A new design methodology for civil aero-engine control
Submitting Institution
University of SheffieldUnit of Assessment
General EngineeringSummary Impact Type
TechnologicalResearch Subject Area(s)
Mathematical Sciences: Applied Mathematics
Information and Computing Sciences: Artificial Intelligence and Image Processing
Engineering: Mechanical Engineering
Summary of the impact
A unified design methodology for tuning gas turbine engine controllers,
developed by researchers in the Department of Automatic Control and
Systems Engineering (ACSE), is being used by Rolls- Royce across its
latest fleet of Civil Aero Trent engines. Trent engines are used to power,
for example, Boeing 787 Dreamliner and Airbus A350 aircraft that have been
adopted by the world's leading airlines.
This new methodology has made economic impact through the introduction of
a new process for tuning gas turbine engine controllers leading to the
adoption of a significantly changed technology. Indicators of impact are:
i) a new control law and design practice, resulting in a unified
approach for different projects;
ii) reduced development effort by shortening and simplifying the
design exercise and rendering it suitable for modular insertion; and
iii) streamlined verification requirements.
Underpinning research
This research was carried out in the Rolls-Royce University Technology
Centre in Control and Systems (RR-UTC), hosted and led by ACSE. RR-UTC was
established in 1993.
Research undertaken by RR-UTC into gas turbine control law and
architecture design, aimed at improving the efficiency and performance of
civil gas turbine engines and at reducing development effort, led to the
improved design methodology.
In 1994-5, as part of an EPSRC grant (GR/K36591, £156K), Chipperfield
(RA) and Fleming (PI) demonstrated [R1] that genetic algorithms afforded
the flexibility to search over both controller structure space as well as
parameter space to identify the best controller for a specific gas turbine
engine (GTE).
In 1998-9, in research funded by Rolls-Royce [text removed for publication],
Thompson (RR-UTC RA), Chipperfield (RA), Fleming (PI) [text removed for publication]
successfully applied evolutionary multi-objective
optimisation in an investigation of alternative distributed control
configurations for GTEs [R2], taking account of a range of competing
criteria (covering cost and weight, risk, fault diagnosis capability and
ease of maintenance).
With Fleming (Supervisor), in the period 1999-2002,
PhD students, Silva and Khatib, used evolutionary multi-objective
optimisation to explore trade-offs between maximising thrust and
minimising fuel consumption and the role of inlet and outlet geometry
modifications in conjunction with control law tuning [R3].
In a DTI-funded project with Rolls-Royce and TRW (£148K, 1999-2002),
Fleming, with UTC RAs (Breikin, Thompson and Hargrave) and PhD students,
(Regunath, Kim and Herbert) investigated low emissions combustion control
for GTEs [R4] and, having identified an optimal fuel split between the
pilot and main zones of the engine combustor, control strategies were
successfully tested at industrial rig test facilities.
Rolls-Royce Civil Aerospace has historically used linear, gain-scheduled
proportional-integral (PI) compensation for control of gas turbine fuel
flow. Engine dynamics vary with flight and power conditions, and a lengthy
design and verification process is required to meet the specification for
all conditions. The RR-UTC research studies carried out on control systems
architectures for GTE's, which took into account functional as well as
operational requirements, informed the choice of the new control law and
design methodology for Rolls-Royce civil aero-engines, which address the
shortcomings of the traditional control approach.
Following a feasibility study within the UTC, the control system
architecture and the design methodology was developed within a PhD
programme (Dr Shahid Mahmood), 2002-2007, championed by UTC Research
Fellow, Dr Ian Griffin. Griffin and Mahmood worked closely with engineers
within the company taking the concept through Technology Readiness Levels
to maturity.
The new control architecture and design methodology are based on a
nonlinear inverse model. Specifically, the nonlinear model [R5, R6], which
describes the response of a dominant engine state variable (high pressure
shaft speed) to known changes of fuel, is inverted such that the required
fuel flow can be computed for a given flight condition. The behaviour of
the resulting controller is analogous to that of a nonlinear PI controller
in which the proportional and integral terms are a function of the state
of the controller and the proportional and integral gains are non-
interacting.
