Long and expensive testing cycles are a common pain point for engineering organisations, and Kautex Textron, an automotive manufacturer, is no exception.
Before employing AI, the engineers at Kautex were frustrated by the time invested and costs involved in predicting how fuel-sloshing noise behaviour impacts vehicle acoustics, which affects the comfort of drivers and passengers.
Long and expensive testing cycles can be a common pain point for engineering organisations for several reasons:
Delayed time-to-market: The longer the testing cycle, the longer it takes for a product (or car) to be released to the market. This can result in missed opportunities and lost revenue, as competitors may launch similar products earlier and expand market share.
Increased costs: Longer testing cycles can also increase costs as more time and resources are required to perform the tests and understand dynamic behaviour. This can result in higher R&D costs, which in turn impacts profit margins.
Decreased quality: When testing cycles are too long, it can become difficult to identify and fix bugs and other issues in a timely manner. This can result in a decrease in product quality, which can negatively impact customer satisfaction and lead to lost sales.
Bottlenecks: Long testing cycles can also create bottlenecks in the development process, as other teams may need to wait for model testing to be completed before they can proceed with their work. This can lead to delays and inefficiencies in the development process.
Resource constraints: Longer testing cycles can also place a strain on resources, as testing teams may need to work longer hours or hire additional staff to complete the testing. This can create additional costs and may be unsustainable in the long term.
Overall, long and expensive testing cycles can have significant impacts on an engineering organisation's ability to deliver high-quality products in a timely and efficient manner. It is important for organisations, like Kautex Textron, to continually evaluate and optimise their testing systems and processes to minimise these engineering pain points and enhance their overall product development practices.
In this article, we will explore how Kautex engineers used data-driven methods to predict fuel-sloshing noise behaviour, resulting in a significant reduction in testing iterations and costs. By training self-learning models, the team was able to better understand the relationship between fuel tank designs, testing parameters, and sloshing noise, as well as reliably predict the expected noise level of new designs from the available data.
This approach has provided Kautex with a competitive edge, allowing them to bring new and improved products to market faster and more cost-effectively, as they better understand the data at their disposal and gain insights from the knowledge accumulated.
Test engineers at Kautex Textron and similar organisations face significant challenges, especially within the classical product development cycle which is traditionally conducted in sequential steps and pre-defined stages.
During the early development stages, engineers can enjoy high degrees of freedom to make somewhat educated choices, and the cost of these design decisions is relatively low. However, as the process progresses towards later stages, the situation is reversed due to the high number of operating conditions that must be considered, as well as the various constraints, and the fact that design is more and more defined. Ultimately, this leads to less freedom of choice and less freedom to change the design cheaply.
The classical product development cycle often struggles to yield meaningful insights (knowledge and or patterns) for engineers from the available data due to the variety of parameters present, often leaving them struggling to determine the exact reasons for model failures encountered when testing. This highlights the importance of having access to the right solution and tools at the appropriate stage of the product development cycle, in order to improve efficiency and reduce money spent.
At Kautex, one specific area where improvement was required was in being able to accurately and reliably predict vehicle acoustics and fuel sloshing noise during vehicle deceleration.
Vehicle acoustics is the study of sound and vibration in vehicles, including cars, trucks, buses, and other types of vehicles. It is a multidisciplinary field that involves acoustics, mechanics, and electrical engineering, among other disciplines.
Vehicle acoustics covers a wide range of topics, including noise reduction, sound quality, vibration analysis, and acoustic comfort. It is concerned with how sounds are generated, transmitted, and perceived in a vehicle, as well as how vibrations affect the performance, safety, and comfort of the vehicle.
In practical terms, vehicle acoustics involves designing and optimising the acoustic properties of vehicles, including the design of engine and exhaust systems, cabin acoustics, and the use of noise and vibration-damping materials. It also involves measuring and analysing the acoustic and vibration properties of vehicles, including sound pressure levels, frequency spectra, and vibration characteristics.
