CybOnLine Manual - Personal WWW Pages
Contents
1. CybOnLine User Manual On line fuzzy modelling for process control applications PROCESS CYBERNETICS LIMITED Published and distributed by Process Cybernetics Limited Copyright 1998 Process Cybernetics Limited All Rights Reserved No part of this publication may be reproduced stored in a retrieval system or transmitted in any form or by any means electronic mechanical photographic recording or otherwise without the prior written permission of Process Cybernetics Ltd Disclaimer Process Cybernetics Limited makes no representation or warranties with respect to the programs or manual and specifically disclaim any expressed or implied warranties of merchantability or fitness for purpose In no event shall Process Cybernetics Ltd be held liable for direct indirect consequential or incidental damages resulting from any defect or omission in the manual or other related items or processes including but not limited to any interruption of service loss of business income anticipated profit use information or any other consequential damages Trademarks Microsoft and Windows are registered trademarks of the Microsoft Corporation a Table of Contents Introduction ccc ccc cece cc ccccee System Requirements 2 Installing CybOnLine 2 Verifying CybOnLine 2 Building Models for use in CybOnLine 4 Building Inferential Models EA2. the c program files cybonline directory double click on it 3 Inthe edit box in the run dialogue at the end of the command line add the name of the model file c program files cybonline tut pcg 4 Press the OK button and CybOnLine should start up 19 CybOnLine User Manual Chapter 3 The Outputs card CybOnLine FUZZMOD Lx tion Name CybOnLine DDE Topic Name FUZZMOD it Comms T riggers ji Inputs 1 Output Tagname DDE Item name Current Value Mass Outval O 0 The CybOnLine faceplate is divided into a number of cards The first of these cards gives details of the model outputs At the top of the faceplate the application name and topic name components of the DDE address are given These address components are common to all the variables in this copy of the program In this case the topic name is FUZZMOD the default name This is because we didn t specify a system name when building the model in CybMod This is not good practice and will give rise to problems if multiple copies of CybOnLine are run You should always specify unique system names for CybMod models which are to be run in CybOnLine On the Output card information is given on the 20 CybOnLine User Manual Chapter 3 model outputs included in this model In this case our model contains only a single sub model The first column gives the tag name Where dynamics are present in the model the maximum lag in
3. A Sos Gr E SR oak ae NA 4 Building What If models E fobs ya el oss Ap E E 7 Linking to a SCADA System Dynamic Data Exchange DDE 12 Triggering the calculation LARGE eRe Bead eho RT acs 14 Automatic triggering ea NR IE is oe E oN it he 14 External Triggering 15 Running CybOnLine 16 Adding CybOnLine to the Windows StartUp Polder 5 sc4cce see eae 16 Adding CybOnLine to the SCADA software s start up list 18 The CybOnLine Faceplate and Linking to the SCADA system 19 The Outputs card 22 4 pene nee e 20 Comms Triggers acres 54 c See es 22 Tipit oe dsn a ehh eth A ES 4 23 SCADA connections 24 Linking what if models Other Services from Process Cybernetics 29 Other products 4is ieiaue ree sakes 30 CybMod A fuzzy modelling program for process control 30 CybFIMC Model based fuzzy control 31 Services from Process Cybernetics 32 Trainings os Sed tor ace epee 32 Consultancy 04 32 Software services 04 33 Contacting US S242 edd be deere eds 34 Introduction Welcome to CybOnLine CybOnLine offers a flexible and easy to use way of integrating fuzzy modelling technology with your existing Windows based SCADA system CybOnLine can be used to provide inferred measurements where a difficult to measure quantity is inferred from other measurements or what i
4. am with the t switch on the command line Running CybOnLine CybOnLine must be running before your SCADA system starts to scan If it isn t running then the SCADA system will try to pick up DDE addresses in CybOnLine which don t exist and scanning errors will result There are two ways of ensuring that CybOnLine starts in time Adding CybOnLine to the Windows StartUp Folder Windows includes a special StartUp folder Programs contained in this folder will automatically be started immediately after the operating system starts Your SCADA system installation has probably put a start up component in this folder to automatically start your SCADA software To add CybOnLine to this start up group do the following 16 CybOnLine User Manual Chapter 3 1 Click on the start button and then point to Settings 2 Click Taskbar and then click the start menu programs tab 3 Click Add and then type the following in the command line edit box c program files cybonline t c program files cybonline model pcg The directory cybonline may be different if you have chosen to install CybOnLine to a different directory The t switch instructs CybOnLine to use external triggering If you want CybOnLine to automatically trigger calculations then simply omit the switch The model pcg is the model file that you wish to load into CybOnLine This model must have been previously created in CybMod and can be given
