This text describes how to use advanced methods and analyses to create a unique control system for a modern statically unstable aircraft such as JAS 39 Gripen, referred to hereinafter as Gripen.
The section starts with a description of the actions that Saab took after an accident that occurred with Gripen in August 1993 during the Stockholm Water Festival. The causes of the accident are described here, as are the actions taken to fix the problems that existed in the control system. This section then explains how we then successfully developed the world’s best control system for military fighter aircraft.
The description provides detailed descriptions of the work carried out to produce control laws for a flight control system. Model-based development was used early on in this work, both to make design more effective and to ensure the quality of the results.
Here we describe how Saab adapted Gripen for export. The requirement for exporting Gripen was to adapt it to NATO standards. Within the context of this work, we explain how to calculate stability as well as handling and flight characteristics in order to handle different types of external store alternatives, meaning that it should be possible to fly with a large number of possible combinations of wing-mounted stores. We describe how heavy and very aft centre of gravity external store alternatives affected the aircraft’s handling and flight characteristics.
This section considers three different perspectives as follows:
The author recommends the following texts that relate to this story: In chapter Creating value for customers under the heading Development of Technology Demonstrators in chapter Keeping unique development skills under the heading Technical capability development for control system and under the heading Designing the world’s best flight control system.
This text concerns the highlighted areas of A Journey of Change in the Aircraft Industry
An accident with Gripen occurred in August 1993 during the Stockholm Water Festival. The accident was caused by the conventional rate limiters used in the control laws.
The major loss of phase with large and fast pilot stick commands created a large time delay, which was the real cause. Right before the accident, the pilot’s commands were so large and fast that they were limited by the conventional servo command control rate limiters so that a PIO of category III occurred.
Gripen is a modern statically unstable aircraft. This means that the control system has to stabilise the aircraft at all times in the right way so that divergence and instability do not occur.
The aircraft’s sensors detect the change in the aircraft’s motion and the change in load factor using the signals from the various sensors, and with that information the control laws can generate a counter-command with the control surfaces to stabilise the aircraft. This forms a feedback control law. The feedback control laws can also be used to reduce disturbances from outside that affect the aircraft, such as gusts of wind, transonic transients and transients from weapon delivery.
In addition, the feed-back control laws can also reduce the effect of passing through a wake vortex. Aircraft with lift generate rotating vortices that are released behind the aircraft; two contrarotating wingtip vortices are formed, which are wound up downstream. The distance between the vortices depends on the wing span of the generating aircraft, and the strength of the vortices depends on the weight, load factor and flight speed of the generating aircraft. The effect of the incoming aircraft depends on how the incoming aircraft hits the vortex packet.
The fact that the vortices survive far behind the generating aircraft can impact and disturb the next aircraft. Conventional statically stable aircraft have a naturally stabilizing restoring aerodynamic moment, while modern statically unstable aircraft do not. The restoring moment must therefore be generated artificially by the flight control system.
With the feedback control laws, the stability and flight characteristics can also be adapted, when the aircraft has heavy external stores mounted.
An accident with Gripen occurred in August 1993 during the Stockholm Water Festival.
The reason for the accident was that the pilot was able to generate large, fast pilot stick commands with an undamped mini stick. This saturated the conventional rate limiters that were used in the control laws.
The major loss of phase produced the time delay that caused the accident. The pilot’s commands became so large and fast that they were limited by the conventional control servo rate limiters, a PIO of category III was generated.
If the commands from the pilot and stabilising feedback commands are small and slow enough so they will not hit the conventional rate limitation, the handling and flight characteristics were very good.
In order to correct the error as fast as possible after the accident, flight control system editions with reduced performance were created. This means that the pilot had less load factor and angle of attack available. This action reduced the risk of the pilot’s command signal and feedback signals to be rate limited. The action was taken in order to buy time, and give Saab’s engineers time to fix the problem of major loss of phase from the conventional rate limiters, i.e. the time delay.
Carrying out this action meant that the flight testing organisation could continue to carry out test flights to test other systems in the aircraft.
Saab contributed to the accident investigations. A construction group was formed to investigate the incident and the construction group began the work of finding a solution to the problem. A number of foreign investigators monitored Saab’s progress.
