Verification of Railway Interlocking Tables using Coloured Petri Nets

1
Verification of Railway Interlocking Tables
using Coloured Petri Nets

• Somsak Vanit-Anunchai
• somsav_at_sut.ac.th
• School of Telecommunication Engineering
• Suranaree University of Technology
• Nakhon Ratchasima 30000 Thailand
• Supported by National Research Council of
  Thailand

2
Introduction to railway signalling
24
41T
24T
23
103T
16
9T
3T
1T
16T
103
42T
15
1
3
Railway Signalling System divides rail track
into sections. Only one train is allowed in one
section at atime. A section or route comprises
wayside equipment 1) Track Circuits used to
indicate the presence of trains 2) Signals to
allow the train enter into the route. 3) points
(switches ) to diverge the train to another
track. Each wayside equipment has an
Identification number.
3
A typical (small) station
24
41T
24T
23
103T
16
9T
3T
1T
16T
103
42T
1
15
3
Route Released ? Normal
Route 3(2) locked
Require TC
Interlocking Tables or Control Tables are the
tabular representation specifying how the train
move together with the states and actions of
related equipment.
4
Approach Lock Cannot cancel
Signalman can cancel
24
41T
24T
23
103T
16
9T
3T
1T
16T
103
42T
15
3
1
5
Motivation (Problems)

• Problems with manual inspection of railway
  Interlocking table ? labour intensive, erorr
  prone
• State Railway of Thailands projects involves
  300-350 stations
• Existing track layout changed (added)
• ? existing signalling changed.
• Other software tools usually are designed for a
  specific railway company but SRTs Operating rule
  is unique and sometimes changed.
• Need simple formal methods for signal engineers

6
Selected related work (quick look)
Logistic
7
Selected related work
8
Selected related work
9
Our CPN model of the Control Table of the small
station comprises two parts

• Signalling Layout
• Interlocking
• The CPN model comprises
• 72 Places ,
• 12 Fusion places ,
• 21 Substitution Transitions,
• 33 Transitions and 12 ML functions.

10
CPN model of the Control Table

• Signalling Layout
• - The CPN model mimics the signalling plan
• - Provides geographic information how each
  wayside equipment connect to each other
• - Provides ability to simulate the trains
  moving
• - Comprises lower CPN subpages which represent
  the trains movement when passing signals,
  passing point and moving between 2 consecutive
  track circuits
• ? modelling wayside equipments

11


CPN model The southern part of the station
12
Modelling Approach

• The CPN model in the signalling layout part
  depends on the track layout.
• ? It is inevitable.
• But the CPN diagram can be quickly, manually
  built when we have CPN patterns (library).
• The work on CPN patterns for this project is in
  progress .

13
CPN model of the Control Table

• 2. Interlocking part comprises 3 CPN subpages
• 2.1 UserCommand
• ? sets and locks the points along the route
• 2.2 Routesetting
• ? sets the required route
• 2.3 RouteReleased
• ? using the passage of the train restores the
  route to Normal state and unlocks the points

14
Modelling Approach

• The CPN model in the Interlocking part depends
  on the contents in the control table.
• Because of 300 stations (to go), we attempt to
  make the generic net structure.
• The contents of the control table are coded in
  ML functions used in arc inscriptions.
• ? Thus 300 stations can use the same net
  structure of the Interlocking part.

15
CPN Model Route Setting
require_point_normal(route) require_point_rever
se(route)
16
Excel ? XML
XSLT script
It took me 2- man-months to complete the first
model (including analysis). But the double track
station? It took me only 8-man hours to build
the model (not including analysis).
? ML functions are automatically created from XML
control table using XSLT.
17
CPN Model Route Setting
This part is a great help regardless of
assumptions.
require_point_normal(route) require_point_rever
se(route)
18
Assumptions and their affects on the correctness
of the model

• To start building the model we have 10
  assumptions.
• Q The important question is how these
  assumptions affect the model.
• A I consider that there are some differences
  between the real system and the model. However
  the model in this paper can detect a large part
  of errors which we always encounter.
• A larger part something is missing or added
  (extra) in the Control Table.

19
Analysis

• The desired property is no collision.
• No two train in two consecutive track circuits.
• Using ML query functions.
• To convince the model correctness
• After route(s) setting and train(s) movement ,
• The terminal markings shall be as we expect.
• To debug the model using an incremental approach
  . Starting from one route setting - one train

20
Terminal markings
Using query ML and state space search No train
collision is detected in case A,B and C
21
Conclusion

• A control table for the small and typical single
  line railway station is modelled and analysed.
• This CPN model can be adapted and re-used for
  SRTs double track projects (300-350 stations) .
• We propose to convert Control tables to ML
  functions using XSLT.
• Thus the CPN models of other interlocking can be
  rapidly built.
• These models will help to detect errors in
  control tables in the early phase of system
  development.

22
Future work

• Relaxes modelling assumptions
• Revises the CPN subpages and arranges a library
  of CPN patterns
• Create CPN models directly from Track layout
  drawing.

23

• Thank You!
• Questions and comments?

24
(No Transcript)
25
Initial markings

• - noTrain at other places
• setting commands for all 8 routes
• - Both blocks in Coming states
• A Block request command for going toward Bangkok

26
Analysis results
State space sizes
More trains ? less number of possible train
movements Less trains ? more number of possible
train movements Not true in general (e.g. double
track and large stations)