here - M-D-J

sciencediscussionAI and Robotics

Oct 20, 2013 (3 years and 7 months ago)

63 views

User‘s manual


PHP modul



This modul was developed for neural networks monitor

based on the method Optifnet or Z
-
optifnet.

It consists from next parts:

-

NN Monitor configuration

-

Controls

-

NN parameters

-

Limits of NN modul

-

Banks usage

-

Status window

Minutely

will be explain later.


ATTENTION:

every adjustable items are located in configuration file and it is possible to
change it for your program in PLC(it means your option in which item will be inputs, outputs
etc).
Config



here is variable from configurati
on file which you must change to change item
from which the value is loading. Change only items by ENABLE FOR USER CHANGE.
Don’t change items by DISABLE FOR USER CHANGE in any case!!!!!. Configuration file
path is: config/config.php.



Connection to Irul
an server


There are 3 items for connection to Irulan server in configuration file:

$irulan_cfg["host"]


DNS(computer name or domain) or IP address of computer where the
Irulan server is located

$irulan_cfg["port"]


port, on which is Irulan server listen
ing

$irulan_cfg["timeout"]


timeout set


if program is not able to make connection with Irulan
server in this time, it will be terminated with error message




Part NN Monitor configuration


ON

-

starts neural network learning

config
: $irulan_cfg["on_o
ff"]


set to 1


OFF

-

stops neural network learning


config
: $irulan_cfg["on_off"]


set to 0


RESET

-

resets NN modul


stops(OFF), resets initialize bit, resets weights


config
: $irulan_cfg["on_off"]


set to 0



$irulan_cfg["reset"]


initia
lize bit
-

set to 0



$irulan_cfg["weights"]
-

by number of inputs and outputs resets the count of
weights;

count = inputs * outputs



Method



you can choose one of the method Optifnet or Z
-
optifnet when you click on radio
button. Optifnet me
thod is set as default


Weights



Status window

will show matrix of weights when you click on this button.
Reloads every 20 seconds.


config
: $irulan_cfg["weights"]


weights are items started with this address


Passed



graph shows how many samples are al
ready learned so how many „passed“



config
: $irulan_cfg["l/c"]


there is a number of not learned samples in this item


Picture


LEARNING

or
COMPLETED



picture shows neural network learning status

LEARNING


neural network is learning yet;
COMPLETED


neural network is learned
already



Weights graph

-

Status window

will show graph. It is possible to see how is the neural
network learning on this graph


number of graphs is computing by inputs and outputs




number of graphs = 2 * outputs.


There is a

legend under the weights graph which describes each line. Data for graph are saved
in the buffer


so it’s important to load data to modul, which automaticly save data to buffer;
or download data from modul before first load of the graph (see
Upload
), (se
e
Download
from NN modul to buffer
). Reloads every 20 seconds.


Passed



see
Passed

by
Weights



Matrix

-

Status window

will show matrix of dynamical system: A, B, C(matrix D will not
shown because D = 0). It’s important to choose regular method(click on r
egular radio button)
for correct display of the matrix. Reloads every 20 seconds.


Passed



see
Passed

by
Weights


Value of step response



height of step response

Number of steps



number of steps of step response

Step response

-

Status window

will show p
icture with step response of dynamical system
computed by filled height and number of steps, when you click on this button. It’s important
to choose regular method(click on regular radio button) for correct display of step response.

config
: $irulan_cfg["st
eps_0"]


how many steps will be value of step input zero


Passed



see
Passed

by
Weights


Get actual status of NN modul


the whole application will be loaded again with actual values
when you click on this link


Picture

-

ON, OFF or Reseted


indicates

actual status of NN modul. It will change if you
click on some of these buttons: ON, OFF or RESET







Part Controls


Upload this file to NN Modul



you can choose file which you want upload to NN modul
when you click on
Browse

button


but it’s importan
t to set number of inputs and outputs on
the base of your data file(see
Inputs
,
Outputs



Set to modul
); number of active banks
program sets automaticly and learning step gama is not refer with this part of program.
Required data will be load to NN modul w
hen you click on
Upload
button and relevant items
will be set.


Upload one of these demo files instead your's
: there are 3 files prepared like demo, 1 for
Optifnet and 2 for Z
-
optifnet method. It’s possible to upload one of these 3 files instead yours.
Bu
t it’s important to set inputs and outputs by legend and limits (see

NN Parameters a
Limits of NN modul
). ATTENTION: if you choose one of demo files, leave form by
Browse

button blank.