The construction group began investigating the effect of the non-linear conventional rate limiters that were implemented into the control system.
The ability of the undamped mini stick to create large servo commands with high control servo speeds made the pilot and feedback command signals to be rate limited by the conventional rate limiters and this caused the accident.
So was it possible to construct a rate limiter that would limit the rate of the pilot’s control command signals and feedback commands without major loss of phase and time delay? This question was central to the continuing work.
The handling and flight characteristics within the linear range were very good, i.e. when the pilot’s commands and the feedback commands were not rate limited.
The construction group reached a solution after many discussions and meetings. Comprehensive theoretical studies and digital simulations had been carried out in MAHS (Manoeuvre Hydraulic Simulator).
The solution consisted of a rate limiter with a low phase loss and time delay.
The difference between the input and output signals of a conventional rate limiter was fed back via a constant and a filter, and this signal was subtracted from the input signal to the rate limiter.
The input signal to the conventional rate limiter was thus the difference between the input signal and the feed-back signal. The output signal from the adjusted conventional rate limiter then passed through another conventional rate limiter.
The result of the investigations is shown in the figure below. This shows the difference between a conventional rate limiter (red) and the difference in time delay between it and the phase-compensated rate limiter (blue). The input signal to each rate limiter is shown in green.
The purpose of the rate limiters is to limit the command rate to the servos so that the servos are always able to carry out their commands at different hydraulic pressures.
The servos cannot be commanded in acceleration limitation because this causes major time delays and losses of phase in the closed system’s stabilisation and pilot loop. This is why the phase-compensated rate limiters for the pilot commands were made depending on hydraulic pressure. If the hydraulic pressure in the system was low due to many and frequent commands from the pilot, the pilot’s command was reduced to a lower command rate so that the stabilising inner loop always had the authority to stabilise the aircraft.
With the improved phase-compensated rate limiter introduced into the control system on the pilot command and the sum of the pilot command and stabilising feedback, it remained to be demonstrated that the control system and control laws would not cause a PIO of category III or any PIO at all.
A thorough theoretical investigation followed, together with extensive tests with a pilot in the simulator. For this purpose, the highly realistic simulator MAHS was used, with a hydraulic system, a flight control system hardware and other realistic hardware. The pilots tested combination commands in pitch, roll and yaw to try to get the aircraft to diverge.
The pilot’s commands were then optimised and mechanised in the software so that batch simulations could be carried out. These calculations, together with theoretical studies, showed that Gripen’s flight control system was resistant to PIOs in categories I, II and III.
Saab is the world leader in knowledge of the causes of PIOs, and how to correct the problem. Furthermore, the phase-compensated rate limiter was patented, as it is unique in the aviation world and is a very effective rate limiter with a low time delay.
Saab’s invention of the phase-compensated rate limiter was presented at a conference in San Diego in 1996.
What was additionally needed in order to improve the development and test methodology was to make a tool lift for developing control laws for the flight control system. A new working procedure was also introduced in the form of model-based development. The new method that started to be used, Model Based Systems Engineering (MBSE), was modernised and improved in comparison with the previous method, which meant that the tools that were to be used needed to support this working procedure.
This was intended to simplify and improve the production of new flight control system editions and also improve the verification and validation of those editions. In addition, the flight testing should be supported with a highly useful simulator placed in the same building where the control laws were developed. One such simulator, called STYRSIM, was built and was highly beneficial in the production and testing of control laws.
Previously, there had been many time-consuming tasks and ineffective working procedures. One example of this was that, previously, block and logic diagrams of the control laws had first been drawn up by hand in the control system. Subsequently, these were programmed manually in Fortran. The program could then be used in simulations in the digital model and in simulators with a pilot.
Saab therefore created a tool that could be used to easily join finished blocks to block and logic diagrams of control laws, and then easily carry out analyses to see whether the construction was sufficiently good.
When the design was satisfactory, a button in the tool could be pressed to automatically generate a control system code that could be fed into simulation programs: firstly a simulation tool with six degrees of freedom called ARES for desktop simulations, and secondly a simulator for simulations with a pilot. Automatic documentation could be obtained for this.
