Evaluating chaff fire pattern algorithms in a simulation environment JP du Plessis (jdp@imt.co.za) Institute for Maritime Technology South Africa
Overview What is seduction chaff? Chaff solution algorithm SADM External C2 Batched parameter setup Batch results DSADM cluster setup Hardware performance Cluster start-up
What is seduction chaff? A round, releasing fine pieces of foil to reflect radar waves, fired some distance from a ship. Chaff aims to present to a missile seeker a more valid target than the ship itself. Seduction chaff is used to defend against an incoming missile threat already locked onto own ship. Wind relative to ship takes the chaff through the missile's range gate, taking the gate with it. In successful cases, the threat missile's path is altered such that it misses the ship by some miss distance.
Chaff solution algorithm A chaff solution algorithm is part of an ship EW system that calculates an optimised chaff solution for the current situation. Seduction chaff requires a detailed algorithm or lookup table to calculate a solution which is not configurable with parameters alone. In a missile threat situation, this algorithm is automatically used to find a solution based on the current scenario.
Chaff solution algorithm (2) Main input variables: Ship velocity Wind velocity Threat bearing Threat type Main output variables: Number of chaff rounds For each round: Launcher Azimuth Launcher Elevation Time to fire Fuse time
SADM external ship C2 model SADM: Ship Air Defence Model BAE Systems A ship C2 model can be configured by the SADM user to represent unique characteristics of the relevant ship Command and Control system. A great feature of the external C2 model is it gives the SADM user the ability to integrate seduction chaff deployment algorithms. source code - code as compiled for own naval vessel closed format -.dll from chaff system supplier
SADM external C2 model (2) The user has some freedom to narrow down the C2 model to only cater for ship systems of interest You do not need to develop a full blown ship C2 model with all the bell and whistles. It all depends on what needs to be achieved.
Algorithm simulation result
We now know what the typical result is for one scenario. One would naturally want to know: What if the wind velocity was different? What if the threat missile came from another direction? Where does my chaff algorithm fail?
Batched parameter simulation runs Do multiple simulation runs varying one variable at a time between subsequent runs. Parse the log files for each trial run to gather needed information from SADM results (file formats not documented and changed between releases). Dump data into.csv file for easy import into preconfigured spreadsheet. Analyse the combined results.
Batch parameter configuration Algorithm need to be tested with a range of different input parameters and the results analysed. Parameters varied: 4 wind velocities 18 wind directions (0 to 340 in 20 steps) 18 threat bearings (0 to 340 in 20 steps) Number of scenario variations: Number of simulation runs (10 Monte-Carlo runs per scenario): 4 x 18 x 18 = 1 296 1 296 * 10 = 12 960 runs The Ship Air Defence Model have this process automated and the results can be analysed directly in the model or the output data can be analysed directly.
0 340 20 320 40 300 60 280 80 Threat angle Wind direction 260 100 240 120 220 140 200 160 180
Algorithm 1
Miss distance plot - 1 Launcher, 5 chaff rounds Ship heading = 000 deg, speed = 0 m/s 500 400 300 200 100 50 360 340 320 ASM angle relative to ship bow, direction from [deg] 300 280 260 240 Wind 30m/s ( 60kts) Wind 20m/s ( 40kts) Wind 10m/s ( 20kts) Wind 0 m/s (0 kts) Size reference [m] - 220 200 180 160 140 120 100 80 60 40 20 0 True wind angle relative to ship bow, direction from [deg]
Algorithm 2
Miss distance plot - 2 Launchers, 5 chaff rounds each Ship heading = 000 deg, speed = 0 m/s 500 400 300 200 100 50 360 340 320 ASM angle relative to ship bow, direction from [deg] 300 280 260 240 Wind 30m/s ( 60kts) Wind 20m/s ( 40kts) Wind 10m/s ( 20kts) Wind 0 m/s (0 kts) Size reference [m] - 220 200 180 160 140 120 100 80 60 40 20 0 True wind angle relative to ship bow, direction from [deg]
Algorithm 3
Miss distance plot - 4 Launchers, 5 chaff rounds each Ship heading = 000 deg, speed = 0 m/s 500 400 300 200 100 50 360 340 320 ASM angle relative to ship bow, direction from [deg] 300 280 260 240 Wind 30m/s ( 60kts) Wind 20m/s ( 40kts) Wind 10m/s ( 20kts) Wind 0 m/s (0 kts) Size reference [m] - 220 200 180 160 140 120 100 80 60 40 20 0 True wind angle relative to ship bow, direction from [deg]
What happens when the missile seeker range gate depth is reduced to be more representative of modern missile seekers?
