www.PosterPresentations.com ! Five Component Subsystem of Magellan GPS 315 C(:-* D(,= D:*6 @E,:.-. ;(2F.G-(-: HI+F.,*2+-* (J ;(2F5*+, 71:+-1+8 7*+K+-)L .-E C()+ M22.-5+> N.2:,+OPQ.,R5+O H716((> (J 7?)*+2) S M-*+,F,:)+)8 7*+K+-)L T:)6.=6. 76.,2. HI+F.,*2+-* (J ;(2F5*+, 71:+-1+L8 7*+K+-) U-)/*5*+ (J V+16-(>(G?8 WC 1"*!2$+3"'+- 4"5%-,') +'5 6,#2+-,7+3"' "8 $9% :;%/$# "8 1"2'$%&8%,$ 1"*!"'%'$# < <. 4=>?6@>?=AB :CC:1>D =C 1=EA>:FC:?> 1=4G=A:A>D Why do we care? ! In complex multi-component systems, the components are either Original Equipment Manufacturer (OEM), purchased from a trusted supplier or purchased from an unverified source. ! The unverified source could potentially introduce counterfeit components with the purpose of 1) malicious intent 2) espionage or 3) economic advantage. ! Counterfeit components lead to lack of assurance of reliability of complex systems, and it may endanger operational performance and safety. What do we propose? We propose the construction of a stochastic computational model: ! for identifying counterfeiting and studying its effects in the military supply chain, and ! for the simulation to compare expected failures of a system as a whole versus failures due to the counterfeit components of lesser quality. H. 1@D: D>EIJB 4@K:LL@A KGD M. ?AGE> I@>@ C=F >N: 1=4GE>@>?=A@L 4=I:L O. D?4EL@>?=A F:DEL>D <P. 1=A1LED?=AD @AI CE>EF: Q=FR ! The output of the computational model contains two parts: a) Plots of Failure Counts for both Verified and Counterfeit Components b) Visualization of the System CONCLUSIONS ! We build a stochastic agent-based computational model using the Stochastic Pi Machine (SPiM). ! We discuss the combinatorial complexity of the agent-based model. ! We implement a visualization of the system. ! We perform statistical tests to analyze the difference in multi-component system’s performance relative to the proportions of verified and counterfeit components. ONGOING AND FUTURE WORK ! Compare field data. ! Handle counterfeits with non-failure properties. For example, performance of counterfeit components can be affected by exposure to heat. ! Apply our stochastic computational modeling approach to other complex interdisciplinary domains. Figure 1: Computational Modeling Approach S. D>@>?D>?1@L >:D>D T. 1=4U?A@>=F?@L 1=4GL:V?>J =C >N: @K:A>WU@D:I 4=I:L Figure 2: A handheld and waterproof GPS used for hiking X. 1=4GE>@>?=A@L 4=I:L?AK @GGF=@1N ① Identify and index all components in the system and populate sets P and CP, where P i P, if P is a component from a verified source (OEM or trusted supplier) and CP i CP, if CP is a component from an unverified source. ② Baseline the system performance as per contractual component performance requirements. ③ Understand Counterfeiting Effects: ④ Use a proportions test to statistically identify changes in performance based on the failure counts. Figure 3: Agent-Based Model of Magellan GPS 315 Y. 1=4GE>@>?=A@L 4=I:L?AK ?A D>=1N@D>?1 G?W1@L1ELED Stochastic Pi-Calculus ! Allows modular description of concurrent and non-deterministic systems. ! It contains: • Processes – Agents • Channels – Inter-agent communication (!/? Handshake) • Stochastic rates associated with channels. • Both channels and processes can be dynamically created. Figure 4: Process Model of Effects of Verified and Counterfeit Components ! Assembled Systems – 4 Types ! Failed Assembled Systems – Non-Counterfeit Case • Failure due to 1 Verified Component • Failure due to 2 Verified Components • Failure due to 3 Verified Components • Failure due to 4 Verified Components • Failure due to 5 Verified Components ! Failed Assembled Systems – Counterfeit Case • Failure due to Single Counterfeit Component • Failure due to One or More Counterfeit Components Z. D>=1N@D>?1 G? 4@1N?A: [DG,4\ 1=I: let Unit() = do ?a1(); ?a2(); AUnit(0.0); !c0() or ?ca1(); ?a2(); AUnit(1.0); !c1() or ?a1(); ?ca2(); AUnit(2.0); !c2() or ?ca1(); ?ca2(); AUnit(12.0); !c12() and AUnit(n:float) = do ?cf1(); if n = 1.0 then !fc1() !fc1(Send) and ?cf1(Receive) else if n = 12.0 then !fc121() else () or ?cf2(); if n = 2.0 then !fc2() else if n = 12.0 then !fc122() else () or ?cf1(); ?cf2(); if n = 12.0 then !fc1212 else () Let P1() = !a1(); AP1() and AP1() = [email protected]; !f1() Let CP1() = !ca1(); ACP1() and ACP1() = [email protected]; !cf1() Let P2() = !a2(); AP2() and AP2() = [email protected]; !f2() Let CP2() = !ca2(); ACP2() and ACP2() = [email protected]; !cf2() Figure 5: 3 Runs; Run 1 (Black), Run 2 (Blue), Run 3 (Red) Table 1: Input data for the Computational Model Figure 6: 1 st time stamp – Configuration of assembled systems (ASystem) Figure 7: Last time stamp – Failed assembled systems (FSystem) Table 2: Statistical Hypothesis Testing for Simulation Results in Figure 5 <<. F:C:F:A1:D 1. Vishakha Sharma, Adriana Compagnoni and Jose Emmanuel Ramirez-Marquez. Computational Modeling of the Effects of Counterfeit Components. To appear in Proceedings of the Summer Computer Simulation Conference (SCSC), Monterey, CA, July 6 - 10, 2014. " "