When an edition of the control system had been constructed and simulated sufficiently in simulator programs and simulators with a pilot, the block diagrams were coded manually in the ADA programming language into the flight control system’s computers. Further simulations were then conducted with the relevant hardware, and ground tests were conducted in Saab’s simulators with a pilot before flying.
The manually coded ADA code was also compared bit by bit with the simulated auto code, to ensure that the functions had been transferred correctly to the flight control system’s computer.
In order to investigate what was on the market, various institutions and aircraft manufacturers in the USA were visited in 1993. It turned out that the only that met Saab’s requirements for a model-based design tool for control systems was one called SystemBuild.
Previously, block diagrams were drawn by hand and then programmed in the Fortran programming language. After this, the Fortran block diagram program was linked together with other parts of the control system for further simulation and testing.
Using the SystemBuild tool, it became possible to use finished blocks. The blocks could then be put together into a finished block diagram in the tool. The block diagram part of the control system could then easily be tested separately with finished commands in the tool. After that, the block diagram part was linked together with the next part and then with the rest of the control system’s block diagrams for further testing and simulation.
When the desired addition and function had been tested and self-verified, automatic code was generated by pressing a command button in the tool. The automatic code was then entered into the ARES simulation tool for desktop simulation and for simulation with a pilot. In ARES, simulations were done in six degrees of freedom.
In addition to introducing the SystemBuild tool, a simulator with very good outside world presentation was also needed. A very important stage of the work is when the pilot and engineer assess the control system edition’s control characteristics. These assessments relate to sensitivity and growth in aircraft response to a pilot command the aircraft handling characteristics, the aircraft’s flight characteristics with regard to frequency and damping, and whether the aircraft behaves correctly in gusts of wind.
During a visit to a simulator conference in Albuquerque in the USA, we saw an F-16 simulator with a 180-degree outside world presentation both in front and on each side and above. This simulator provided a very realistic flight experience.
A simulator of that type, STYRSIM, was built by Saab for use while designing control laws for the flight control system together with a pilot. The simulation was also used to verify these control laws. Here, the simulator could be used to set up test flights and to run through test schedules, so-called “knäblock”, that the pilot has for flight test. The simulator was also used for a dry flight test before the real test flight, to test different combinations of pilot commands. After a test flight, the test team was able to gather in the simulator, discuss the outcome of the test flight and compare it with the result from the simulation in STYRSIM.
The pilot and engineer sit very close so that they can discuss the control law design. The pilot also makes assessments of handling and flight characteristics using the Cooper-Harper scale.
Cooper-Harper ratings are a well established scale in the aviation industry, where pilots make subjective assessments and give scores to the flight and control characteristics of an aircraft.
As early as during the development of JA 37 Viggen, it was found that, when decelerating through the transonic range (decelerating through the speed of sound), a large transient was developed due to pressure changes and air flow separations, the so-called transonic transients.
This happened in particular in cases of rapid deceleration through transonic under load factors commands from supersonic speeds to subsonic speeds.
In order to mitigate this transient in JA 37 Viggen, the aircraft was changed so that the front part of the aircraft was angled down a little and bump was added to the aircraft body. Despite this, there was still a significant transonic transient when decelerating through the transonic.
An attempt was therefore made to use the autopilot in JA 37 Viggen to produce a counter-reaction to the transonic transient in order to mitigate the problem. Initially, a solution was attempted in which a countercommand on the elevons to the transonic transient was commanded out as a function of Mach number, a so-called feedforward compensation. This type of solution did not work because the transient did not always occur with the same exact magnitude at a specific Mach number.
Test aircraft no. 2 of Gripen demonstrated that this type of transonic transient existed also in Gripen at decelerations through the transonic range.
The transonic transient was primarily caused by a flow separation in front of the wing trailing edge control surfaces on the Gripen. The transient arose if the angle of attack exceeded approximately 6 degrees and the size of the disruption was dependant on how fast the deceleration through the transonic range was and what load factor the aircraft had. In addition, the transonic transient also dependeds on the external stores hanging on the aircraft’s pylons.
Test flights with Coca Cola shaped wingtip pylons showed that it was not possible to remodel the aircraft, to reduce the flow separation in front of the trailing edge control surfaces.