Vehicle acoustics is an important field for automotive manufacturers and suppliers, as it can have a significant impact on the perceived quality and performance of vehicles. By optimising vehicle acoustics and structural vibrations, manufacturers can improve customer satisfaction, reduce noise and vibration-related complaints, and differentiate their products in the marketplace.
Fuel sloshing noises are sounds that occur in vehicles when fuel in the fuel tank moves around due to the motion of the vehicle. These noises are often described as sloshing or splashing sounds and can be heard from the inside of the cabin or outside the vehicle.
Fuel sloshing noises are more commonly heard in vehicles with a larger fuel tank, such as trucks and SUVs, as the fuel has more space to move around. The noise can be particularly noticeable when driving on uneven roads, accelerating or braking, or cornering.
In addition to being a nuisance, fuel sloshing noises can also be a safety concern. The noise can be distracting to the driver and make it difficult to hear other sounds such as emergency vehicles or warnings from the vehicle itself.
To reduce fuel sloshing noises, vehicle manufacturers may use various techniques such as baffles or foam inside the tank, optimising the shape and size of the fuel tank, and minimising the space between the fuel tank and the body of the vehicle. Computer simulations (such as Computational Fluid Dynamics) and physical testing can also be used to optimise the design and placement of these features.
Overall, while fuel sloshing noises are a common occurrence in vehicles, they can be reduced through careful design and testing to improve the driving experience and safety of the vehicle.
In traditional product development processes, fuel sloshing is tested by setting up a sled test where the fuel tank is allowed to run before being abruptly stopped at different fill levels, resulting in the sloshing of the fluid, replicating a decelerating tank.
Experimental test rig of Kautex. This rig is able to accelerate and brake a fuel tank on a track, simulating the impact that the movements have on a vehicle.
This test is conducted under varying conditions including different speeds, fill level, and configurations. Engineers use around three to four microphones mounted on the sled and accelerometers to measure the vehicle acoustics and monitor how the sloshing noise is transported through the air and structures.
However, this testing process is often time-consuming and costly, and the insights gained by engineers are often not fully exploited. This results in engineering organisations being unable to make the most of the data and the value they get from their tests as much as they would like and need. They are unable to reliably predict how vehicle acoustics will be impacted without a significant outlay of time and resources.
Self-learning models are AI models that can be optimised by training them on existing historical data as well as data that becomes more available over time. Engineers do not have to build new AI models from scratch every single time, as the self-learning system does this itself.
There are many benefits to utilising self-learning models in the product development process. Self-learning models are considered strong enablers for AI adoption, allowing engineers to capture and structure historical data.
Furthermore, new data can improve the self-learning models on an ongoing and systematic basis by recording the design characteristics, test conditions, and test results. Over time, the predictions made by self-learning models become much more accurate with the wealth of knowledge acquired.
Kautex Textron's innovative use of Monolith's self-learning models has revolutionised the fuel-sloshing noise prediction development process. By incorporating artificial intelligence models, Kautex's engineering team can now predict sloshing noise for new designs or configurations, such as different filling levels, saving time and money and allowing them to maintain a competitive advantage in the market. Self-learning models have given the test engineers at Kautex a way to reliably predict vehicle acoustics in different operating conditions, making better use of the data available to them.
The modelling approach in the case study of Kautex combines acoustic test data, 3D CAD fuel tank shapes, and different inner component setups. This CAD-based approach using Autoencoders helped the Kautex engineers gain new and deeper insights into their complex products as well as the creation of a toolchain that provides valuable predictions significantly faster than what state-of-the-art simulations can provide. Both 3D and tabular approaches were used to train models and make predictions of sloshing noise behaviour based on new, unseen 3D designs of fuel tanks.
With AI, engineers can reduce the time spent on testing while gaining more insights from these tests. The business impact of this time reduction is significant. Kautex engineers no longer have to wait for weeks or months for test results, as Monolith provides them in just a matter of days.
