Invention content
It is an object of the present invention to provide a kind of oil truck oil and gas leakage intelligent monitor system based on wireless sensor network, should
System monitors oil truck in traveling humiture on the way, pressure and the security information such as whether leaks in real time, to avoid unexpected thing
Therefore generation problem.
The invention is realized by the following technical scheme:
A kind of oil truck oil and gas leakage intelligent monitor system based on wireless sensor network is realized and is joined on way to oil truck
Number is detected and oil truck oil and gas leakage concentration intelligent predicting, it is characterised in that:The intelligent monitor system includes being based on nothing
The oil truck of line sensor network is in the acquisition of way state parameter and intelligent predicting platform, oil truck oil and gas leakage concentration intelligent predicting
Model;
It is described to be acquired with intelligent predicting platform by multiple inspections in way state parameter based on the oil truck of wireless sensor network
Node and on-site supervision end composition are surveyed, and it is flat in the acquisition of way state parameter and intelligent predicting to be built into oil truck in an ad-hoc fashion
Platform;The detection node is made of sensor group module, microcontroller and wireless communication module NRF2401, is responsible for detection oil truck
The temperature on surface, the actual value of pressure and gas concentration;It realizes and oil truck is managed in way parameter in the on-site supervision end
It is predicted in way leakage gas concentration with to oil truck;
The oil truck oil and gas leakage concentration intelligent forecast model includes multiple NARX neural network prediction models, Fuzzy C-
Means clustering algorithm grader, multiple empirical mode decomposition models, multigroup ANFIS Network Prediction Models and the ELman based on PSO
Network oil truck oil and gas leakage concentration Fusion Model;Each NARX neural network prediction models output is used as Fuzzy C-Means Clustering
The input of algorithm classification device, Fuzzy C-Means Clustering Algorithm grader carry out multiple NARX neural network prediction models output valves
Classification, each type of NARX neural network prediction models export the input as each empirical mode decomposition model, Mei Gejing
Test input of the multiple outputs of mode decomposition model as every group of ANFIS Network Prediction Model, every group of ANFIS Network Prediction Model
Input of the fusion forecasting value as the ELman network oil truck oil and gas leakage concentration Fusion Models based on PSO, based on PSO's
The output valve of ELman network oil truck oil and gas leakage concentration Fusion Models is that oil truck leaks gas concentration value.
The further Technological improvement plan of the present invention is:
The multiple NARX neural network prediction models carry out the oil and gas leakage concentration of each test point in oil truck surface
The input of prediction, multiple NARX neural network prediction models is each test point gas concentration value, Fuzzy C-Means Clustering Algorithm
Grader exports multiple NARX neural network prediction models according to the characteristic value that multiple NARX neural network prediction models export
Value is classified.
The further Technological improvement plan of the present invention is:
The output valve of each type NARX neural network prediction models is defeated as each empirical mode decomposition model
Enter, the output valve of each type NARX neural network prediction models is decomposed into low frequency trend portion by each empirical mode decomposition model
Point and multiple high-frequency fluctuation parts, low frequency trend part and multiple high-frequency fluctuation parts are respectively as every group of ANFIS neural network forecast
The input of model, the output equal weight of every group of each ANFIS Network Prediction Model mutually sum it up to obtain every group of ANFIS neural network forecast mould
The fusion forecasting value of type.
The further Technological improvement plan of the present invention is:
The fusion forecasting value of every group of ANFIS Network Prediction Model is as the ELman network oil truck oil gas based on PSO
The input for leaking concentration Fusion Model, the ELman network oil truck oil and gas leakage concentration Fusion Models based on PSO are realized to multigroup
The fusion of the fusion forecasting value of ANFIS Network Prediction Models leaks gas concentration value as oil truck.
Compared with prior art, the present invention having following obvious advantage:
One, the input of the used NARX neural network prediction models of the present invention includes one section of oil truck oil and gas leakage concentration
Time outputs and inputs historical feedback, and the input of this partial feedback is it is considered that contain the oil truck oil and gas leakage of a period of time
CONCENTRATION STATE historical information participates in the prediction of oil truck oil and gas leakage concentration, for a suitable feedback time length, prediction
Good effect is obtained, the NARX neural network prediction patterns of this patent provide the effective prediction oil truck oil gas of one kind and let out
Leak concentration method.