In the late 1990s, development was focused on using the flight control system to reduce the transonic transient that arises when the aircraft decelerates through transonic with a load factor command.
Unfortunately, wind tunnel tests cannot demonstrate this phenomenon because a wind tunnel produces reflexion flows from the wind tunnel walls at near speed of sound speeds, and the reflexion flows affect the measurement results. In addition, the phenomenon depends to a large extent on the rate of deceleration through the transonic range, as well as the load factors with which the aircraft is decelerating.
Previous flights had shown large transients due to flow separations in front of the trailing edge control surfaces. The phenomenon was demonstrated by means of tufting on the top of the wing. The phenomenon could be filmed with a camera mounted on the aircraft. The pressure drop in front of the trailing edge rudder gave rise to a fast and powerful disturbance. This occurred in a very narrow speed range – 0.92–0.95 of the speed of sound – and at an angle of attack greater than about 6 degrees.
The ARES digital simulation model with six degrees of freedom should be used to design the transonic control laws that reduce the transonic transient to an acceptable magnitude. It became clear that the ARES model did not give the correct magnitude at all for the transonic transient when Gripen decelerates quickly through the transonic range with a load factor pilot command. The basis for the model in the transonic range was thus uncertain and unsatisfactory.
Experience from previous aircraft showed that the aircraft’s sensors can only be used for the purpose to reduce the transonic transient such as angle of attack sensors, accelerometers and gyro signals to compensate and give counter-commands to the canards and wing trailing edge control surfaces (elevons), to reduce the transonic transient. The aircraft’s sensors detect changes in the aircraft’s movement, changes in load factor and loads from the transonic transient.
Creating a control law that effectively reduces the transonic transient requires that the control law compensates for the time loss that arises. First of all, it is necessary to measure what happens with the aircraft using the aircraft’s sensors in the load factor and angle of attack, then the control law takes actions and reduces the transonic transient with a canard and elevon command.
Since it takes a certain amount of time before the autopilot can command a counter-command, the time loss must be compensated by a prediction.
Some form of prediction of what would happen with the aircraft was therefore required. It was necessary here to use a predicted angle of attack. Another contributing factor to the problem with the transonic transient is the change in maximum load factor limit that exists when flying from supersonic speeds to subsonic speeds. The maximum limit loadfactor is 1 g less than when flying at subsonic speeds.
The deceleration through the transonic range is often very fast, which means that the change in the load factor command can itself be experienced as a transient.
This is dealt with by allowing the difference in the maximum load factor limitation of 1 g to depend on the magnitude of the deceleration rate, so that load factor limitation will change more slowly if the aircraft decelerates quickly.
The control laws for dealing with the transonic transient flow separation problem consist of a prediction of the angle of attack. This happens because a signal passes an amplification and a lead lag filter to command the canards and trailing edge control surfaces (elevons) as early as possible with the right magnitude.
Because the control surface efficiency of the trailing edge control surfaces (elevons) are three times greater than that on the canards. This is compensated for on the canards so that they contribute roughly with the same pitching moment.
To help get started with the design of an effective control law, an additional function was implemented in the ARES simulation model to produce a similar transonic transient as in real flight. Through extensive tests of this function, it was possible to establish the structure of the compensating control law in the flight control system.
In the control law, a number of control law gains were implemented that could be changed during flight tests using the flight test function (FTF). In the pilot’s test program, “knäblock”, a number of variations of these control law gains were added, from which the pilot was able to select following a decision from the ground test station (Houston) during the test flight itself.
The control law gains were increased as much as possible in order to reduce the magnitude of the transonic transient at fast decelerations through the transonic range with pilot load factor command.
A theoretical sensitivity analysis was carried out before the test flights were executed with the various FTF settings. The input data for the sensitivity analysis consisted of how much the aerodynamic data could vary in the speed sound range during these manoeuvres. Modern control theory was used to carry out these analyses.
Various external store alternatives were also tested, from aircraft with only wingtip guided missiles to aircraft with wing-mounted drop tanks and asymmetrical external store alternatives. The decelerations through the transonic range were carried out with different throttle settings, but mostly with flight idle, which gives the lowest motor thrust, which also gives the fastest deceleration through the transonic range.
