07-07-2014, 03:15 PM
Optimal Self-Diagnosis Policy for
Optimal Self-Diagnosis.pdf (Size: 370.12 KB / Downloads: 20)
Abstract
-FADEC (full authority digital engine control) is widely adopted for gas turbine engine
controllers. The advancement of microelectronics produces high speed general purpose programmable
logic controllers (PLC) with low price. When they are adopted for FADEC, we can expect high cost
performance engine controllers. However, these PLCs were originally developed for ordinary industrial
machinery controllers and PLC makers prohibited their use as gas turbine controllers because of
their low reliability. Engine makers should give some measures to hold enough reliability when they
apply PLCs for FADEC. In this paper, a FADEC is self-diagnosed at the nth control calculations.
Introducing an expected cost until self-diagnosis and an expected cost per unit time, optimal policies
which minimize them are discussed. Numerical examples are finally given. @ 2003 Elsevier Ltd. All
rights reserved.
Keywords-FADEC, Gas turbine, Diagnosis policy, Expected cost, Optimal policy.
1. INTRODUCTION
The original idea of gas turbine engines was proposed by Barber in England at 1791, and the
engine was first realized in the 20 th century. After that, they advanced greatly during World
War II. Today, gas turbine engines are widely utilized as main engines of airplanes, high performance
mechanical pumps, emergency generators, and cogeneration systems because they can
generate high power compared to their sizes, their start times are very short, and no coolant
water is necessary for operation.
Gas turbine engines are mainly constituted with three parts: compressor, combustor, and
turbine. The engine control is performed by governing the fuel flow of the engine. When we
operate gas turbine engines, we should pay attention to surge, stool, and catastrophic high
temperature of turbine inlet gas because these phenomena cause serious damage to the engine.
To prevent such dangerous phenomena, the turbine speed, inlet temperature and pressure, and
exhaust gas temperature of gas turbine engines are monitored and engine controllers determine
appropriate fuel flow based on these data.
0895-7177/03/$ - see front matter @ 2003 Elsevier Ltd. All rights reserved.
doi: 10.1016/S0895-7177(03)00337-6
Typeset by AM-T@
1244 K. ITO AND T. NAKAGAWA
The gas turbine engine operates in a harsh environment and a hydromechanical controller
(HMC) is adopted for engine controller for long periods because of its high reliability, durability,
and excellent responsibility. However, the performance of gas turbine engines have advanced
and customers need to reduce the operation cost, and so, HMC could not follow these advanced
demands and the engine controller has been electrified. The first electric engine controller which
was a support unit of HMS was adopted for 547-17 turbo jet engine of the F86D fighter in the late
1940s. By the change of devices from vacuum tube to transistor and transistor to IC, the roll of
electric engine controllers has changed from the assistant of HMS to the full authority controller
because of the increase in reliability. In the 196Os, the analogue full authority controller could not
follow the accuracy demand of engines, and a full authority digital engine controller (FADEC)
was developed [l-3].
A FADEC is an electric engine controller which performs the complicated signal processing of
digital engine data. Aircraft FADECs which are expected to have high mission reliability and
are needed to decrease weight, hardware complication, and electric consumption generally form
a duplicated system [4-61.
Industrial gas turbine engines have been advanced in absorbing key technologies which were
established for aircraft gas turbine engines. FADECs which were originally developed for aircraft
have also been adopted for industrial gas turbine engines. Comparing between general industrial
gas turbine FADECs and aircraft FADECs, the following differences are recognized.
(1) Aircraft gas turbine FADECs have to perform high speed data processing because the
rapid response for aircraft body movement is necessary, and inlet pressure and temperature
change greatly depend on height. On the other hand, industrial gas turbine FADECs are
needed to have such high performance compared to aircraft ones because they are operated
at steady speed on ground.
(2) Aircraft gas turbine FADECs have to be reliable and tolerable for faults, and so, they form
a duplicate system because their malfunction in operation may cause serious damages to
aircraft and crews. Industrial gas turbine FADECs also have to be reliable and tolerable,
and still be low cost because they have to be competitive on the market.
Due to the advance of microelectronics, small, high performance and low cost programmable
logic controllers (PLC) have been distributed on the market. Their origins were relay sequencers
and still were utilized as sequencers of industrial automatic systems. Applying numerical calculation
ability of microprocessors, these PLCs occupy the analogue-digital and digital-analogue
transformer and can perform numerical control. When we use such PLCs, a high cost perfor-
.mance FADEC system can be realized. However, these PLCs are developed as general industrial
controllers and PLC makers do not permit them for use with high temperature and pressurized
hot gas controllers. Thus, gas turbine makers which apply these PLCs as FADEC have to design
some protective mechanism and have to assure high reliability of FADEC.
In this paper, we consider self-diagnosis policies for the gas turbine FADEC, and obtain the
expected costs. Further, we discuss optimal policies which minimize the expected costs.
CONCLUSION
We have considered two optimal self-diagnosis policies for FADEC: A FADEC performs the
control calculation at time interval TO and its self-diagnosis is performed at the nth calculation.
The expected cost per unit time and the expected cost per cycle are derived, where the time
interval from a diagnosis to the next diagnosis is defined as a cycle. We have shown that there
exist optimal values n* in both policies. Numerical examples are given when the failure time has
an exponential distribution.