Two, NARX neural network prediction models of the present invention are that one kind can be effectively dense to oil truck oil and gas leakage
The dynamic neural network model that non-linear, the nonstationary time series of degree are predicted, can be in the non-stationary drop of time series
The precision of prediction to oil truck oil and gas leakage concentration-time sequence is improved in the case of low.With traditional prediction model method phase
Than the method has processing nonstationary time series effect good, and calculating speed is fast, the high advantage of accuracy rate.By to non-stationary
Oil truck oil and gas leakage concentration experiment data practical comparison, this patent demonstrates NARX neural network prediction models to oil tank
The feasibility of vehicle oil and gas leakage concentration-time sequence prediction.Meanwhile experimental result also demonstrates NARX neural network prediction models
It is more excellent than conventional model performance in non-stable time series forecasting.
Three, the present invention utilizes NARX neural network oil truck oil and gas leakage concentration prediction models, is introduced due to passing through
Time delay module and output feedback establish the dynamic recurrent neural network of model, it will output and input vectorial Time-delayed Feedback and introduce network instruction
In white silk, new input vector is formed, it includes not only being originally inputted to have good non-linear mapping capability, the input of network model
Data also include the output data after training, and the generalization ability of network is improved, and makes it in non-linear oil truck oil gas
Leaking static neural network more traditional in concentration-time sequence prediction has better precision of prediction and adaptive ability.
Four, the present invention is exported original deformation NARX neural network prediction models by empirical mode decomposition model (EMD)
Sequence is decomposed into the component of different frequency range, each component shows the different characteristic information lain in former sequence.With drop
Low sequence it is non-stationary.High frequency section data correlation is not strong, and frequency is relatively high, represents the ripple components of original series, tool
There are certain periodicity and randomness, this is consistent with the cyclically-varying of oil truck oil and gas leakage concentration;Low-frequency component represents
The variation tendency of former sequence.It can be seen that EMD can decomposite step by step the ripple components of oil truck oil and gas leakage concentration, periodic component and
Trend components, each component decomposited itself include identical deformation information, reduce different characteristic to a certain extent
Interfering between information, and the original oil truck oil and gas leakage concentration Deformation Series of each component variation curve ratio decomposited are bent
Linear light is slided.It can be seen that EMD can effectively analyze the oil truck oil and gas leakage concentration deformation data under multifactor collective effect, decomposition obtains
Each component be conducive to the foundation of ANFIS Network Prediction Models and be better anticipated.ANFIS nets are established respectively using to each component
Network prediction model, to avoid extreme learning machine from inputting the problems such as randomness that dimension is chosen is lost with component information, first to each point
Phase space reconstruction is measured, is finally superimposed each component prediction result to obtain final fusion forecasting result.What case study showed to be carried
Fusion forecasting result has higher precision of prediction.
Five, the characteristics of present invention is according to oil truck oil and gas leakage concentration prediction parameter differences between samples, build fuzzy C-mean algorithm
Cluster (FCM) grader classify to oil truck oil and gas leakage concentration multiple spot forecast sample parameter, design multiple EMD models and
Multigroup ANFIS Network Prediction Models predict the sample parameter of oil truck oil and gas leakage concentration prediction again respectively, in oil tank
In vehicle oil and gas leakage concentration prediction continuous and discrete process, fully consider oil truck oil and gas leakage concentration when space characteristic,
Similar in the origin cause of formation, the data of relative homogeneous are extracted from the data of magnanimity grade, and to establish, specific aim is stronger, can more react
Random time stage oil truck oil and gas leakage concentration prediction model improves precision of prediction.
Six, multigroup ANFIS Network Prediction Models proposed by the present invention are a kind of fuzzy based on Takagi-Sugeno models
Inference system is the novel fuzzy inference system structure for organically combining fuzzy logic and neuroid, is passed using reversed
Algorithm and the hybrid algorithm adjustment premise parameter and consequent parameter of least square method are broadcast, and automatically generates If-Then rules.It is multigroup
ANFIS Network Prediction Models as a kind of very distinctive neural network, it is same have with arbitrary accuracy approach it is arbitrary linear and
The function of nonlinear function, and fast convergence rate, sample requirement are few.Model calculation speed is fast, as a result reliably, has obtained effect
Fruit.
