On May 11, 1997, an IBM supercomputer named Deep Blue defeated world chess champion Garry Kasparov in a six game chess match. This was a historic moment for supercomputing. A super computer beat a world champion human player for the first time in history. Deep Blue’s victory over a human signaled a shift in technology. Advances in complex computer models and parallel processing have paved the way for today’s computer processing power. Today this technology is used in almost every aspect of our daily lives.
Processing Advancements
Deep Blue used brute force in calculating large numbers of moves. This allowed it to determine how different moves could affect the outcome of the game. By 1997, Deep Blue’s algorithms could calculate up to 200,000,000 moves per second using 30 parallel processors. It’s parallel processing and complex algorithms could uncover hidden patterns in massive datasets. These advancements inspired further development of computer learning algorithms. Improved algorithms minimized searches and made processing more efficient and powerful. IBM demonstrated this level of technology through development of a machine named Watson. In 2011, this supercomputer took on two of the all-time most successful human players of the game show Jeopardy! and won. Watson could process and reason about natural language which demonstrated a whole new generation of human-machine interactions.
Human and Computer Partnership
It seems computers maybe chess (and maybe game show) champions of the world from this point forward. However, it’s actually not that simple, and we haven’t even delved into airplane design and certification yet. Kasparov could have just accepted the idea that computers surpass human performance. Instead, Kasparov started developing the concept of the human and computer partnership. He wanted to show that combining human with machine could push performance even further. In the world of chess, Kasparov invented the term “advanced chess”. This is where human players utilize a companion computer to calculate moves. The combination of human intuition and computer data processing power provides a partnership which is mutually beneficial. The combination boosts performance beyond what either is capable alone.
An Aircraft Designer’s Perfect Companion
Aircraft design and certification organizations have been using a human/computer partnership for decades. Aircraft design is a multidimensional problem with a massive number of variables and boundary conditions. The entire design process utilizes sophisticated software from CAD/CAM technology to analytical modeling such as Finite Element solutions. Optimization engines have also become very sophisticated allowing analysts to setup optimizations across technical disciplines (aerodynamics, structures, etc.). This is done in an attempt to find the most efficient design solutions. As sophisticated as computers have become, aircraft design still requires a human partner to guide the enormously complex compromises.
Design for Certification
It’s challenging to design aircraft to meet performance, weight, and cost targets. Add certification and it becomes a monumental task. Certification cannot be an after thought. Aircraft must be designed for certification from the very beginning. Addressing all of the certification considerations during the design process requires intimate knowledge of the massive regulatory and guidance libraries of the FAA (and any validating agencies in other countries). A great deal of certification work is also done through the use of industry standards. These are produced by Standards Development Organizations (SDOs) such as SAE, ASTM, RTCA, etc. SDOs further broaden the required knowledge base for certification. All of this requires human understanding of the enormous library of compliance information and how to apply it during the design process.
Requirements Buildup
Every certification project requires a complete evaluation of the design relative to the certification requirements. That sounds relatively straight forward. However, just defining the requirements alone could be a significant task if there are new and novel features. The requirements “build up” process can take a significant amount of time for a certification project. The level of safety set by the regulators requires human intelligence to understand. In some cases, regulations don’t even exist to address the design. In cases like this, it’s the applicant’s responsibility to fully brief the certifying agency on the design features. This initiates discussions regarding possible special conditions that may be required. Technology like IBM’s Watson isn’t going to be able to evaluate a new and novel design and develop a set of requirements for certification.
Findings of Compliance
Can a computer make a finding of compliance? In a previous post we discussed ‘what exactly is a compliance finding‘. A compliance finding is a verification that the engineering data presented by the applicant complies with the applicable airworthiness standards (requirements). It’s essentially a process by which either the FAA or their designee reviews the compliance data and makes a compliance determination. That means a computer would have to somehow process the design information and compare that with the requirements. Often, the method of compliance used is a “design review”. The FAA or their designee does this by reviewing drawings to identify the design features which comply with the requirements. It’s hard to imagine a computer having the capability to process such information. The multitude of different ways to create a compliant design is far too great.
An Essential Partnership
Artificial intelligence may not generate an FAA special condition or make a finding of compliance, but it can be an outstanding assistant. Computers have the ability to analyze massive datasets and find hidden patterns which humans couldn’t possibly do manually. Along with intuition, humans understand the advantages and limitations of computer algorithms. Human reasoning and decision making partnered with the processing power of computer algorithms creates an essential partnership for aircraft design and certification. While more and more computers are essential, aircraft certification will continue to be a human endeavor.
Joel Heck has been with Cessna, now Textron Aviation, for 24 years in engineering. His entire career has been focused on the design approval aspects of FAA certification, and his technical work experience in certification spans the Cessna, Beechcraft, and McCauley product lines.