The following lessons were learned from the design of the transonic transient control laws:
Aircraft with lift generate rotating vorticies that are released into the air behind the aircraft. Two contrarotating wingtip vortices are formed, which are wound up downstream. The distance between the vortices depends on the wing span of the generating aircraft, and the strength of the vortices depends on the weight, load factor and flight speed of the generating aircraft. The effect of the incoming aircraft depends on how the incoming aircraft hits the vortex packet.
The figure shows passage through the wake turbulence of a generating aircraft.
The two vortices survive relatively far behind the generating aircraft until they finally break down; the survival of the vortices may affect and disrupt the following aircraft.
Conventional statically stable aircraft have a naturally stabilizing restoring aerodynamic moment, while modern statically unstable aircraft do not. The stabilizing moment must then be generated artificially by the flight control system.
During a air to air aiming practice with two Gripen from the air force wing at F 7 Såtenäs, flying over Vänern, the following aircraft crashed when passing through the target’s wake vortex in a tight diving turn. The aircraft stalled and the pilot ejected. The investigation of the accident showed that Gripen’s behaviour when passing through a wake vortex had to be changed.
A work group was formed in autumn 1999 to create a simulation model with the effects of a wake vortex, to as soon as possible start with flight mechanics simulations. It needed to be possible to carry out the simulations both as batch simulations using the ARES simulation model and as simulations together with a pilot.
The physical models to be implemented into the simulation models had to be as simple as possible. The result was that an additional model was made that calculated additional forces and moments, added to the ordinary simulation model in ARES.
The angle of attack vanes on each side of the nose measure the airflow on the aircraft. When passing through a wake vortex, they are affected by the change in flow from the wake vortex. This information can be used by the flight control system’s control laws to reduce the aircraft’s movement in gusts and winds. Information from the angle of attack vanes makes it possible to ensure that the variation due to gusts when aiming is as small as possible.
For this reason, a model for how the angle of attack vanes are affected when passing through a wake vortex was also added to the ARES simulation model.
When the ARES simulation model had been updated with these simulation additions from the wake vortex forces and moments and the effect on the angle of attack vanes when passing through a wake vortex, a pre study could be initiated.
The purpose of the pre study was to produce a control law in the flight control system that would reduce the disturbance to the aircraft as far as possible when passing through a wake vortex. The modern statically unstable aircraft should behave like a conventional statically stable aircraft or better when passing through a wake vortex.
In the pre study, it was established that the effect on the angle of attack vanes from the wake vortex caused a disturbance to the aircraft that could lead to large angles of attack if the generating aircraft had a high load factor and a large angle of attack.
The pre study showed that it was possible to distinguish between dangerous passes through wake vortices and wind gusts. A detection function was therefore created, which switched control laws when the aircraft passed through a wake vortex. The control laws switched signals and then reverted back to the normal angle of attack information once the wake vortex had been passed. This technique greatly reduced the disturbance when passing through a wake vortex.
In order to verify this new control law with wake vortex detection and disturbance reduction when passing through a wake vortex, the ARES model with six degrees of freedom was used with the additional effects from passing through a wake vortex. The effect of the wake vortex on the incoming aircraft is caused by the load factor of the generating aircraft and the load factor that the incoming aircraft has, as well as the flight case that the aircraft is in and how the wake vortex is passed through.
The outcome of these batch simulations showed very good results. The control law in the flight control system that ensures that Gripen disturbs very little when passing through the wake vortex of a generating aircraft was verified, validated and ready for a test flight.
Flight tests were carried out with a wake vortex generating Gripen that would produce the wake vortices with different load factors. The incoming aircraft was also a Gripen, with complete measurement equipment.
In this way, it was possible to measure how the incoming aircraft hit the wake vortex and its effect on the aircraft, and it was also possible to verify that the control law in the flight control system detected and reduced the effect of the wake vortex when passing through the vortices. Passes were carried out symmetrically through the generating aircraft’s wake vortices, both from below and from above. In addition, passes through wake vortices were done laterally, from the side.
The figure shows various test flight manoeuvres for passing through wake vortices generated by a Gripen.