Seven, ELman networks oil truck oil and gas leakage concentration Fusion Model of the present invention, the Elman god of the model
It is generally divided into 4 layers through network:Input layer, accepts layer and output layer, input layer, hidden layer and output at middle layer (hidden layer)
The connection of layer is similar to feedforward network, and the unit of input layer only plays signal transmission, the linear weighting effect of output layer unit.
Linearly or nonlinearly function can be used in the transmission function of implicit layer unit, accepts layer and is also known as context level or state layer, it is used
Remember the output valve of implicit layer unit previous moment, may be considered a primary delay operator.Elman type neural networks
Feature is that the output of hidden layer passes through the delay and storage of accepting layer, is linked to the input of hidden layer certainly, this to make it from connection mode
There is sensibility, the addition of internal feedback network to increase the energy that network basis is in reason multidate information the data of historic state
Power, to achieve the purpose that dynamic modeling.The characteristics of ELMAN type recurrent nerve metanetworks is that the output of hidden layer passes through structure list
Delay, the storage of member are linked to the input of hidden layer certainly, this so that it is had sensibility to the data of historic state from connection mode, interior
The addition of portion's feedback network increases the ability that network basis is in reason multidate information, is conducive to the modeling of dynamic process;The model
Using the feedback link of associated layers dynamic neuron, the information of future anticipation network and past prediction network is merged, is made
Network strengthens to the memory of time sequence signature information, to improve oil truck oil and gas leakage concentration prediction precision.
Specific implementation mode
In conjunction with attached drawing 1-7, technical solution of the present invention is further described:
One, the design of system general function
A kind of oil truck oil and gas leakage intelligent monitor system based on wireless sensor network of Patent design of the present invention, it is real
Now oil truck is detected in way surface temperature, pressure and leakage oil gas parameter and in way oil truck oil and gas leakage concentration intelligence
Prediction, the system is by the oil truck based on wireless sensor network in the acquisition of way state parameter and intelligent predicting platform and oil truck
Oil and gas leakage concentration intelligent forecast model two parts form.It is acquired in way state parameter based on the oil truck of wireless sensor network
Include detection node 1 and on-site supervision end 2 with intelligent predicting platform, they are built into wireless test and control network in an ad-hoc fashion
Realize the wireless communication between detection node 1 and on-site supervision end 2;Detection node 1 is by the oil truck of detection in way state parameter
It is sent to on-site supervision end 2 and preliminary treatment is carried out to sensing data;On-site supervision end 2 shows oil truck in way status predication
State parameter.Whole system structure is as shown in Fig. 1.
Two, the design of detection node
Using largely detection node 1 based on wireless sensor network as oil truck in way state parameter perception terminal,
Multiple detection nodes 1 and on-site supervision end 2 are made up of the information interact system of wireless sensor network self-organizing.Detection
Node 1 includes acquisition oil truck surface temperature, pressure, the sensor of oil gas and the micro- place corresponding signal conditioning circuit, MSP430
Manage device and NRF2401 wireless transport modules;The software of detection node mainly realizes wireless communication and oil truck in way conditions Ambient
The acquisition and pretreatment of parameter.Software is designed using C programmer, and degree of compatibility is high, substantially increases Software for Design exploitation
Working efficiency enhances the reliability, readability and portability of program code.Detection node structure is as shown in Fig. 3.
Three, on-site supervision end software
On-site supervision end 2 is an industrial control computer, and on-site supervision end 2, which is mainly realized, joins oil truck in way state
Number is acquired and oil truck oil and gas leakage concentration intelligent forecast model, the information exchange of realization and detection node 1, on-site supervision
It is messaging parameter setting, data analysis and data management and oil truck oil and gas leakage concentration intelligent forecast model to hold 2 major functions.