If you have checked
Reset synaps after upload
, program will resets we
ights moreover.
Progress of upload is possible to see in
Status window
.



config
: $irulan_cfg["values_separator"]


separator of values in file


$irulan_cfg["bank0"]


start address of the first bank


$irulan_cfg
["bank1"]
-

start address of the second bank


$irulan_cfg["bank2"]
-

start address of the third bank


$irulan_cfg["bank3"]
-

start address of the fourth bank



$irulan_cfg["bank_count"]


start address

for usage of each bank


Download from NN modul to buffer


data from NN modul will be save to buffer when you
click on this button. Buffer is important for
Weights graph
drawing
.
So if other user already
loads data to NN modul, you don
’t need to
load it a
gain, but for the correct show of the
Weights graph

you must save data from NN modul to buffer. Progress of download is
possible to see in
Status window
.


Save data from NN modul to file

Name of file
: fill this form with name of file, in which you want to
save data from NN modul.
Data will be save after you click on this button. Progress of download is possible to see in
Status window
. File is after download located in:
path_to_application/data/

like
name.dpl



Part NN parameters:

Inputs



inputs


config
:

$irulan_cfg["inputs"]


Outputs



outputs


config
: $irulan_cfg["outputs"]


Active banks
: active banks


config
: $irulan_cfg["banks"]



Gama
: gama


learning step


this number speed up or slow up learning of neural network


config
: $irulan_cfg["gama"]


Set
to modul


you can set up number of inputs, outputs, active banks and learning step gama
to your values with which you fill the relevant forms when you click on this button. Program
change only that items which value was changed.


Get from modul



program

loads the following values: inputs, outputs, active banks, gama,
limits and graphical representation of usage of each bank when you click on this button (see
Part Limits of NN modul and
Banks usage
).



Part Limits of NN modul:


Set limits to NN modul
-


y
ou can set upper and lower limits to your required values here
when you fill relevant form with values and click on the button. Attention: set up only upper
limits. Program will automaticly set up lower limits to values:



lower limit = upper limit * (
-
1)
.


Limits will be save to items on the base of the configuration file.


config
: $irulan_cfg["lim"]


start address of limits


their count is set by number of
outputs: for each outputs is one pair of limit(upper and lower limit)


Upper and

Lower limits



there are values of upper and lower limits in these forms


config
: $irulan_cfg["lim"]



Part Banks usage:


Graphs shows percentual usage of each bank. These values are changed after you upload new
data to NN modul and will refresh after you click on
Get
from modul

or
Get actual status of
NN modul


config
: $irulan_cfg["bank_count"]


start address for usage of each bank




Order


It’s important to set up every items in configuration file like your PLC settings. It’s important
to set up connection to Irulan

server too.


Configuration of NN modul:

The best way how to start is to reset NN modul with
RESET
button. After that is
necessary to set up number of inputs, outputs on the base of your data file or demo file which
you want to upload to NN modul. The numb
er of inputs and outputs for demo files is show by
radio button. You make this action when you change the relevant values and click on the
Set
to Modul
button. It’s important to set up limits then. Set up only upper limits, lower limits will
be changed aut
omaticly. You make this when you change the values in relevant forms and
click on
Set limits to NN modul
button
.


When you have finished the configuration, you can load the actual values when you click on
link
Get actual status of NN modul
, which reloads
the whole window.



File upload:


Now we can upload file to NN modul. You can choose your file or one of demo files
for upload. If you don’t reset the NN modul on the start, you can check the
Reset synaps after
upload

and program for upload will reset the
weights too. Program fill the buffer neccesary
for
Weights graph
drawing too
.


Start running of NN modul:


If we have correctly configured the NN modul and loaded data, we can start running of
NN modul


it means start learning of neural network with the
O
N
button
.

We can stop
learning of neural network with the
OFF
button anytime.


Monitoring :


Now we can monitor how the neural network is learning. There are 4 possibilities
what we can monitor:


1, weights matrix


Weights

button


2, weights graph


Weiht
s graph
button


3, matrices of dynamical system


Matrix
button


4, step response


Step response
button after we fill the forms
Value of step response

and

Number of steps

ATTENTION: if data are loaded in NN modul already,
for the correct show of the
Weigh
ts
graph

you must save data from NN modul to buffer



Download from NN modul to buffer
button

ATTENTION: it’s important to choose a regular method(click on regular radio button) for
correct show of the matrix of dynamical system and step response.

It’s pos
sible to see how many samples are learned already on the
Passed

graph and
status of learning


it indicates picture LEARNING or picture COMPLETED. Also is
possible to change learning step gama and limits during the monitoring.


Learning result :


Weights
matrix, matrices of dynamical system A, B, C and step response computed by
the end of learning of neural network


COMPLETED status.


Save data to file:


If data are loaded in NN modul already, you can
download them to file with name you
want. This file yo
u can use for next using this application or for another program
.