5. any file name you fancy 4 Click the Next button and then click the StartUp folder 5 Type the name you want to see appear in the Start up menu and then press the Finish button The name won t affect the programs operation and is really for your own information Multiple copies of CybOnLine can be run 17 CybOnLine User Manual Chapter 3 simultaneously Simply add more start up command lines to the StartUp folder remembering to use different model filenames for each Adding CybOnLine to the SCADA software s start up list Most SCADA packages include the facility of automatically starting additional software on SCADA start up The details of how to do this will be in your SCADA packages documentation Remember that you will need to be able to edit the command line for CybOnLine to allow the model file and trigger switch to be specified Again multiple copies of CybOnLine can be run simultaneously by including multiple start up command lines 18 CybOnLine User Manual Chapter 3 The CybOnLine Faceplate and Linking to the SCADA system In this section we shall look at CybOnLine loaded with the Gas tank model you produced in the last chapter Start CybOnLine with the gas tank model and automatic triggering by doing the following 1 Press the Windows Start button and then Run 2 Press the Browse button and look for the cybonline exe file When you find it it should be in
6. apter 3 page Create three trigger outputs and point them at the trigger variable addresses on the Comms Triggers page The type of output digital or analogue and the value transmitted is irrelevant the calculation will be triggered when anything is sent to these locations Add a sequence to your SCADA system which does the following a The operator presses a button to request a calculation b The current value of the measurement and trial input are sent to CybOnLine c Something is transmitted to all three triggers to start a calculation d Wait for five seconds this is more than enough time and in most cases can be reduced e Acquire the new value of the predicted process outputs from the three CybOnLine outputs CybOnLine User Manual Chapter 4 Other Services from Process Cybernetics We hope that you find CybOnLine a useful and easy to use piece of software All our software is written in house and we are continually developing our products We are very interested in receiving your feedback on our products If you have any comments to make or suggestions for improvements to our products then please get in touch with us at the address given at the end of this chapter The rest of this chapter outlines some of the other products and services available from Process Cybernetics 29 CybOnLine User Manual Chapter 4 Other products CybMod A fuzzy modelling program for process c
7. ariables blocks would have to be created and initialised An analogue output temperature pointing at App Name CybOnLine Topic FUZZMOD Item DDEInput 0 Sample time Same as that at which the data for identification was gathered An analogue output pressure pointing at App Name CybOnLine Topic FUZZMOD DDE Application Name CybOnLine DDE Topic Name FUZZMOD Outputs i Comms T riggers if Inputs 1 l Input Tag DDE Item Current Value Input Tag DDE Item Current Value Temperature DDEInput 0 Pressure DDEInput 1 Item DDElInput 0 24 CybOnLine User Manual Chapter 3 Sample time Same as that at which the data for identification was gathered An Analogue input Mass pointing at App Name CybOnLine Topic FUZZMOD Item Outval 0 Sample time Same as that at which the data for identification was gather but slightly offset from the sample intervals of the two inputs to allow time for processing 5 seconds is plenty A Digital input Communication error alarm pointing at App Name CybOnLine Topic FUZZMOD Item ErrFlg 0 Sample time Same as that for the other variables 25 CybOnLine User Manual Chapter 3 Linking what if models What if models will normally be executed at the request of the operator which means that CybOnLine has to be set up using the external triggering option Using the c program files cybonline tut2 pcg the model f
