From Aircraft Health Monitoring to Aircraft Health Management

Do you remember the iconic movie: 2001: Space Odyssey? It is a 1968?Do you remember the iconic movie: 2001: Space Odyssey? It is a 1968 epic science fiction film produced and directed by Stanley Kubrick. The screenplay was written by Kubrick and science fiction author Arthur C. Clarke, and was inspired by Clarke's 1951 short story "The Sentinel". In that movie the American spacecraft Discovery One is bound for Jupiter, with mission pilots and scientists Dr. Dave Bowman and Dr. Frank Poole on board, along with three other scientists in hibernation. Most of Discovery's operations are controlled by HAL, a HAL 9000 computer with a human personality. We are already talking about AI in 1951. At certain moment HAL said to astronaut Dave: “Dave, High gain antenna control device is going to fail in 48 hours!”.

Dave retrieves it in an extravehicular activity (EVA) pod, but finds nothing wrong. HAL suggests reinstalling the device and letting it fail so the problem can be verified.?

The movie was released in April 1968. Remember that date. I will come back to this later.

Aviation is complicated business. Besides all operational stuff, there is a lot of technique deployed. To make it easier. The aircraft is divided into systems. Generally speaking, the brain of each system is LRU (Line Replaceable Unit). Let’s assume that unit fails during the flight. In present airplane there will be a massage at one of the displays announcing to pilot that certain unit or function failed. Pilot will insert the entry in maintenance log and hand over the airplane to ground crew upon arrival. The technician will pool the unit from the aircraft and send it to the store handling unserviceable parts. He will order the new LRU from serviceable store, install it in the aircraft, accomplish the functional test and release the aircraft for service.?

The unserviceable unit will arrive to the store together with the whole paperwork. On one of the documents there will be the original complaint issued by the pilot. The unit will than be send to repair shop. The shop technician will receive the unit, repair and test it and release it with serviceable tag certifying that the unit is airworthy. The unit is than send to serviceable store and set on the shelf until next request. The circuit is closed. That described above is a pool concept. It is much more complicate, but the idea is the same.?

The interesting part is the system called health monitoring. It is introduced with introduction of 747-400 aircraft. That was one of the first sophisticated aircraft Internal systems in the aircraft were monitoring at least 20.000 parameters. Practically all testing of many systems could be done from the cockpit. There is also a CMC (Central Maintenance Computer) installed on board and real time monitoring of the systems is done during the flight. If an LRU fails, the AICAS massage appears on the screen in the cockpit and audio sound is heard from the speaker. An engine-indicating and crew-alerting system (EICAS) is an integrated system used in modern aircraft to provide aircraft flight crew with instrumentation and crew annunciations for aircraft engines and other systems. During the flight the crew could be alerted if there is any of parameters out of limits or if there is a failure in the system.?

Crew will than troubleshoot the system and isolate the fault. They will also insert the entry in maintenance log. Some airlines have the procedure that crew sends the ACARS massage to the operational center providing them with the list of failures during that particular flight. Few years later when SATCOM system was installed in the aircraft, the crew used to call Operation center two hours before landing and verbally report the failures (if any). Operational center analyzes the complaint and issue the advice to maintenance department and get them prepared and ready when aircraft arrives. In most of situations the technician would be at the arrival gate with LRU in his hands. When aircraft arrives, he would immediately replace the unit (LRU), accomplish the test, and release the aircraft. That operation can go very smooth. But there are some disadvantages/risks. It is possible that there is no serviceable LRU and aircraft can’t be repaired immediately. The consequence could be a delay and AOG situation. Meaning that the part should be borrowed from another aircraft or from another operator (high costs and logistic problems). Sometimes, it is decided to deactivate the system and temporary dispatch the aircraft with all additional risks (additional delays in case of failure). Sometimes it is just a delay in replacing part due to no time between two flights. It is not unusual that the repair is diverted to next A check due to no time or due to no material. That all is introducing a lot of organizational and financial problems.?

That is only operational part. Behind the scenes there is much more. Here is a department measuring all removals and providing the statistics. The department will indicate which parts are bad actors and set engineers to work to design modifications to improve LRU’s. There is and MPD (maintenance planning document) which provides the whole maintenance pattern to optimize the maintenance and keep the LRU’s in top condition and in the aircraft. Every removal is costly. Therefore, everybody is working hard to make LRUs more reliable and limit number of removals and failures.?

The Statistic coming from statistic department also indicate NFF (No Failure founds). Some call it FNF (Failure not Found). That is very painful phenomenon. What is NFF? The test in the aircraft is showing thet LRU is broken (not performing well) but when the same unit is tested in the shop at the test set using CMM (Component Maintenance Manual) the unit is performing well, and it is certified serviceable and airworthy. If the unit is installed in the aircraft, it can/might fail again. The process is repeatable. Sometimes, 40% or 80% of the units are getting that terrible label NFF. The process is very costly because the technicians in hangar are removing units, which are tested and send back without any maintenance action. Calculate the costs? The costs are very high, and many people are involved sometimes for weeks or month until final resolution to the problem.?

To make that complicate story short, the following.?

The condition of each component is monitored. When component fails, it is reported, and the appropriate action will take place to replace faulty unit. It is important to understand that we are passively waiting that unit fails and that we start the corrective action.?