Using Monolith self-learning models, Kautex can move along the product development process faster, spending less time waiting for prototypes to be made. This means engineers can complete testing iterations quickly and benefit from in-depth learning about the vehicle acoustic behaviour of new tank designs and shapes, making data more valuable than ever.
Kautex's existing information and data can now be leveraged to access previously inaccessible information, making previously unusable data usable. In short, AI has transformed Kautex's product development process, enabling them to make better and faster decisions that drive its success in the market thanks to the lessons learned from testing.
Vehicle acoustics is just one of the various areas where self-learning models can be used to improve the product development process. Learn more about self-learning models here in our blog post and explore how they can utilise engineering companies' data to solve intractable problems such as vehicle acoustics and beyond.
In conclusion, the case of Kautex Textron demonstrates how the integration of AI and data-driven methods can revolutionise product development processes. By using self-learning models to predict fuel sloshing noise behaviour, Kautex engineers could significantly reduce testing iterations and costs, resulting in a competitive edge in the market.
The success of Kautex's approach reflects the growing importance of AI in research and development, as technical decision-makers recognise its potential to drive innovation and improve and develop customer experiences.
The results of a recent study commissioned by Microsoft shed light on the future of testing and research and development. The study surveyed 536 technical decision-makers responsible for AI, who agreed that implementing artificial intelligence is crucial to maintaining a competitive advantage in their industries. In fact:
These findings suggest that technical decision-makers understand the importance of AI and its transformative potential. Monolith is leading the charge by empowering visionary engineers with the tools they need to cut product development in half using AI.
Our vision statement is that by 2026, Monolith will empower 100,000 visionary engineers to use AI to cut their product development cycle in half. Interested in becoming one of the 100,000?
As AI becomes more accessible and easier to deploy, Monolith's vision to empower 100,000 engineers to use AI to cut product development in half represents a promising future for the field. By leveraging the power of AI, engineers can make better and faster decisions, ultimately driving the success of their businesses.
AI can help engineers with decision-making in several ways:
Predictive maintenance: AI can be used to predict when a machine or system is likely to fail, allowing engineers to schedule maintenance proactively and avoid unexpected downtime.
Simulation and modelling: AI can be used to create highly accurate simulations and models of complex systems, allowing engineers to test and optimise designs before physical prototypes are built.
Quality control: AI can be used to monitor and analyse production processes in real-time, identifying potential issues and allowing engineers to make adjustments before defects occur.
Optimisation: AI can be used to optimise designs and processes, using algorithms to search for the best possible solutions based on a wide range of parameters and constraints.
Decision support: AI can be used to provide engineers with real-time data and insights, allowing them to make more informed decisions and respond quickly to changing conditions.
Natural language processing: Natural language processing (NLP) is a branch of AI that deals with the interaction between computers and human language. It involves the development of algorithms and computational models that enable computers to understand, interpret, and generate natural language. AI can be used to analyse unstructured data such as customer feedback and support tickets, providing engineers with insights on how to improve products and services.
In summary, AI can help engineers make better decisions by providing them with insights and recommendations based on data analysis, simulations, and predictive modelling. This can help engineers optimise designs and processes, improve product quality, and reduce costs and downtime.
Learn more about how AI helped engineers at Kautex Textron improve vehicle acoustics in this webinar.
This webinar showcases how the team at Kautex Textron was able to utilise Monolith’s machine learning capabilities to produce accurate AI models for predicting car fuel sloshing behaviour. The methods developed helped engineers gain new and deeper insights into their complex products as well as create a toolchain solution that provides valuable predictions significantly faster than what state-of-the-art simulation tools such as computational fluid dynamics (CFD) can provide.
In the session, our team and Kautex focus primarily on cost and time-critical test applications, as well as data-driven self-learning models that are increasing ROI, without disrupting existing engineering workflows or systems.