OPINION: How Artificial Intelligence Eliminates Risk from Aerospace and Defense Quality Teams

April 15, 2024

The complexity of aerospace and defense (A&D) supply chains has outstripped the capabilities of human processing.

An A380 comprises 4 million parts, distributed across over 10,000 suppliers. The challenge for Original Equipment Manufacturers (OEMs) is how to source this large number of parts whilst maintaining key quality and performance metrics.

Notionally, large manufacturers have strong supply chain management systems in place. There are three primary methods:

1. External certification, such as AS9100.
2. First Article Inspection (FAI).
3. Source inspection.

The problem is that none of these provide real-time, granular data on supplier processes. External certificates and source inspectors offer snapshots; and FAI is not process-based.

As a benchmark, let us take the data quality of financial investigators. If regulators suspect an individual of insider trading, they have the capacity to track the specific trade, the specific time, the communication with other members, and so on. In other words, financial investigators have extremely granular data on the trades which individuals execute - at all times.

Compare that to aerospace and defense regulators. Carnegie’s internal statistics from our user data estimate that over 65 percent of quality escapes go unnoticed. That means that poor quality products float through the supply chain and no-one can spot them. The FAA believes 2 percent of parts on commercial airliners are counterfeit. Who knows what percentage of those parts are of sub-optimal quality.

And we cannot trace these quality breakdowns.

The Root Cause

Why does this discrepancy between financial data and aerospace and defense quality data occur?

The majority of aerospace and defense quality processes are paper-based - especially at smaller Tier 3s and 2s. The implication is that we cannot track data as it passes through the system.

And why are the majority of A&D quality processes paper-based?

Unlike aerospace and defense hardware — which sits on the very forefront of human capability — A&D software is outdated. For context, in the same period of time between the launch of SAP’s R/3 ERP in 1992 and today, we went from the Wright Flyer to the Spitfire. From cotton-muslin gliders to

war-winning fighters in less than the time it takes for the development of novel A&D ERP software.

Without good operational software, quality teams cannot move away from paper-based work. And we - as an industry - cannot capture critical data that would otherwise eliminate human error, produce better-quality products, and increase the rate of innovation.

The next question: why is A&D software so antiquated? At Carnegie, we believe the answer is threefold:

• High switching
• Long-term
• Development

In other words, for a manufacturer to switch ERP takes a huge amount of effort - often shutting down production for a month or more. However, the payback period is long (amortizing over many years), which is misaligned with CEO compensation incentives. Finally, in order to build A&D operational software, you need high-quality data, which often means implementing an entirely new Manufacturing Execution System (MES).

So, this is the root cause behind supply chain inefficiency in A&D. High switching costs; misalignment of timelines with CEO incentives; development dependencies.

The Role of Artificial Intelligence

The rise of AI will fundamentally reshape the future of A&D manufacturing. When you think of AI, it is useful to consider it as an alternative to human beings. Unlike previous software tools which can automate specific tasks, AI can “think” like a human being and eliminate entire jobs.

Take the work of a quality liaison manager. That person will spend 90 percent of their time on a series of manual, repetitive tasks. However, because the input variables for those tasks — i.e. the data, or the communication — span multiple channels and platforms, previous software was not able to automate the role.

Now, however, AI can “think” and “understand”. AI can understand the context of a given activity by that liaison manager and intelligently automate it. At Carnegie, we achieve this through our fine-tuned AI models. Unlike relying on ChatGPT — which is trained on a diverse dataset found across the internet — we use a variety of proprietary and public A&D data sets to fine-tune AI models. This means that the output of our automation is always specific to A&D manufacturers. The result is a higher-level of performance - and fewer errors.

Over time, as we receive more data, we can further improve the performance of our fine-tuned models. In particular, we can use the data from a single manufacturer - or single facility - for fine-tuning. In this instance, the output of our AI models is specific to a unique use case. This provides optimum performance.

In the future, quality managers will never perform manual, repetitive tasks. They will move from task monkeys to intelligent optimizers. Rather than wasting their time on manual, repetitive tasks which have a high human error rate, quality managers will focus their energies on strategic tasks. The result is the immediate identification and rectification of quality breakdowns across the supply chain.

Future

Real-Time Monitoring and Traceability

One critical advantage of AI in A&D quality management is the capability for real-time monitoring and traceability. Through the integration of Internet of Things (IoT) devices and sensors across the supply chain, AI systems can track the movement, condition, and quality of parts in real-time.

This level of visibility ensures that any deviations from the required quality standards are identified and addressed immediately, dramatically reducing the risk of counterfeit parts entering the supply chain. In addition, this real-time data collection feeds into AI models that continuously learn and improve, enhancing the overall robustness of the quality management system.

Automated Compliance and Reporting

Traditional methods of ensuring compliance are often cumbersome and prone to human error. AI, however, can automate much of the compliance monitoring and reporting process. By constantly analyzing data against regulatory requirements and standards, AI systems can flag non-compliance issues instantly, ensuring that they are addressed before they escalate. In addition, AI can automate the generation of compliance reports, reducing the administrative burden on quality teams and allowing them to focus on strategic tasks.

AI Predictive Analytics

The depth and breadth of insights provided by AI extend beyond operational efficiency and risk mitigation. AI's ability to analyze complex data sets offers quality teams and decision-makers unprecedented insights into the performance and reliability of suppliers and parts. These insights empower OEMs to make informed decisions regarding supplier selection, quality control measures, and strategic planning. As a result, AI not only enhances the immediate quality management processes but also contributes to the strategic positioning of A&D manufacturers in a competitive marketplace.

Conclusion

The introduction of AI into the A&D sector's quality management processes represents a pivotal shift towards more efficient, accurate, and reliable manufacturing operations. By using AI's capabilities for predictive analytics, real-time monitoring, automated compliance, and enhanced decision-making, A&D manufacturers can significantly reduce risks associated with quality control. This not only ensures the safety and reliability of A&D products but also positions manufacturers for success in an increasingly complex and competitive global market.

As we look to the future, the role of AI in transforming A&D quality management is not just promising; it is essential for the continued advancement and innovation in the industry.

This article was written by George Bradshaw, CEO and Co-founder of Carnegie Aerospace. For more information, visit www.carnegieaero.com .

---