The management software has selected Microsoft Visual++6.0 as developing instrument, and the Mscomm communication controls of calling system come
Communication program is designed, on-site supervision end software function is as shown in Fig. 4.Oil truck oil and gas leakage concentration intelligent forecast model designs
Following steps:
1, NARX neural network prediction models design
NARX neural network prediction models such as Fig. 2, NARX neural networks (Nonlinear Auto-Regression with
External input neural network) it is a kind of dynamic feedforward neural network, NARX neural networks, which are one, to be had
The nonlinear auto-companding network of oil truck oil and gas leakage concentration input, there are one the dynamic characteristics of multi-step delay for it, and by anti-
If the dried layer of feedback connection close network, NARX recurrent neural networks are most widely used a kind of dynamics in nonlinear dynamic system
Neural network, performance are generally better than total regression neural network.One typical NARX recurrent neural networks mainly by input layer,
It hidden layer, output layer and outputs and inputs delay and constitutes, the delay exponent number output and input, hidden to be generally determined in advance before application
Layer neuron number, the output at that time of NARX neural network prediction models depend not only on past output y (t-n), also depend on
In the delay exponent number etc. of input vector X (t) and input vector at that time.NARX neural network prediction model structures include input
Layer, output layer, hidden layer and time delay layer.Wherein oil truck oil and gas leakage concentration passes to hidden layer by time delay layer, and hidden layer is to input
Signal handled after be transmitted to output layer, hidden layer output signal is done linear weighted function and obtains final neural network by output layer
Predict that output signal, the signal that time delay layer exports the signal of network-feedback and input layer are then delivered to hidden layer into line delay.
NARX neural networks have the characteristics that non-linear mapping capability, good robustness and adaptivity, suitable for oil truck oil gas
Leakage concentration is predicted.X (t) indicates the external input of neural network, i.e., the oil truck oil and gas leakage concentration of multiple test points
Value;M indicates externally input delay exponent number;Y (t) is the output of neural network, i.e. the oil truck oil and gas leakage of subsequent period is dense
Spend predicted value;N is output delay exponent number;S is the number of hidden layer neuron;It is hereby achieved that j-th implicit unit is defeated
Go out for:
In above formula, wjiFor the connection weight between i-th of input and j-th of hidden neuron, bjIt is j-th of implicit nerve
The bias of member, the value of the output y (t+1) of network are:
Y (t+1)=f [y (t), y (t-1) ..., y (t-n), x (t), x (t-1) ..., x (t-m+1);W] (2)
2, cluster (FCM) classifier design of fuzzy C-mean algorithm
If finite aggregate X={ x1,x2,…xnIt is the set that n sample forms, they are each NARX neural networks respectively
The maximum value average value and minimum of prediction model, C are scheduled classification number, mi(i=1,2 ... c) it is each cluster
Center, μj(xi) it is degree of membership of i-th of sample about jth class, clustering criteria function is defined as by membership function:
In formula:||xi-mj| | it is xiTo mjBetween Euclidean distance;B is FUZZY WEIGHTED power exponent, is that can control cluster
As a result the parameter of fog-level;M is the Fuzzy C Matrix dividing of X, and V is the cluster centre set of X, and the result of FCM clustering algorithms is just
It is to obtain the M and V for so that criterion function is reached minimum.In fuzzy C-means clustering method, it is desired to person in servitude of the sample to each cluster
The sum of category degree is 1, i.e.,:
The minimum that formula (3) are asked under conditions of formula (4), enables J (M, V) to mjAnd μj(xi) partial derivative be 0, can obtain minimum
The necessary condition of value is:
FCM algorithms can be completed according to following iterative step:
A, clusters number c and parameter b is set, algorithm terminates threshold epsilon, iterations t=1, and permission greatest iteration number is
tmax;
B, each cluster centre m is initializedi;
C, membership function is calculated according to formula (4) with current cluster centre;
D, with current membership function all kinds of cluster centres are updated by formula (3);
E, suitable matrix norm is chosen, if | | V (t+1)-V (t) | |≤ε or t >=tmax, stop operation;Otherwise, t
=t+1, return to step C.
When algorithmic statement, all kinds of cluster centres and each NARX neural network prediction models output valve sample pair are obtained
In all kinds of degrees of membership, completes fuzzy clustering and divide.Fuzzy clustering result is finally subjected to de-fuzzy, fuzzy clustering is changed
Classify for certainty, realizes final NARX neural network prediction model output valves classification, as shown in Figure 2.
