Fig. 1 Structure flow of single
－
neuron PID controller
In Figure 3,
ω
1
,
ω
2
,
ω
3
is the weighted coefficient, which can be adjusted by the ability of the
neuron's self-learning. K is the neuron's adjustable ratio coefficient.
The single neuron adaptive
controller realized the adaptive and organizing function by adjusting the weighting coefficients. The
adjustment of weight coefficient can adopt different learning rules, and thus constitute different
control algorithms. In the experimental project of virtual laboratory, the advanced algorithms are
modular, and the sub VI is constructed to be called by other experimental modules. And the
realization of the single neuron PID controller will be the use of LabVIEW another way of calling.
LabVIEW in the dynamic link library call was achieved through the CLF (Library Function Call)
node. In the course of implementation of the single neuron
PID controller, single neuron PID control
algorithm was coded in the integrated environment of VC++, including set corresponding function
name, input and output parameters and return type etc.. Then the generated source code files are
compiled, and generate DLL files. So you can call these functions directly in LabVIEW. Then, the
ω
1
,
ω
2
,
ω
3
and adjustable ratio coefficient K are initialized in the main program flow. Figure 2 shows a
single neuron PID control algorithm flow chart. Figure 3 shows the VI single_n. flow chart.
Fig.2 Program flow of single-neuron PID
controller
Fig. 3 Program flow of single_n. vi
In the experiment, the controlled object is put into the implementation of step input. That is to set
the boiler level from 200 mm up to 250 mm. The experiment is repeated debugging.
The paper
selected the appropriate adjustable coefficient, learning rate and other control parameters: K=50.0,
Cp = 100.0, Ci = 0.1, Cd = 1.0, ts = 2 s. After the experiment, the level of the liquid level is stable. The
local end of the single neuron PID controller level response curve shown in Figure 4, we can see that
the system response fast, and no overshoot. At the same time, when the external disturbance is added
to the system, it is able to return to steady state quickly without oscillation. The experimental results
show that the single neuron PID controller in this paper has the ability
to be adaptive to the
time-varying and large time delay of the system, and it has better dynamic and static performance and
robustness.
21
Fig. 4 The local operation curve of single-neuron
PID controller
Fig. 5 The operation curve of single-neuron
PID controller of the remote-clien
At the same time, the system can achieve remote control experiments
on the Internet by virtual
laboratory remote client. Because the virtual lab system is based on the virtual experiment server
located at the center of the central server, which is responsible for the management and scheduling of
each experiment module and the field experiment instrument. On the one hand, the system receive the
Web server through the remote client's instructions, and call the relevant experimental module to
experiment and download
the experimental parameters; On the other hand, the system collected field
control end of the experimental data through the interface of OPC server DataSocket protocol, and
transmitted in real time to the remote client. The experimental results was shown in Figure 5. The
remote client operating curve and the local end is basically the same, the delay is less than 1s. It is
proved that the virtual network laboratory can achieve the basic goal of the network based remote
experiment.