8. e of 1 means that a problem exists A DDE problem will fail the sub model on which it occurs and stop 22 CybOnLine User Manual Chapter 3 further calculations pending a reset The error flag item address should be connected to a SCADA digital input and tied into an alarm or error warning The Comm Status column simply gives a visual indication of the current status of the sub models links A communication failure will be shown by a failure message on a red background Good links are indicated by OK on a green background The Reset button is used to reset individual sub models after a communication failure With automatic triggering it has to used in between samples to avoid sequencing problems Inputs Click the Inputs 1 tab There will always one or more cards associated with the model inputs These cards give the variable name the DDE item name assigned to it by CybOnLine and the variables current value A particular variable can be used as an input for several sub models CybOnLine only allocates a single DDE item name for each variable regardless of the number of times it is used in different sub models If a variable is used several times and automatic triggering is in use then all the variables in the affected sub models need to have the same sampling time to avoid sequencing 23 CybOnLine User Manual Chapter 3 problems SCADA connections For the model in this example the following SCADA v
9. f modelling to support process operators CybOnLine uses models which have been developed in CybMod Process Cybernetics fuzzy modelling package The models are loaded at the command line as CybOnLine starts CybOnLine then acts as a DDE server allowing DDE capable SCADA software to link to it simply by setting a few I O addresses in the SCADA package CybOnLine runs in the background carrying out the model calculations and passing the information to the SCADA package allowing you to easily integrate the fuzzy modelling results into your control schemes 1 CybOnLine User Manual Chapter 1 We have developed CybOnLine from leading edge research in fuzzy modelling and control in the process industries Process engineers have designed the package with the process industries in mind We are continually developing our products and would be delighted to receive your feedback System Requirements Pentium PC running Microsoft Windows 95 98 or NT A minimum of 16 MB of RAM is recommended Installing CybOnLine Insert the diskette number 1 into the floppy disc drive Choose the Run command from the Start menu and in the Run window type a setup exe and then press enter Follow the on screen instructions to complete the installation Verifying CybOnLine 2 CybOnLine User Manual Chapter 1 The first time you run CybOnLine you will be taken to alicence verification dialogue Simply enter your password and press t
10. he OK button then follow the on screen instructions Each copy of CybOnLine is licensed to run on only a single machine The number of installations is strictly limited by the protection software and these installations are only available to help you in the event of a hardware problem If you exceed the number of permitted installations you should contact Process Cybernetics Process Cybernetics reserves the right to charge for or refuse to supply additional registrations 3 CybOnLine User Manual Chapter 2 Building Models for use in CybOnLine The two main uses for CybOnLine are for inferential and what if modelling Although the models in both cases will be built using our modelling package CybMod there are slight differences in the model building process Building Inferential Models An inferential model is used to estimate the value of a difficult to measure variable from the values of other easy to measure variables The most commonly inferred variables are compositions e g estimating the mole fraction on a distillation column plate from temperature and pressure data 4 CybOnLine User Manual Chapter 2 A feature that is common with most inferential systems is that they don t involve dynamics The measurements used to infer the difficult value are related to it in a purely functional way that involves neither differential equations nor time delays The data used for inferential modelling ca
11. his way the effects of a particular change can be seen as a series of snapshots at different times As a tutorial let s build a set of what if models using method 2 The data we shall be using in this example what if on the last disk in the CybOnLine distribution set was generated by from a simple second order process supplied with two inputs Both inputs are random but one changes four times as frequently and more powerfully than the other The data is arranged in rows of samples and three fields of data The first data field in the frequently 8 CybOnLine User Manual Chapter 2 changing input the second the other input and the third is the output Load the data into CybMod Let s assume that we want to build a what if model based on the first input The first thing we have to do is to construct a standard dynamic model to find the minimum lags which can be applied to the process inputs Using cross correlation of the raw data and residuals and trial and error we found the best model of the data in first order dead time format to be Output k R Output k amp 1 Inputl k amp 4 model built with two reference sets per variable and tested using a free running predictor The model above is a standard dynamic model To convert it into a what if model we need to make a few changes In operation we will know the current value of the output and will know the value of the input we want to tr
