Description
DS200PLIBG1ACA GE Mark VI
2 Leveraging big data tool chains
After the data collected from the manufacturing product value chain is stored in the database, a data analysis system is required to analyze the data.
The manufacturing data analysis system framework is shown in Figure 1. Data is first extracted, transformed, and loaded (ETL) from different
databases into a distributed file system, such as Hadoop Distributed File System (HDFS) or a NoSQL database (such as MongoDB). Next,
machine learning and analytics tools perform predictive modeling or descriptive analytics. To deploy predictive models, the previously mentioned tools
are used to convert models trained on historical data into open, encapsulated statistical data mining models and associated metadata called Predictive
Model Markup Language (PMML), and Stored in a scoring engine. New
data from any source is evaluated using models stored in the scoring engine [9].
A big data software stack for manufacturing analytics can be a mix of open source, commercial, and proprietary tools. An example of a
manufacturing analytics software stack is shown in Figure 2. It is known from completed projects that existing stack vendors do not currently
offer complete solutions. Although the technology landscape is evolving rapidly, the best option currently is modularity with a focus on truly distributed
components, with the core idea of success being a mix of open source and commercial components [10].
In addition to the architecture presented here, there are various commercial IoT platforms. These include GE”s Predix ( www.predix.com ), Bosch”s IoT
suite (www.bosch-iot-suite.com), IBM”s Bluemix ( www.ibm.com/cloud-computing/ ), ABB based on Microsoft Azure IoT services and people platform
and Amazon’s IoT cloud (https://aws.amazon.com/iot). These platforms offer many standard services for IoT and analytics, including identity management and data
security, which are not covered in the case study here. On the other hand, the best approaches offer flexibility and customizability, making implementation
more efficient than standard commercial solutions. But implementing such a solution may require a capable data science team at the implementation site.
The choice comes down to several factors, non-functional requirements, cost, IoT and analytics.
Reviews
There are no reviews yet.