Test flights were also carried out to test that the detection function did not accidentally detect strong gusts of wind, but rather only wake vortices.
In addition, the effects of flying behind a tanker used for air refuelling were checked to see whether Gripen was affected by these slipstream disturbances. This test was carried out with a Hercules aircraft. The results of these test flights showed that there was no problem with flying very close to a tanker used for air refuelling with propellers producing slipstreams; Gripen remained steady in place.
Foreign aircraft manufacturers showed a good deal of interest in Saab’s work in this area.
In order to export Gripen, it was necessary to adapt the aircraft to NATO standards by making it possible to carry the external stores that the new customers required. It also needed to be possible to refuel the aircraft while airborne. For this reason, Gripen subseries 3 was created, which is available in two versions: Gripen C, with a single seat, and Gripen D, which is a two-seater version. There were also important changes here, such as larger multicolour indicators in the cabin and other improvements for international adaptation.
In order to extend flight time during a mission, the ability to refuel the aircraft while airborne via a refuelling probe was introduced in Gripen C and D. The refuelling probe is extended from the body of the aircraft behind the left canard (like a telescope) when it was time to refuel in the air.
The fact that the refuelling probe was on the left-hand side, right behind the pilot, could be a problem for the pilot’s ability to fly the refuelling probe right into the basket. Foreign aircraft manufacturers did not believe that the ordinary control law would be sufficient to fly the probe into the basket. They believed that a new, advanced control law was required. Test flights with air to air refuelling using various tanker aircraft showed that the ordinary normal control law in the flight control system was sufficient and worked well, with no PIO tendency problems.
It is, of course, possible to construct an automatic basket capture control law for the refuelling probe in the flight control system if suitable sensors are available, to make things even easier for the pilot.
Subseries 3 of Gripen was NATO-adapted to fly with heavy external stores in the various aircraft pylons. Some external stores resulted in very aft aircraft centre of gravity positions.
This means that there were a large number of possible combinations of wing-mounted stores that the aircraft could fly with. The heavy external stores with very aft aircraft centre of gravity positions affected the aircraft’s handling and flight characteristics. The aircraft’s centre of gravity was affected by many heavy external stores in that the centre of gravity ended up far aft, which made the basic aircraft more statically unstable.
This means that the stability and the handling and flight characteristics change to the worse according to the equation below. The basic aircraft’s Cm-alpha (blue) becomes more positive, i.e. more unstable, and the normal alpha feedback gain then becomes too low to achieve the final desired artificial C*m-alpha with the desired stability and the handling and flight characteristics.
CmalfaT = CmalfaF + KALFDC*Cmdc + KALFDE*Cmde
The blue Cm-alpha is the basic aircraft’s unstable characteristics. When the centre of gravity moves backwards, this Cm-alpha becomes still more unstable and therefore needs to be stabilised with extra alpha feedback gain to KALFDC and KALFDE.
KALFDC is the gain K from an angle of attack vane sensor to a canard command and KALFDE is the gain K from an angle of attack vane sensor to the trailing edge control surfaces.
In 2002, test flights were carried out with a heavyily loaded Gripen. It was then possible to show that this method was a feasible way of improving both the stability and the handling and flight characteristics for very heavy very aft centre of gravity external store alternatives.
With the Flight Test Function (FTF), the pilot was able to test various extra angle of attack feedback gain additions to KALFDC and KALFDE in the flight control system’s control laws.
Test flights were carried out with the extra angle of attack feedback gain addition using the FTF function, which showed that this was a feasible way of improving handling and flight characteristics when flying with heavy external stores. It was then possible to show that the types of requirement that export customers set in this regard could be met.
By adding an adapted extra angle of attack feedback gain that depends on the stores hanging on the aircraft, the control laws can improve the handling and flight characteristics for very heavy external stores that resulted in very aft centre of gravity positions.
Flight characteristics are then roughly the same as the characteristics of a light aircraft, if they are compared with the same lower load factor limit and angle of attack limit that applies to the heavy external store configuration aircraft.
CmalfaT = CmalfaF + (KALFDC+ KDALFD)*Cmdc + (KALFDE + KDALFD)*Cmde
The figure shows the effect of extra angle of attack feedback gain on the aircraft’s characteristics.