12. ile you saved earlier as an example the steps in setting up the what if model are 1 Run CybOnLine using the external trigger option e g cybonline t tut2 pcg 2 Create two analogue outputs in your SCADA system one for the current value of the process output and one for the test value of the input and connect them to the appropriate DDE addresses in CybMod The sampling time can be anything you want and can be a mixture of sampling and exception processing if this is available on your SCADA system 3 Create three analogue inputs and connect them to the three model output DDE addresses 26 CybOnLine User Manual Chapter 3 DDE Application Name CYBONLINE DDE Topic Name FUZZMOD if Comms T riggers if Inputs 1 DDE Item name Current Value Output 4 Outval O 0 Output 8 Outval 1 0 Output 16 Outval 2 0 The three outputs are the outputs of each of the three sub models you created in CybMod Output 4 refers to a model where the maximum lag is 4 in this case Output k R Output k amp A Input k amp 4 4 Create three digital inputs and point them to CybOnLine FUZZMOD Lx DDE Application Name CYBONLINE DDE Topic Name FUZZMOD Output Tagname Comm Error Flag Comm Status Trigger Item Output 4 EnFla Q OK Trig dd 0 Output 8 EnFlg 1 0K Trig dd 1 Output 16 EnFla 2 OK TrigAdd 2 the error flag variables on the Comms Triggers 27 CybOnLine User Manual 28 Ch
13. n come from two sources The required data may sometimes be available in tables of data e g boiling point data and in this case the fuzzy model simply serves as a convenient way to transform this tabular data to a continuous form More commonly the data is available from historical plant data Difficult to measure quantities are often obtained from routine sampling and analysis Historical records of these analytical results can be merged with the recordings of the other data and used to provide a data set for modelling purposes To demonstrate the procedure involved in generating an inferential model a sample data set is included with this package You can find the data in the last disk in the CybOnLine distribution set under the filename press dat The data consists of several samples each of three fields of data which represent respectively the temperature the pressure and the mass of gas contained within a vessel The data were generated from an ideal gas type equation i e Mass 2 646 10 _P essure_ Temperature 5 CybOnLine User Manual Chapter 2 This data was generated purely for tutorial purposes In areal situation if you had an equation of this sort it would of course be much more sensible to include it directly in the SCADA systems rather than building a fuzzy model For the tutorial let s build a model which infers the mass of gas in the vessel from the temperature and pressure First of all load the data int
14. o CyobMod see the CybMod manual for instructions Change the field names as follows Field_0 to Temperature Field_1 to Pressure and Field_2 to Mass Unlike data from a dynamic system the samples in this data file are in no particular sequence The samples are completely unrelated to time and this means that the cross correlation results are mostly meaningless The correlation coefficient at zero lag is useful however as it represents the correlation between the various fields in the data with no dynamic lag between them Create a model by selecting Mass as the output and pressing the Create button on the main window dialogue A card containing the new model will appear Note that the model currently has a single input Mass at a lag of one We don t want to create a dynamic model so the first thing we need to do is change this lagged output into a model input that we do want Double click on the lagged input change the variable to Temperature 6 CybOnLine User Manual Chapter 2 and change the lag to zero Now add the other input we want on the model the Pressure at zero lag Identify the model using sequential presentation and all the data Save the model as c program files cybonline tut pcg for use in the tutorial in the next chapter Care has to be taken when interpreting the results of the model analysis tools which have been primarily designed for dynamic modelling The predic
15. ontrol CybMod is the core of our current range of products It generates relational fuzzy models from process input output data and saves themina form suitable for use in CybOnLine and CybFIMC CybMod includes tools to analyse and adjust the raw I O data and model construction and identification is quick and easy CybMod also includes a range of tools to evaluate the quality of the models and to diagnose any problems 30 CybOnLine User Manual Chapter 4 CybFIMC Model based fuzzy control CybFIMC adds fuzzy model based control capability to your existing Windows based SCADA system It takes a model generated in CybMod and incorporates it into a Fuzzy Internal Model Control FIMC scheme This provides full non linear feedback control with dead time compensation and feed forward action on any included disturbances CybFIMC like CybOnLine is a DDE server making it really easy to link to existing SCADA installations CybFIMC is a MISO Multi input Single output controller it controls a single output variable which can be subjected to several inputs Multiple copies of CybFIMC can be run simultaneously providing as many loops as you require 31 CybOnLine User Manual Chapter 4 Services from Process Cybernetics Training Process Cybernetics offers courses on various aspects of fuzzy modelling and control These courses include a mixture of both theory and practise If you would like to be kept informed of u
