Artificial Intelligence and Deep Learning in OEM Service Operations
Twice each year Cognitran hosts the ITIS User Group meetings. These provide an opportunity to update OEM representatives on new technologies being deployed in ITIS to deliver enhanced functionality and security.
At the recent summer 2017 event CEO Paul Goulbourn outlined how Cognitran is implementing Artificial Intelligence (AI) to safeguard our clients’ intellectual property and improve the customer experience. The presentation prompted further discussions on the opportunities AI and deep learning could offer for OEM service operations.
Investment in Artificial Intelligence (AI) and deep learning is growing rapidly, with potential applications across every business sector. Globally, investment in AI is forecasted to exceed $35bn by 2025.
Investment in Artificial Intelligence (AI) and deep learning is growing rapidly As Paul explained:
AI is an active area of research and development for many OEMs. Most of the attention is around autonomous vehicles and how these can operate safely on crowded and often erratically planned traffic systems.There’s also enormous potential to apply emerging AI and deep learning techniques to improve efficiency, security and the customer experience in OEM aftermarket operations.
Lower Cost of Entry
Within the last year to 18 months the cost of entry to AI has tumbled thanks to the development of new tools. This has facilitated a new focus on practical applications with the potential to deliver real business value at a reasonable cost.
We’re already implementing deep learning methodologies to detect anomalies in user behaviour that might indicate attempts to compromise security or intellectual property.
No two users will use an information repository in exactly the same way. However, by analysing the behaviour of a large number of users it’s possible to build mathematical models of characteristic or normal behaviour.Monitoring a number of parameters you can build a normal distribution model of expected behaviour. Anomalous behaviour at the extremes of the distribution curve could be a vital indicator that a user is not accessing the information for a ‘normal’ purpose.
The types of behaviour indicated could include an attempt to download large sections of repair information with the intention of selling the data. Anomalous behaviour could also identify brute force attempts to compromise security algorithms or key programming codes.
The advantage of AI is that it doesn’t require you to pre-empt what every fraudulent user might attempt to do. You aren’t looking for a specific pattern of behaviour, just anything that deviates significantly from normal observed behaviour.
The Aftermarket Customer Experience
Looking at the customer experience of OEM service operations there are a number of ways AI can be applied. Among the most straightforward applications involves the booking in process for vehicles at a service department. Scanning a number plate or VIN could automate and fast track much of a process that customers can find frustrating at busy times. As Paul points out:
The experience can be enriched if the system is able to ‘learn’ customer preferences. Service advisers can automatically have all relevant vehicle and customer details to hand, without a customer having to repeat details they have probably already provided.
Deep learning can also improve the delivery and use of diagnostic information; potentially reducing warranty costs as well as avoiding customer dissatisfaction resulting from misdiagnosis.
Traditional diagnostic routines are highly prescriptive. They often dictate a fixed sequence of tests to be conducted in a defined order. In practice, the routines are rarely revisited to take account of known common problems or to eliminate or downgrade diagnostic tests that almost never identify the fault.
Deep learning can mine diagnostic and repair data to identify ‘best test’ options based on the vehicle specification and age. By focusing on most likely causes and eliminating irrelevant tests it would be possible to reduce warranty costs, improve diagnostic accuracy and accelerate the time taken to fix faults.
The ITIS detailed repair data is used for this analysis.
One of the most exciting applications is in the area of prognostics: identifying characteristic factors in the service and repair histories of OEM equipment that could indicate a high probability of an expensive repair or a major breakdown.
Paul explained the thinking behind this:
It’s likely that there will be a history of fault codes, none of which appears significant on its own and may not even produce noticeable symptoms. These could, however, be a characteristic ‘fingerprint’ of seemingly minor faults that culminate in a major failure. Analysing fault code data and repair histories builds the knowledge needed to pre-empt major and expensive repairs. In the world of commercial vehicles this would support the ‘zero downtime’ goal. OEMs armed with this knowledge could have a significant competitive advantage.
Paul concluded with this thought:
The application of AI is one of the most exciting areas of development across many automotive and OEM businesses. Forward thinking organisations like ours will remain at the forefront of applying these technologies to deliver tangible business benefits.