Miss distance plot - 4 Launchers, 5 chaff rounds each, Short range gate Ship heading = 000 deg, speed = 0 m/s 500 400 300 200 100 50 360 340 320 ASM angle relative to ship bow, direction from [deg] 300 280 260 240 Wind 30m/s ( 60kts) Wind 20m/s ( 40kts) Wind 10m/s ( 20kts) Wind 0 m/s (0 kts) Size reference [m] - 220 200 180 160 140 120 100 80 60 40 20 0 True wind angle relative to ship bow, direction from [deg]
The Remedy A well designed algorithm will optimally calculate a chaff pattern. Maximize amount of chaff inside missile range gate. Ensure maximum missile miss distance from ship in as many situations as possible. Every time check results by repeating the batch process.
DSADM: Distributed SADM The process above on one machine can take many days to simulate (single thread application using one CPU core). Sometimes between the external C2 and SADM something goes wrong and then the whole batch run needs to be restarted (this is vastly improving with newer releases). By using DSADM to spread the load to a cluster of machines significantly reduces the simulation time Adding virtual machines to the mix gives another order of magnitude improvement as will be shown next.
DSADM cluster setup with an external ship C2 Only one SADM process and external C2 set, managed by DSADM service, can run on a machine. To take advantage of modern multi-core CPU's, virtual machines are the logical solution. Typically one virtual machine for each CPU thread seems to be optimal for CPU utilisation and robustness. ExternalC2.exe (and dll's connecting to it) is only in one place from which all nodes run it automatically. Thus only one place to update it.
DSADM cluster setup with c2extern
DSADM cluster hardware performance Core i7 machines found to perform extremely poorly when using Windows (XP Pro 64bit) as operating system hosting the Virtualbox virtual machines. Network traffic between DSADM controller and nodes is quite low, although the controller does run into issues when its machine is stressed by also running some virtual machine nodes on it.
DSADM Cluster hardware performance
DSADM cluster hardware performance (2)
DSAM cluster startup with external ship C2 Each DSADM node mounts the shared directory containing c2extern.exe upon start-up (.bat file used) In each DSADM node another.bat file in the locally mounted shared directory starts up c2extern.exe and restarts it each time it closes. Machine sharing directory with c2extern.exe and dlls use Ubuntu as operating system (Windows has issues with number of connections)
mount_c2_dir.bat run by node after booting @echo off if exist (z:\nul) net use z: /delete echo waiting for mapped network drive (z:) to become available... :start net use Z: \\10-0-0-5\shared /persistent:no dummy_pwd /user:dummy ping -n 2 127.0.0.1 > nul if not exist z:\nul goto start echo z:\ now available, can move path there and run batch starting c2_extern.exe z: start_c2.bat
start_c2.bat in c2extern.exe directory @echo off if not exist z:\c2_log\nul mkdir z:\c2_log :start ping -n 3 127.0.0.1 > nul echo ------------------------------------------------------echo starting external C2 and restart when done c2_extern.exe /log goto start
In essence three simulation clusters function together in some chaff simulations On each node of the simulation cluster the following runs: SADM simulation environment Ship external C2 model Ships with effectors and sensors, threat missiles, signal propagation and the environment Includes specific features of SAN Frigate Proprietary chaff algorithm dll e.g. RheinMetall MASS algorithm Wibu cluster license manager
Chaff sky writing! Since SADM 5.2 the number of chaff rounds have been greatly improved and launchers became trainable in elevation. The following message was proof sent to the developers of the RheinMetall MASS chaff system that SADM finally had the ability to simulate their chaff patterns.
Questions?