16. pcoming courses then tick the boxes on your registration card or contact us directly at the address given at the end of this chapter Consultancy Process Cybernetics can provide consultancy to help support your modelling and control projects The level of our consultancy involvement can range from telephone support to full project management We shall provide a fixed price quotation for any work we undertake If you are interested in our consultancy services then contact us at the address given at the end of the chapter 32 CybOnLine User Manual Chapter 4 Software services All our software is written in house and as a result we are able to offer a bespoke tailoring service to make our software suit your particular requirements This could involve changing the look of the software changing the way it operates or integrating it with other applications If you have an application in mind that requires special software then contact us at the address at the end of the chapter We will discuss your requirements and provide you with a fixed price quotation for the work 33 CybOnLine User Manual 34 Chapter 4 CybOnLine User Manual
17. same as the system name specified in CybMod or FUZZMOD if no name was specified The item name is the name which connects to a particular piece of data CybOnLine generates item addresses automatically and lists them against the data field tag names on the CybOnLine faceplate Most SCADA systems will allow DDE servers to act 13 CybOnLine User Manual Chapter 3 as data I O sources for a wide variety of input and output variable types eg analogue outputs analogue inputs digital outputs and digital inputs The usual way to set up a link is to specify DDE as the I O device and then to include the DDE address as a device I O address The method of forming the individual components of the DDE address application topic and item into a single I O address varies between SCADA systems consult your SCADA documentation to find out how to do this Triggering the calculation CybOnLine provides you with a choice of ways to trigger a calculation Automatic triggering In this mode CybOnLine keeps track of input information which has been passed from the SCADA package A model calculation is triggered when a complete set of input data has been received Once a calculation has been made all of the input counters are set to zero and a new and complete set of input data has to be sent to trigger a new calculation When operating in automatic mode CybOnLine will generate an error if data arrives out of sequence i e if
18. simple to integrate with your existing SCADA package This chapter provides you with the general approach which can be used but the details will be dependent on your particular SCADA software Dynamic Data Exchange DDE Dynamic Data Exchange is a method available in the Microsoft Windows operating systems for exchanging data between programs DDE works by using an area of shared memory It is a bit like using the clipboard to transfer information between programs but occurs completely under program control without user intervention 12 CybOnLine User Manual Chapter 3 CybOnLine is a DDE server which means that it acts primarily as an information source To connect to CybOnLine your SCADA software must be able to act as a DDE client Consult your SCADA documentation to confirm this The key to allowing data to be exchanged through DDE is the DDE address of the information to be connected to The DDE address is made up to three components the application name the topic name and the item name The application name is the name of the server which is to be used to supply or receive the data This will always be CYBONLINE for all cases where the SCADA system and CybOnLine are running on the same machine it is not recommended to run the two pieces of software on different machines over a network The topic name is generally used to indicate a particular group of data In CybOnLine the topic name is set to be the
