyors in Manufacturing Systems, Ph.D Thesis Chongqing
perimental data of the belt weigher. Experiments should ensure the University, China, 1997.
[9] Daqian Dong, Research on Dynamics Behaviour of Large Scale Belt Conveyor and Its
sample size to guarantee the robustness and applicability of the

631
B. Li et al. Measurement 129 (2018) 625–632

Application, North China Electric Power University, 2008 (Ph.D Thesis). [15] H.B. Bahar, D.H. Horrocks, Dynamic weight estimation using an artificial neural
[10] Fansheng Wang, Study on the Flexible Multi-body Dynamics Analysis Method of network, Artificial Intell. Eng. 12 (1–2) (1998) 135–139.
Belt Conveyor, China University of Mining and Technology, 2010 (Ph.D Thesis). [16] Jie Yan, Yan Zhao, Junli Zhang, Study on static nonlinear error compensating for
[11] Kun Chen, Static And Dynamic Analysis On The Measurement System Of An weighing sensor based on BP neural network, Chinese J. Sensors Actuators 21 (6)
Electronic Belt Scale, Anhui Polytechnic University, China, 2014 (M.D Thesis). (2008) 1025–1028.
[12] Liang Zhu, Study on the Accuracy Compensation of Continuous Weighing System of [17] Yufeng Zhuang, Hu. Xiaojin, Yu. Zhai, Nonlinear compensation of micro scale
Bulk Materials, Nanjing University of Science & Technology, China, 2015 (Ph.D capsule dynamic condition weighing unit based on BP neural network model,
Thesis). Chinese J. Sci. Instrument 35 (8) (2014) 1914–1920.
[13] Yanqiang Yuan, Discussion of error theory and durability of Belt scale, 11th [18] Song Mei, Yuee Zhou, et al., The relational analysis of sag and tension based on the
Weighing Technology Seminar, Nanjing, China, (2012). method of least square curve fitting, Manuf. Automation 19 (2012) 123–126 (In
[14] Martin T. Hagan, Howard B. Demuth, Mark H. Beale, Neural Network Design, PWS Chinese).
Publishing Company, USA, 1996.

632

You might also like