3, empirical mode decomposition modelling
Empirical mode decomposition (EMD) is a kind of self-adapting signal screening technique, have calculate it is simple, intuitive, be based on experience
With adaptive feature.The trend step-sizing that it can will be present in different characteristic in signal comes out, and obtains multiple high-frequency fluctuations
Partly (IMF) and low frequency trend part.EMD decomposite come IMF components contain signal from high to low different frequency sections at
Point, the frequency resolution that each frequency band includes changes with signal itself, has adaptive multiresolution analysis characteristic.It uses
The purpose that EMD is decomposed is exactly to more accurately extract fault message.IMF components must simultaneously meet two conditions:1. waiting for
In decomposed signal, the number of extreme point is equal with the number of zero crossing, or at most difference one;2. on any time, by office
The envelope mean value that portion's maximum and local minimum define is zero.Empirical mode decomposition method is directed to NARX neural network prediction moulds
" screening " process steps that type exports value signal are as follows:
(1) all Local Extremums of NARX neural network prediction models output value signal are determined, cubic spline is then used
Line connects the Local modulus maxima of left and right to form coenvelope line.
(2) local minizing point of NARX neural network prediction model output valves is being connected into shape with cubic spline line
At lower envelope line, upper and lower envelope should all data points of envelope.
(3) average value of upper and lower envelope is denoted as m1(t), it finds out:
x(t)-m1(t)=h1(t) (7)
X (t) is NARX neural network prediction model output valve original signals, if h1(t) it is an IMF, then h1(t)
It is exactly first IMF component of x (t).Remember c1(t)=h1k(t), then c1(t) first for signal x (t) meets IMF conditions
Component.
(4) by c1(t) it separates, obtains from x (t):
r1(t)=x (t)-c1(t) (8)
By r1(t) it is used as initial data to repeat step (1)-step (3), obtains the 2nd point for meeting IMF conditions of x (t)
Measure c2.Repetitive cycling n times obtain the n components for meeting IMF conditions of signal x (t).Pass through empirical mode decomposition model in this way
The output of NARX neural network prediction models is decomposed into low frequency trend part and multiple high-frequency fluctuation parts, experience decomposition model
As shown in Figure 2.
4, multigroup ANFIS Network Prediction Models design
Since fuzzy reasoning itself does not have self-learning function, application is greatly limited, and artificial neural network
Fuzzy language cannot be expressed again, actually similar to a black box, lacks transparency, so the reasoning of human brain cannot be expressed well
Function.Adaptive nuero-fuzzy inference system system ANFIS based on neural network, also referred to as Adaptive Neuro-fuzzy Inference
(Adaptive Neuro-Fuzzy Inference System), the two is organically combined, and can play the excellent of the two
Point, and respective deficiency can be made up.Fuzzy membership function and fuzzy rule in adaptive network-based fuzzy system are logical
It crosses and the study of a large amount of given datas is obtained, the feature of ANFIS maximums is namely based on the modeling method of data, rather than is based on
Experience or intuition are any given.The system that this is not understood those characteristics also completely by people or characteristic is extremely complex
It is particularly important.The input of ANFIS Network Prediction Models is respectively each multiple high-frequency fluctuation parts (IMF) and low frequency trend
Partial maximum value average value and minimum, steps are as follows for the major calculations of ANFIS Network Prediction Models:
1st layer:By the data obfuscation of input, each node corresponds to output and is represented by:
Patent of the present invention is 3 nodes, is the pole of each multiple high-frequency fluctuation parts (IMF) and low frequency trend part respectively
Big value, average value and minimum.Formula n is each input membership function number, and membership function uses Gauss member function.
2nd layer:Implementation rule operation exports the relevance grade of rule, and the regular operation of ANFIS Network Prediction Models, which uses, to be multiplied
Method.
3rd layer:The relevance grade of each rule is normalized:
4th layer:The transmission function of each node is linear function, indicates local linear model, each adaptive node i
Output is:
5th layer:The single node of this layer is a stationary nodes, and the output for calculating ANFIS Network Prediction Models is:
The consequent parameter of conditional parameter and inference rule that membership function shape is determined in ANFIS Network Prediction Models can be with
It is trained by learning process.Parameter is declined the algorithm adjustment combined with gradient using Linear least square estimation algorithm and joined
Number.ANFIS Network Prediction Models each time transmit along network forward direction until the 4th layer first in iteration by input signal, fixes at this time
Conditional parameter adjusts consequent parameter using least-squares estimation algorithm;Signal continue along network forward direction transmit until output layer (i.e.
5th layer).ANFIS Network Prediction Models by the error signal of acquisition along network backpropagation, with gradient method update condition parameter.