19. the same input data is received twice without 14 CybOnLine User Manual Chapter 3 an output calculation in between This can be useful for warning the user of problems with the DDE connections either missing or erroneous connections Automatic triggering is useful when all the input data is gathered and modelling calculations are to be carried out at the same sample interval Automatic triggering is the default mode of operation for CybOnLine External Triggering In this mode CybMod only carries out a calculation when it receives an input from a trigger variable The trigger variable can be an analogue or digital variable and the value that is sent to CybMod is irrelevant It is simply the receipt of a value into the trigger address which causes the calculation to be executed When operating with an external trigger CybMod no longer keeps a track of which inputs have been received it is the responsibility of the SCADA system to ensure that a complete set has been sent External triggering can be used in situations where a calculation has to be carried out on demand rather than at particular intervals e g for operator decision support It is also useful where model 15 CybOnLine User Manual Chapter 3 input data is sampled at different rates or where a model output is needed less frequently than the sampling rate of the input data External triggering is enabled in CybOnLine by starting the progr
20. the sub model is given in brackets after the output tagname this is mainly of use in working with what if models where the same output may be modelled in several different sub models In this case we have a steady state model no lags and nothing is shown after the tag name The next column gives the DDE item name to be used for linking this variable with the SCADA system This item name along with the application and topic name uniquely identifies this particular output The last column on the output card gives the current value of the model outputs No calculations have yet been carried out and so this value is blank 21 CybOnLine User Manual Chapter 3 Comms Triggers Click the Comms Triggers tab DDE Application Name CybOnLine DDE Topic Name FUZZMOD gers if Inputs 1 Comm Status Outputs Output Tagname Comm Mass ErFlg 0 OK Reset This card gives information on the status of the DDE communications for each sub model When external triggering is selected it will also show details of the trigger address for each sub model The first column on this card the Output Tagname simply identifies the particular sub model The next column the Comm Error Flag gives the item name for a digital output which indicates the current communication status of this sub model A value of 0 indicates that no problems have been encountered in DDE communications and a valu
21. tion plot will be fine for point to point comparisons but the shape of the curve is completely meaningless the data is in no particular sequence The residuals plot will be useful actually more so than with dynamic models but the residual cross correlations will be useless When testing the model be sure to use the Free running predictor Using the other predictors will give misleading results Building What If models Sometimes it is useful for operators to have models which allow them to forecast the results of particular actions These what if models are often used as a half way house to fully automatic fuzzy control and allow confidence to be built up amongst the process operators It should be noted however that the control results from what if supported manual control will be inferior to that 7 CybOnLine User Manual Chapter 2 which could be achieved by fully automatic control Usually it is necessary to obtain predictions of the output over a horizon rather than at a single point This can be achieved in CybMod in two ways 1 The SCADA system can be used to run CybMod as a multi step ahead predictor This involves the configuring the SCADA system to repeatedly call CybMod using previous predictions as inputs for future predictions This is rather complicated to do and is not recommended 2 Multiple models up to nine of the process can be created each at a different prediction lag In t
22. y in other words we know Output k and Input1 k and want to predict Output k From the dynamic model we know that the minimum lag that we can apply to Input for the best prediction results is 4 This means that our pase what if model is 9 CybOnLine User Manual Chapter 2 Output kYA R Output k Input k which can be expressed in backward difference form as Output k R Output k amp 4 Input k amp 4 We can create a whole series of what if models with lags greater than 4 which will allow us to predict more than 4 samples into the future As we increase our prediction range however the accuracy of the what if models will fall This is because one or both inputs can change within a long prediction period Construct the following three sub models using the data Output k R Output k amp 4 Input k amp 4 Output k R Output k88 Input k amp 8 Output k R Output k amp 16 Inputl k amp 16 Test the models using a single step ahead predictor which is the mode in which these what if models will be used You will see that the prediction accuracy falls off with increasing prediction range at long ranges the prediction may be useless 10 CybOnLine User Manual Save the model as Chapter 2 c program files cybonline tut2 pcg for use in the next chapter CybOnLine User Manual Chapter 3 Linking to a SCADA System CybOnLine is designed to be
Download Pdf Manuals
Related Search