Without further this extra angle of attack feedback gain additions, the flight characteristics of the aircraft with heavy external stores become slow in response and sluggish (see the round pole in the imaginary/real part diagram above).
With the addition of the extra angle of attack feedback gain, the aircraft with the heavy external stores flight characteristics can come closer to the precise and quick characteristics of the light aircraft when the round pole approaches the light aircraft’s poles.
The figure below shows a maximum pilot stick command comparison between a light aircraft and an aircraft with heavy external stores with extra angle of attack feedback gain added to the control laws.
The figure shows that the heavy aircraft has roughly the same load factor response to a pilot stick command in pitch as for a light aircraft.
The extra angle of attack feedback gain KDALFD depends on the stores hanging on the pylons of the aircraft.
To determine the magnitude of the extra angle of attack feedback gain KDALFD, the stores mounted on the pylons needed to be classified with regard to weight, volume and aerodynamic effects. For this reason, AeroData codes were created for the various stores and weapons.
The extra angle of attack feedback gain addition for weapons hanging in pylon position 2 is used to compensate for the change in centre of gravity for a store hanging in pylon position 2.
For pylon position 3, the aerodynamic effect of the store in pylon position 3 is calculated. The store in pylon position 3 is also compensated for the change in centre of gravity when calculating the additional angle of attack feedback gain KDALFD.
To avoid overshooting the load factor limit and angle of attack limit when the pilot commands large pilot stick commands, there is an extra effect that increases the angle of attack feedback gain.
In addition to the implemented extra angle of attack feedback gains in the control laws, the aircraft with external stores under the wings is classified into various performance groups (PG). Each performance group has its angle of attack and load factor limit and maximum roll rate.
With the extra angle of attack feedback gain and performance groups introduced into the control laws for the flight control system, batch simulations were carried out with different stores configurations. This was done to see whether stability and handling and flight characteristics were at level 1 for heavy external stores as expected.
In addition, simulations were carried out to make sure that angle of attack, load factor and roll rate limits were not exceeded for maximum pilot stick commands and in combinations of maximum pitch, roll stick and pedal commands in specified time sequences.
In order to carry out these tests, a tool called KLAS (“Klassificering av LASter”, classification of loads) was used. KLAS was based on batch simulations with combinations of maximum pilot pitch and roll stick and pedal commands in specified time sequences. The results from using KLAS summarised how well a store configuration dealt with combined pilot commands against angle of attack, load factor and roll rate limits.
To avoid getting into an uncontrolled situation (departure), with large angle of attack and sideslip angles, a safe permissible angle of attack and sideslip angle envelope was defined.
The aircraft’s response to the pilot’s combination of maximum pilot pitch and roll stick and pedal commands had to fall within this envelope. This envelope in alpha and beta limits is called the thermos criterion.
In addition, tests were carried out to ensure that the load factor and lateral acceleration and the roll rate limits for each external store alternative were within the specified limits.
As previously shown, there soon proved to be many combinations of external store alternatives. The results of the batch simulations of the various external store alternatives were analysed, and the external store alternatives with undesirable results were then selected for flight testing.
Before the test flights were carried out, the external store alternatives with undesirable results were simulated with a pilot in STYRSIM; this made it possible to prepare difficult combination manoeuvres. One pilot manoeuvre was to be executed in both pitch, roll stick and pedal commands in a particular time sequence where the time between the commands was important to get the worst possible outcomes.
Before a test flight, a briefing was always held with the pilot and responsible personnel from various different disciplines.
The test flight was analysed in real time in flight with a parallel real-time 6-degree of freedom model for aerodata, the ROMAC simulation model. In the event of a discrepancy between the model and reality, the aerodata was corrected as needed. The results of the test flight were often better than the simulations indicated.
The test flight was monitored in a control station; data was sent by telemetry to the control station from the test aircraft with a delay of approximately 0.5 seconds. There was access to sensor data such as angle of attack, sideslip angle, speeds etc., as well as access to all control surface signals and commands and a number of internal control system variables.
As well as personnel from the flight control system, there were also personnel from other technological disciplines taking part in the work of monitoring the test. These personnel came from aerodynamics, flight mechanics, hydraulics, engine and structure.
