AI & its Bearing on the Future of Test
03 December 2023
In any manufacturing or lab environment, there’s measurement hardware and systems, plus related software, all in operation. Through such items, sequenced tests can be actioned. These tests may then indicate whether a product - like a motor, power inverter, battery pack or wafer - is successfully meeting (or conversely deviating from) the predefined specs.
However, all these tests are running in silos, and there might be hundreds of them being conducted within sites around the world. The ability to draw all this data together using artificial intelligence (AI) is going to be at the heart of test activities moving forwards.
With connectivity becoming ubiquitous at high bandwidth and a low price, if you learn about an anomaly from one machine and understand its root cause, you can apply this knowledge across multiple systems. This is like Google Maps: when you hit traffic congestion, the cloud-based system reroutes you automatically through an autonomous mesh of connected systems. Similarly, we believe that the future of test is not an instrument, but an autonomous hyper-automated amalgamation of systems that brings together hardware, software, data, procedural workflows and intelligence. This will contribute to the empowering of a connected future.
One key factor that this ‘system of systems’ hinges upon is emerging technologies - and their ability to accelerate transformation. Technologies like AI will enable us to reimagine how test and measurement work is executed, so that improved business outcomes can be derived.
The potential of AI in test
Use of AI in test is opening up new opportunities in many areas. These include the semiconductor, automotive, telecoms, aerospace/defence and education sectors. Automotive presents an interesting challenge for test, as we move towards higher levels of autonomous driving. In the past, we typically relied on stimulus/response-based testing. With embedded AI now a reality in many products, scenario-based testing is proving crucial for success. As a result, it is necessary to identify an infinite number of complex scenarios (both planned and unplanned) that a car might end up in. These then need to be tested before the car is safe to operate within the real world.
Another example is 6G communications - the next generation of wireless telecoms, which promises ultra-high data rates above 100Gbps, plus end-to-end latency delays of less than 1ms. These capabilities will enable computing to become much more immersive and revolutionise machine-to-machine (M2M) communication. With the advent of 6G, we will also need a standardised way to collect, manage and store data. NI is working with other organisations to create this - including open sourcing a platform for researchers to record and test data.
Once test data has been collected, it is possible to create scenarios to investigate. For instance, with 6G the potential interference scenarios can be considered, and how to organise everything so that communications are more effective and efficient. Likewise, how to cope with the heightened data demands of a stadium or an airport can be looked into.
In order to detect anomalies, we need to look more at how things change over time, and in particular the variations in usage characteristics and the load on the system. Those are dimensions we haven’t really looked at traditionally in test, but they are familiar concepts from software systems and in relation to everyday life - remember how cellular networks used to jam up when everyone tried to send SMS messages at midnight on New Year’s Eve! In this example, the density of people in one place changes the behavioural characteristics of systems. Those characteristics appear in batches whenever that density increases, and when you look over time at the factors that lead to anomalies, there is a degradation pattern.
Data and complexity
With effectively infinite scenarios, it isn’t possible to test them all. So, can we use an AI engine to test the AI itself? That is another area which NI is beginning to research.
Figure 1: The future of test is an autonomous hyper-automated system of systems
Every day, systems are getting much more complex. There are a greater number of moving parts coming together to make such systems run, with more sub-modules that get assembled in different permutations and combinations. That is why the opportunities for test and measurement continue to grow. And even though we’re working with much more data, the idea is not to store it all. Instead, we need to find the key patterns that lead to an anomaly and aggregate the relevant data. We can then cold store it and compress it, or archive or even purge it as needed. You don’t need every line item of data. The trick is figuring out what is the key aggregate to store, and conversely what may be thrown away.
Challenges for AI-based testing
The big challenge for AI, particularly in the 6G space, will be getting the data and creating a critical mass of scenarios which are actually meaningful. That is going to be the first big challenge, and it is something we have to work out with our design partners - namely the carriers, the semiconductor manufacturers, the device vendors, etc.
The second part is how to secure and protect this data, and do it at a price point that is consumable. The cost of computing should not outweigh the business value that comes from it. How to traverse these two, carefully and delicately, is something that needs to be figured out. There is a lot of learning to be done in the next 12 to 18 months, as these capabilities begin to converge.
AI at the heart of test
NI has a long history of investments in advanced analytics, which has led to innovations being seen with regard to AI. For example, NVIDIA has been using NI’s advanced analytics software to manage multiple contract manufacturers around the world.
The end goal of AI in test is to increase resiliency, with the capabilities derived helping our customers to achieve faster time to market with fewer failures. To achieve this, we need a network of scenarios that can cover anything, and we need to first virtualise the test instrumentation - that is how we will democratise test and measurement as a next logical evolution.
The future of test is not an instrument, or just one system working in isolation. It is an autonomous, hyper-automated system of systems, with AI at placed right at its centre. We cannot lose sight of the end goal - all of this technology working together to achieve the output of excellent business or product performance for the end customer.
Contact Details and Archive...