Compared with the established approach, the new control law minimises
dependence against engine power level and removes dependence against
altitude. The new control law also delivers improved transient performance
whilst maintaining robustness.
Further, and perhaps most importantly, the new design methodology
requires fewer tuning parameters (a ten-fold reduction, from 55 down to
5), thus accelerating development time and significantly reducing the cost
of downstream software verification.
During the course of the five-year programme within the UTC, besides
demonstrating the enormous practical advantages of this new design,
difficult tuning and architectural problems were overcome by the
introduction of a number of practical innovations, e.g. [R6].
References to the research
Key papers providing evidence of the quality of the underpinning
research:
R1. Chipperfield A and Fleming P: Multiobjective gas turbine engine
controller design using genetic algorithms, IEEE Transactions on
Industrial Electronics, vol. 43, no. 5, pp. 583-587 1996. doi: 10.1109/41.538616
R2. Thompson HA, Chipperfield AJ, Fleming PJ and Legge C: Distributed
aero-engine control systems architecture selection using multi-objective
optimisation, Control Engineering Practice, vol. 7, no. 5, pp.
655-664, 1999. doi: 10.1016/S0967-0661(99)00011-8
R3. Silva VVR, Khatib W and Fleming PJ: Performance optimisation of gas
turbine engine, Engineering Applications of Artificial Intelligence,
vol. 18, pp. 575-583, 2005. doi:
10.1016/j.engappai.2005.01.001
Additional references:
R4. Breikin TV, Herbert ID, Kim SK, Regunath S, Hargrave SM, Thompson HA
and Fleming PJ: Staged combustion control design for aero engines, Control
Engineering Practice, vol. 14, pp. 387-396, 2006. doi: 10.1016/j.conengprac.2005.02.005
R5. Davies C, Holt JE, Griffin IA (2006). Benefits of Inverse Model
Control of Rolls-Royce Civil Gas Turbines, Proc UKACC International
Control Conference, Control 2006, Glasgow, 2006.
R6. Mahmood S, Griffin IA, and Fleming PJ: Robust control of a gas
turbine engine with variable power offtake, Proc ASME Turbo Expo 2006,
GT2006-91266, Barcelona, Spain, 2006.
Details of the impact
Approach to impact
Initially, the research (see Section 2) was undertaken by the team in the
Rolls-Royce UTC in Control and Systems Engineering within ACSE who
rigorously worked the programme through Technology Readiness Levels (TRLs)
2 and 3 from 2002-2004.
In 2005, the academic team were invited to work closely with an
industrial research team [text removed for publication] to demonstrate the
feasibility of the concept for large civil engines, working through TRLs 4
to 6.
The role of the RR-UTC team was to provide technical advice, design
procedure detail and supporting evidence through detailed simulations. A
team of experts representing systems, engine operability, performance and
control monitored the scheme.
The scheme was submitted to an Approval Meeting [text removed for publication]
for acceptance, and it proceeded through the remaining TRLs.
Economic impacts
1. Rolls-Royce's business performance has been improved through the
introduction of the new design process
The new design process achieves a ten-fold reduction in the number of
tuning parameters (from 55 to 5) without detriment to control system
performance [S5, S6]. This simplification of the design has led to cost
savings as a result of improved design practice, reduced development
effort, and streamlined verification requirements.
Improved design practice:
The simplified design process provides a unified approach across
different projects. The simplified process is intuitive and makes the
controller easy to tune, thereby making it easier to train engineers to
use it. By adopting the same design methodology across the emerging
Rolls-Royce fleet, the company has made continued cost savings across
projects.
The process has proven robust. A key characteristic of the design
methodology is the invariance of the designed controller to changes in the
engine design throughout the Engine Development Programme.
There is no need to obtain a linearised model at various set points and
no need for an associated frequency-domain analysis. Dispensing with the
need for frequency-domain analysis enables the Rolls-Royce control
engineers to align with the non-linear time-domain models used by the
company's Performance Department.
Reduced development effort:
The design exercise is:
shortened - requiring fewer tuning parameters;
simplified - it is no longer necessary to develop linearised GTE
models, nor is it necessary to use these models to define gains at all
power conditions and all altitudes; and
suitable - for modular insertion within the existing architecture
and, subsequently, to become part of the Rolls-Royce Product Line.
