As commonly used, the term “big data” refers to a collection of data sets too large and complex to process using on-hand database management tools or traditional data processing applications. The challenges of big data include capture, curation, storage, search, sharing, transfer, analysis, and visualization. The trend to larger data sets results from the additional information derivable from analysis of a single, large set of related data, as compared to separate, smaller sets with the same total amount of data. Big data supports making correlations that facilitate the clear and rapid understanding of trends, which enables better and timely decisionmaking.
In the automotive sector, manufacturers are finding that the key to leveraging big data is contextualizing that data to make it actionable on the factory floor. “First and foremost, big data is not very valuable without applying contextual analytics. Without contextualization it would be like driving on a one week vacation to a place you haven’t been before without a GPS, map and road signs. If you eventually arrive you find you only have one day to enjoy before heading back home. For manufacturers this is lost opportunity due to the lack of speed in quality decision making and could be the difference between profitability or losses.” says John Billings, Vice President & Head of Automotive USA for Siemens.
Contextualization is about providing perspective (e.g., from the customer, product developer, manufacturing engineer) across the enterprise to create a closed-loop learning and adaptive ecosystem that can be used anywhere in the product lifecycle. This is particularly valuable in the automotive sector, where surprisingly rapid and often volatile shifts in how customers perceive vehicles and the driving experience itself have forced automakers to rethink everything from materials to marketing to manufacturing to maximize efficiencies, revenue, and profits.
The Right Combination
In the manufacturing context, big data is data generated, exchanged, and integrated between the shop floor and the top floor of an enterprise. What is making data really “big” in the automotive sector is the increased use of virtualization and simulation, as well as the growing trend of crowdsourcing and social manufacturing (i.e., leveraging social media-generated data in the design and manufacturing process). This development is moving automotive manufacturing towards what can be called the digital enterprise. In the digital enterprise, data connects manufacturing to systems such as PLM, SCM, and MES; these systems are integrated into ERP.
From Siemens’ perspective, one of the keys to successfully leveraging big data is to take advantage of both virtual and real data, and combine them. A tremendous amount of “virtual” data is created in preparation for both the building and manufacturing of a vehicle, such as modeling for product development or process planning for manufacturing operations. Then there’s the “real” data, what’s generated on the shop floor by sensors, PLCs, and so on. Uniting the virtual and real data is key to enabling a learning and adaptive manufacturing enterprise. Thus ensuring product quality, as well as understanding the requirements early in the design process versus later in manufacturing.
The benefits of this union are significant. During early planning, leveraging data that has been collected over time in the “real” world allows manufacturers to see the downstream impact of decisions made in the design process. Because data is created while product is produced on the plant floor, it can be compared in real time to the data prepared in the planning stages to understand if requirements are being met properly. The result can be huge improvements for automotive manufacturing, in productivity, efficiency, flexibility, quality, speed-to-market, and more.
“This approach allows automotive manufacturers to transcend the gaps between the product, process, and design functions by closing the loop between them and manufacturing per se,” says Billings. It leads to an adaptive and self-teaching ecosystem that can inform and improve real-world processes on the auto plant floor.
Applications that Work
Billings points to a number of areas where automakers are making significant strides in the use of big data, including quality, operational efficiency, and improved product launch. First example: the paint shop. Today automakers are routinely leveraging big data to determine the causes of paint variance and improve the quality of paint jobs. Data is analyzed to consider all parameters that could affect the quality of paint and painting: suppliers, environmental factors, process parameters, shift conditions, and so on. “They take this broad spectrum of data and use it to learn and understand what factors support or diminish quality in work previously done, then apply those lessons to ensure best quality moving forward,” notes Billings.
Another key area where automakers are leveraging big data is in operational efficiency: using analytics to understand how they can run existing plants at ever-higher utilization rates. “Automakers are not looking to open up new plants; therefore, the knowledge big data can deliver in terms of how to optimize the use of existing assets is key,” says Billings. He shares that many automotive manufacturing plants are running at or above 100 percent capacity utilization already, and virtually all are running at over 80 percent. Big data is being used to drive greater efficiency by providing better insight into operational problems, their root causes, what can be done to reduce mean-time-to-repair rates, and capturing all that is learned so that it is in place to speed response should a failure occur in the future.
Using big data to enable predictive launch is an increasingly important application in the sector. According to Crain’s Detroit Business, the North American automotive industry is preparing for 32 new and remodeled vehicle launches in 2014. Consequently, manufacturers and their suppliers are finding ways to collect and analyze big data to improve planning, streamline product development, and meet the industry’s increasingly aggressive launch schedules.
An important tactic being adopted is what Billings calls “predictive launch.” Say a manufacturer wants to start up on a specific date. The goal is full production, 75 units an hour. What needs to be in place? The product.
The product moves into design for manufacturing. (Can we build it?) Then it goes to line builders. (This is what is needed to build.) Continued product changes impact the line build, which makes it difficult to meet timing, budget and functionality. How do the line builders ensure that the line can build 75 jobs an hour by the target date? What are the many parameters that come to bear to make that a reality? What did we learn from the last launch? Such questions are posed across all functions (contextual knowledge). What did design learn from the product side; how is that captured and applied moving forward? Data across the ecosystem is analyzed as a means to visualize and understand the launch before it happens, greatly improving the likelihood that launch objectives will be met.
One element that is leveraged to help address some of these questions is to simulate the production line. For example, an automotive company has a four week window to rip out an old bodyshop, then install and startup a new bodyshop. To address these constraints they are simulating the production line and its automation before its installed. This will help with the mechanical installers to make sure they are installing correctly and with the automation engineers to ramp quickly to full production.
The Importance of Integration
Discrete manufacturing possesses more stored data than all other industries in the United States, including government, healthcare, and communications sectors. Moreover, when compared to other sectors, the gains available to manufacturers such as automakers through analysis of readily accessible data are significant and relatively easy to secure.
The low-hanging fruit might be found in connecting disparate data stores. Manufacturers work off of different databases: software application platforms, typically used in early stages of design and development, and a manufacturing platform that collects, stores, and manages all manufacturing. “We integrate across those platforms so that data identified as appropriate or related to a particular product is targeted where it sits and can be accessed and visualized,” says Bill Carrelli, vice president of strategic marketing at Siemens. This is the contextualization automakers are seeking.
By having these integration layers in different platforms, automakers and their suppliers can search across them by attributes, and find data related to whatever work is being done (i.e., design, manufacturing). Once the data is identified, it can be viewed, imported via the integration layers, and delivered to those who need it when they need it.
The idea is not to create one huge database or force that to exist. Instead, the goal is to look at databases that are federated in different parts of the globe, in various parts of an enterprise or supply network, and integrate them to identify attributes, commonalities, etc., of data. Then bring that data together. That’s a powerful tool for automotive manufacturers in helping them make better decisions at any point in the product lifecycle process.
“This process improves the velocity of effective decisionmaking,” says Billings. “In a sector that moves as rapidly as automotive, an increase in that velocity can be a significant competitive advantage.”Have an Inquiry for Siemens about this article? Click Here >>