As you can see it is complicated business. If you are outsider, you would get lost in all procedures and details. No wander that airlines are frustrated about maintenance costs. The maintenance actions are most of time unexpected and difficult to plan. The failure can occur any time, but it also occurs in the worst possible and at the most unexpected component. It can cause delay, AOG or cancellation of the flight with sometimes high costs. And it is almost impossible to plan. The process can’t be made proactive but due to many interventions of engineers, we managed to make the LRU’s reliable, and they are not failing so often. In good old times the LRU’s were flying around 5000 flight hours without failure on average. Good engineering work improved that number and the LRU’s are nowadays flying 80.000 or 100.000 flight hours without failure. That is great improvement but still not predictable and the chance to get a failure is always there.????

The described situation is already 50 years normal operation. Why? Because the technique was not sophisticated enough to do it different. And still, we are not mastering the situation. Fortunately, there is some development. We could say that we are at the edge of new era called aircraft health management. The process includes both “on-board” and “off-board” sequences. Here are the elements of process: Data is collected from the aircraft and can improve the levels of safety and efficiency which can increase aircraft on-time performance and improved dispatch reliability. It can introduce a significant cost reduction for the operator (planned removals). The process should be data driven and without warm body interface or intervention. I call intervention of people warm body interface.??

Health management requires that a lot of data should be collected during the flight. I mentioned 20.000 parameters on 747-400. That should be 10 times more on modern aircraft and it is. 787 is generating 500 GB of data per flight. And that is a lot. One part of process is to sense and collect all the data during the flight. The data can be packed and filtered and send to the ground systems. It can be sent via Hi speed link or via gate link. Immediately after the data is received the algorithms supposed to run to filter and analyze the data and provide the indication about health of LRU’s in the aircraft.?

Here is the moment of truth.?

There is a lot of data but almost no algorithms. So, the data is useless. Some airlines are claiming that 95% of data is just thrown away. Obviously that we have a big, big problem. Thousands of flights every day and terabytes of data generated every day and practically no algorithms and no data analyses. Basically, we are falling in old situation called health monitoring.??

Just to set the stage before we go further. In software eco system we have zero trust situation. Obviously, many parties are involved, and they don’t trust each other and also, they don’t want to share the data afraid that somebody else will use it to make money. The whole army of engineers are ready to develop algorithms, but they need a lot of data, big data. Data owners are stressed under management pressure. Management of airlines is correct when they say it is our data, but they also prevent any data exchange because they are afraid that data could be used also for other purpose (generate money for others). Engineers would like to analyze the big data and develop algorithms to predict failures. They need a lot of data, but they can’t get it. Even if they get the data the process of developing algorithm is not easy, and it is often very long. It takes a lot of time, and they need a steady stream if data. The algorithm should learn itself to analyze the data and come to the correct conclusions. Big data is a collection of data from various flights in various circumstances and in various time. It is all variable (I hate variables!).?

Because of unavailability of data, engineers even came to idea to send algorithm to the location of operator and analyze the data locally and only provide the results. Can you imagine that you purchase the algorithm from certain OEM. You keep it in house and analyze your data and get the conclusions. In that case you don’t provide the data to OEM. Here is again the trust issue. Would their algorithm perform on your data. It makes sense to question it.

It that new environment we are not talking any more about aircraft health monitoring. We are talking about aircraft health management. Her e is the definition:??

Aircraft Health Management (AHM) is the unified capability of using health monitoring of aircraft structure and systems (including propulsion system) to control the scheduling of aircraft needed maintenance actions; could be resumed to the process stages of Sense, Acquire, Transfer, Analyze and Act (SATAA).?

Now the short version of new world: big data generating by the aircraft is analyzed and the algorithm calculated that certain component will fail very soon. The notification, which component and on which aircraft will fail is send to maintenance department and they replace the component before actual failure.

Here are some remarks about Health Management:?

? It can be planned when the most convenient time is to replace the component.

? You don’t need the component on stock waiting month for to be used. You order it on demand. Saving of spares. (New spare part calculation is needed!)

? The repairs are cheaper because there is no collateral damage.

Note: If 787 generator fails due to bearing failure, it happens sometimes that all internal parts are broken because of heavy forces. Sometimes is very expensive damage because the whole explosion took place inside the chassis. But if you detect that the bearing is degrading, you just remove the generator at next convenient moment and just replace the bearing before it causes damage of the whole interior.????

? Some processing can be done in the aircraft real time during the flight.?

? NFF (No Failure Found) should be redefined.

? We need standards for data exchange, zero tolerance and algorithm development.

? There is urgent need for MSG-4 analyses. (All present activities are based on MSG-3 which is not sufficient for Health Management).

Back to health management. In the ideal world, we could collect the data from all available aircraft. Then split the work between software engineering groups which will analyses the data. Say, each airline, OEM or engineering team will develop one, two or three algorithms based on their capability. The rule is that each develop different algorithm. In a year or two they can develop the majority of algorithms and made them available to all others. The idea is to develop all required algorithms in one or two years and cover the whole flying fleet.?

Of course, that I understand that this will not happen. It is utopia. But with some luck we can come very close to ideal situation. We could for example organize the data marketplace. (Multi-Sector - About ExchangeWell (sae-itc.com))

The algorithm developers could open the digital booth and sell their algorithm to others. Also, they could go shopping and visit the booth of other algorithm developer. There might be something they need.??

The farmers also have market and thet are selling their stuff: tomatoes, oranges, and pumpkins. There is no reason that Software engineers can’t open the booth and sell their products on the software market.?

If we vanquish all obstacles, we might land in the situation Dave Bauman had on the way to Jupiter when HAL9000 said: Dave, High gain antenna control device is going to fail in 48 hours!”. Well on our case it would be a routine flight from Amsterdam to LAX and board computer could say to pilot: “Captain, when you return to Amsterdam again, the Flight Control Computer #3 should be replaced because it will fail in 3 days. “??