As a result, development effort is both reduced and simplified, leading
to the need for fewer and less specialist design engineers.
Streamlined verification requirements:
The costs associated with safety-critical aerospace software
certification are substantial and inroads into these costs are hugely
beneficial commercially to the company. The significant reduction in the
number of tuning parameters automatically leads to a simplification of
low- level software verification effort. The power dependency is now
encapsulated in a single internal table and the altitude dependency is
expressed within a simple thermodynamic scaling term.
In the context of the certification-driven environment for introducing
changes to civil engines, the adoption and implementation of this new
control methodology to be used by Rolls-Royce engineers across the range
of configurations across the Trent engine fleet is a remarkable
achievement [S1]. "This is the first new radical control law for over
20 years" (Control Systems Design Architect, Rolls-Royce).
2. A sector has adopted a significantly changed technology
Control laws designed using this methodology are now implemented across
the latest Trent fleet of engines. The Rolls-Royce Trent 1000 engine,
achieved on-time certification on August 7, 2007, over seven months ahead
of the rival General Electric engine for the Dreamliner [S3]. In 2013
Rolls-Royce was awarded certification for its higher efficiency and thrust
"package C" variant of the Trent 1000 engine that will power the Boeing
787-9 Dreamliner aircraft [S4]. The new control system was granted a US
Patent in 2012 [S2].
The Dreamliner's first flight was in December 2009 and it entered service
with All Nippon Airways (ANA) on October 26, 2011. The engine, has
delivered a dispatch reliability of better than 99.9%, which is a record
for a wide body aircraft engine [S4].
The control law is currently in use on Trent 1000 (Boeing 787
Dreamliner), Trent XWB (Airbus A350 airframe) and BR725 (Gulfstream G650)
engines, and is planned for use on all future variants.
Over 20 customers have selected the Trent 1000 to power their 787
Dreamliners and these include All Nippon Airways, Air China, Air Europa,
Air New Zealand, British Airways, Delta, Icelandair, International Lease
Finance Corporation, LOT, Thai Airways and Virgin Atlantic. Overall
commitments for the Airbus A350, powered by Trent XWB, total more than 592
with 34 customers. With spares, this will mean a requirement for more than
1100 engines. Firm aircraft orders include customers such as Emirates,
Qatar, Cathay Pacific and Singapore Airlines.
The control law has also been implemented and validated on industrial
engines - evidence of its suitability as part of the wider Rolls-Royce
Product Line. Important reported benefits concerning its implementation on
industrial engines include simpler code implementation, reduced CPU
processing time and less code to maintain as a result of the control code
being reduced by approximately 40%.
3. Performance has been improved through highly skilled people taking
up specialist roles drawing on their research
[text removed for publication]
The UTC is also proud to record that well over twenty of its engineers
have been recruited into Rolls-Royce and Aero Engine Controls; a
significant proportion already hold senior technical positions within the
company.
Sources to corroborate the impact
Factual Statement:
S1. Letter from the Chief of Research and Technology, Electrical Power
and Control Systems, Rolls-Royce, plc. This corroborates the claims made
in this case study regarding the impact of the research undertaken in
RR-UTC, which led to the adoption of a new design methodology for use
within the company for civil aero-engine control and industrial gas
turbine engine control.
Patent granted in USA:
S2. S. Mahmood: Control System. US Patent, US8321104 B2 http://www.google.co.uk/patents/US8321104.
Reports, reviews, web links and other documented sources of information
in the public domain:
S3. R-R press release 2007: Trent 1000 certification and reliability http://tinyurl.com/qfc5k4z
S4. R-R press release 2013: Trent 1000 `package C' variant certification
http://tinyurl.com/okc6s46
Confidential reports and documents:
S5. DNS104232 - `Boeing 787 Airframe / R-R Trent 1000 Control System
Sub-System Definition Document (SSDD): Control Laws'. Cerith Davies.
Original Issue 1 dated June 2005.
S6. DNS122060 - `User Guide for the Fuel Control Laws Rolls-Royce Inverse
Model (RRIM)'. C Davies/I Griffin. Original Issue 1 dated March 2007.