IPS NPS Overview of Research Capabilities at

IPS @ NPS Overview of Research Capabilities at NPS in support of ESRD and the ESO

NPS • Fully accredited university. • Primarily MS degrees, increasing numbers of Ph. Ds. • Engineering and non-engineering programs. • Around 2000 resident students and 200 faculty. • Students are mostly military (all branches) and a few Do. D civilians and contractors. 2

Students • Students: – Graduate students (Master, Engineering Degree, Ph. D) – Post-doc • The ECE and MAE departments are ABET accredited at the graduate level and students have to write a thesis of research quality. Systems Engineering Department accreditation is pending. • Faculty members have to fund at least one quarter support (>44 days) per academic year, from non NPS sources. 3

Personnel • Naval Systems Engineering: – Dr. Fotis Papoulias, MAE, Total Ship Systems Engineering (TSSE) – Dr. Cliff Whitcomb, SE, Systems Engineering • Power Systems: – – Fr. Robert Ashton, ECE, Power Systems Dr. Alex Julian, ECE, Power Systems Dr. Roberto Cristi, ECE: Control Systems and Signal Processing Dr. Giovanna Oriti, ECE, Power Systems • Weapon Systems: – Dr. Bill Colson, PH, Directed Energy Weapons – Dr. David Jenn, ECE, Sensors – Dr. Bill Meier, PH, Rail Gun 4

Education • Graduate Certificate Courses: – Electric Machinery Theory – Advanced Electrical Machinery Systems – Solid State Power Conversion – Advanced Solid State Power Conversion • Courses offered both as resident and on a Distance Learning basis (VTC and Web-based) 5

Related Research • Power Systems, on-going research by Julian and Ashton. • Electric ship trade-offs: Ongoing by Papoulias and Whitcomb in collaboration with MIT 2 N program. • Control Systems and Digital Signal Processing: Recent work by Stevens and Cristi, ongoing. 6

A Polynomial Approach to Unknown Input Observers Applied to a DC Zonal Electrical Distribution System for Robust Model-Based Fault Diagnosis R. Cristi, J. D. Stevens, LCDR, USN

Fault Tolerant Controller/Plant Re-design Fault Diagnosis Supervisory Level Execution Level Controller Plant 8

System of Interest: DC ZEDS Model 9

How to detect and isolate the fault For a System of Systems: sensors 10

How to detect and isolate the fault See each system: sensors Use a mathematical Model of each subsystem to generate residual error: Math. Model observer Fault occurs residual 11

Application to Fault Detection Given the system with inputs (actuators) and outputs (sensors) Design a family of observers insensitive to some of the signals: Obs. Fault in 12

Application: DC ZEDS Model Port Bus Power Supply SSCM Zone 1 SSIM SSCM Zone 2 CPL SSIM SSCM Starboard Bus Zone 3 CPL SSIM CPL SSCM Power Supply 13

Application: LAM Power Supply • Assumptions – Ignore signal-conditioning Filters – For commanded bus voltage • Ignore the short circuit/over-current protection • Ignore slew-rate limiter – Ignore anti-windup limits on integrators – Ignore thyristor voltage drop in – Ignore resistance in parallel with • States: • Inputs: • Outputs: 14

Application: LAM Power Supply • • Top Row: UIO insensitive to Bottom row: Observer sensitive to Left Column: Fault on Right Column: Fault on 15

Other Slides 16

Observers and Residuals Modeling Errors sensor noise residual 17 Fault occurs

How to detect and isolate the fault For a System of Systems: sensors Not Available! observer 18

Unknown Input Observers Need for Unknown Input Observers (UIO) UIO • We need a sufficient number of outputs • Some inputs can be observed, but not all • Need to compute residuals to check for faults • Try to estimate missing inputs 19

Available UIO’s A few approaches available: • S. Hui and S. Zak, "Observer Design for Systems with Unknown Inputs, " International Journal of Applied Mathematics and Computer Science, vol. 15, pp. 431 -446, 2005; • J. Y. Keller and M. Darouach, "Two-stage Kalman Estimator with Unknown Exogenous Inputs, " Automatica, vol. 35, pp. 339 -342, 1999. ; • W. S. Kerwin and J. L. Prince, "On the Optimality of Recursive Unbiased State Estimation with Unknown Inputs, " Automatica, vol. 36, pp. 1381 -1383, 2000. Observations: 1. They all try to reconstruct the full state, thus they are subject to constraints. 2. We just want to computer the “residuals” and check for consistency between observations and mathematical models; 3. We developed our own approach based on Partial State and Polynomial Matrices. It is more general and it seems more suitable to our problem. 20

Input Observability and Estimation • If system satisfies input observability criteria, then input estimation is possible. 21

Input Observability and Estimation • Scaled Port Bus Model: • Input is subject to unknown disturbance with known 1 st and 2 nd moment statistics. • One input is subjected to a completely unknown disturbance • Object: determine 22

Input Observability and Estimation • Structured Residual Set • Input/output consistency is calculated in each channel • Channel 1: – estimate inputs/outputs without using • All other channels – will determine estimates using * - faulted signal 23

Input Observability and Estimation • Input residuals from UIOs are along the columns. • Column U~234 shows no residual except in residual based on (first row). 24

Input Observability and Estimation • Output Residuals – No residuals in Column 1, except – is dependent on the faulted input • Output residuals indicate fault, but does not estimate the fault. 25