The conditional parameter given in ANFIS Network Prediction Models is adjusted by this method, can with it is concluded that parameter it is global most
Advantage not only can reduce the dimension of search space in gradient method, can also improve ANFIS Network Prediction Model parameters
Convergence rate.One group of ANFIS Network Prediction Model, which is realized, predicts a kind of NARX neural network prediction models output valve, one
Output low frequency part and multiple high-frequency fluctuation part of the input of group ANFIS Network Prediction Models for empirical mode decomposition model,
Each ANFIS Network Prediction Models output equal weight mutually sums it up to obtain one group of ANFIS network in one group of ANFIS Network Prediction Model
The fusion forecasting value of prediction model is as shown in Figure 6.
5, the ELman network oil truck oil and gas leakage concentration Fusion Models design based on PSO
ELman network oil truck oil and gas leakage concentration Fusion Models based on PSO are as shown in figure 5, Elman networks can be seen
Work is a feedforward neural network with local mnemon and local feedback link.Other than hidden layer, there are one especially
Associated layers.The layer receives feedback signal from hidden layer, and there are one corresponding association node layers to connect for each hidden node
It connects.Associated layers by the hidden layer state of last moment together with current time network inputs together as the input of hidden layer, be equivalent to
Feedback of status.The transmission function of hidden layer is generally Sigmoid functions, and output layer is linear function, and associated layers are also linear function.
In order to efficiently solve the approximation accuracy problem in fault diagnosis, enhances the effect of associated layers, propose a kind of ELman networks oil tank
Vehicle oil and gas leakage concentration Fusion Model is as shown in Figure 5.Z-1Indicate step delay operator.If the input layer of network, output layer, hidden layer
Number be respectively m, n and r;w1, w2, w3And w4Indicate structure layer unit to hidden layer, input layer to hidden layer, hidden layer to defeated respectively
Go out the connection weight matrix of layer, structure sheaf to output layer, then the expression formula of network is:
cp(k)=xp(k-1) (15)
If the ideal output of network is yt, reality output y, error function is:
ELman network oil truck oil and gas leakage concentration Fusion Models based on PSO are as shown in Figure 5.The model includes Elman
Neural network structure determination, PSO algorithm optimizations and Elman neural network prediction three parts.Basic step is as follows:
Step1, the M groups input of given neural network, output sample are standardized as training set, and by initial data;
Step2, Elman neural network structures are determined according to input, output parameter number, so that it is determined that PSO algorithm particles
Length;
Step3, by all interneuronal indirect weights, threshold coding in Elman network structures at real numerical code expression
Individual.If the N-dimensional vector that will be made of N number of weights, threshold parameter comprising N number of optimization weights, threshold value, each individual in network
To indicate initialization population;
Step4, to predict Error Absolute Value and as ideal adaptation angle value, and according to fitness value obtain individual extreme value and
Global extremum;
Step5, judge whether global extremum meets PSO termination conditions, if satisfied, exiting PSO optimizing, go to Step6;If
It is unsatisfactory for, updates each particle rapidity and position, go to Step4;
Particle corresponding to Step6, decoding global extremum, and in this, as the initial weight of Elman networks, threshold value;
Step7, the optimal initial weight and threshold value that are obtained in Step6 are assigned to Elman networks, training simultaneously determines network mould
Type predicts load with trained neural network model.
Four, oil truck is in the acquisition of way state parameter and intelligent predicting platform plane layout drawing
According to the structure of oil truck detection node, prison are placed in the upper surface, lower surface and front and back sides of oil truck oil tank
Surveying oil truck, whether leakage situation occurs for pot temperature, pressure and gas on the way in traveling.System arranges multiple detection nodes 1
With the horizontal layout installation diagram at on-site supervision end 2, wherein 1 equilibrium of detection node is arranged in the upper following table of detected oil truck oil tank
Face and front and back sides, whole system horizontal layout are shown in Fig. 7, wherein upper and lower surface layout layout with front and back sides it is similar, in figure with
Explained for layouting above, by the system realize to oil truck way parameter acquisition with to way oil and gas leakage concentration into
Row prediction.
The technical means disclosed in the embodiments of the present invention is not limited only to the technological means disclosed in the above embodiment, further includes
By the above technical characteristic arbitrarily the formed technical solution of combination.It should be pointed out that for those skilled in the art
For, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also considered as
Protection scope of the present invention.