The tests were managed by a member of test flight personnel as test leader, often with a safety pilot as support in the control station. The test leader always obtained approval from all the technological disciplines present after a test item was carried out. The pilot in the test aircraft carried out the test positions on the test chart (knäblock) after approval from the test leader.
The personnel in the control station made sure that no limits were exceeded and that the manoeuvres were carried out correctly. They also assisted if any errors occurred on the aircraft.
After a test flight for an external store alternative, a debriefing was always held, where the results were reviewed and discussed. This debriefing was attended by personnel from the various technological disciplines, together with the pilot.
This section summarises how control laws have been developed and qualified through verification and validation carried out in various simulators and through test flights.
The design phase for flight control systems is made up of four steps. First, a pre study is carried out on what is to be changed in the control laws, which may relate to an improvement or an adjustment of a problem, and is carried out by the control law designer.
After that, a preliminary design is created using the SystemBuild tool, where block diagrams and logic are produced. Using the same tool, the control system’s stability and characteristics can be studied linearly, and in that way it is possible to see the damping, frequency and stability margins of the changed control law.
From the block diagram in the SystemBuild tool, automatic code is then generated of the control laws in the control system, which are then implemented into a model of Gripen with six degrees of freedom, called ARES. The ARES model can be used to carry out non-linear digital simulations of the flight control system edition with six degrees of freedom.
ARES can also be used to carry out batch simulations for various types of manoeuvres and flight cases. This work is performed by the control law designer as part of self-verification. For the purpose of assistance, the control law designer has results from batch simulations to classify various loads. The simulations are carried out using the KLAS tool and consist of batch simulations with various types of maximum pilot pitch and roll stick and pedal commands. Various time sequences for different external store alternatives are simulated here to make sure that Gripen is carefree.
The ARES model is then implemented with the new automatic control system program code in the STYRSIM simulator as a real-time application. Here, the control law designer checks that the new edition of the control laws in the control system work as specified. A pilot is then called in to review the result together with the control law designer and to provide comments on the changes adopted and potentially propose adjustments and changes.
The last step before the new control law change is soft-frozen is to code the change in the ADA programming language. ADA is used in the actual flight control system computers in the SYSIM hardware simulators.
Soft-freezing means that the control law changes can progress to the next stage: verification.
The change is decided to be implemented on a decision-making meeting called the flight control board (FCB). Simulations with a pilot are also carried out in SYSIM to make sure the function has been transferred correctly.
When the pilot and control law designer and FCB are in agreement that a simulated change works as specified, the FCB soft-freezes the new control law change and thus also the software in the flight control system. The FCB’s decision also takes into account the result from the batch simulations mentioned above.
When the new control law change in the control system has been soft-frozen, an independent verification group takes over and carries out batch simulations and simulations with a pilot in accordance with the verification group’s own independent test schedules.
If these tests produce an outcome, the new change in the flight control system returns to the design phase and the control law designer.
In the verification work, batch simulations in ARES and simulations with a pilot in STYRSIM are used, and the verification group also uses the SYSIM hardware simulator with ADA code in the flight control system computer.
If the outcome of the verification shows that the change follows the specification and has not produced an undesirable outcome, the software – and thus also the control laws in the system – is hard-frozen.
Hard-freezing means that the control law change in the flight control system is approved for validation.
The validation group is a group of independent experts that carries out tests on the flight control system with Gripen simulations. Parts of these tests are fixed and consist of error simulations on the flight control system and other systems that is a part of Gripen. If the validation tests produce an outcome, a new edition of the flight control system must be created.
When flight testing new flight control system editions, an edition check of the change adopted is carried out, after which the test flight can continue. ROMAC, a real-time version of the ARES simulation tool, is available for assistance in the test flight.
ROMAC runs parallel to the pilot during the test flight, where the pilot’s commands and the aircraft’s speed and height are fed into ROMAC before each test point in the test flight. ROMAC is synchronised with the test flight. This allows the result of ROMAC’s parallel simulation to be compared to the actual aircraft response from the test flight. It is then easy to see any discrepancies in the aerodata basis, fix discrepancies and update the aerodata basis to achieve better and better precision in the simulations of Gripen.