Data and Analytics – Fuel and Engine in Smart Manufacturing

By Josef Kriegmair, Representative Production Turbine Blade / Structure Castings, MTU Aero Engines AG

Josef Kriegmair, Representative Production Turbine Blade / Structure Castings, MTU Aero Engines AG
Josef Kriegmair, Representative Production Turbine Blade / Structure Castings, MTU Aero Engines AG

Raw Material Data
There have been several headlines like data is the new oil, data is the oil of the 21st century, the world‘s most valuable resource is no longer oil, but data.

Gathering of data creates cost, also the processing of data. The main tasks are:

• Capture data
• Bring data into context
• Convert data into insight via analytics
• Use insight to trigger actions

In manufacturing, value comes with successfully completed actions.

Based on my experience, there are two approaches to convert data to business value: Big-Data Approach, Right-Data Approach.

Big Data Approach
The key question of that approach is: Where is the data? Capture all data you can get, process the data, and hope to receive value after the conversion. It is searching without knowing after what you search. The capability of the technology is the core element.

That approach is data-driven during the design phase.

Right Data Approach
The key question is: Where is the value?
The result is Right Data: Required data is captured to create value. Domain knowledge is used in order to focus on the relevant data. The amount of data is often much lower than that of the Big Data Approach. Therefore, the conversion causes less cost than that of the Big Data approach. Right data also means sufficient data integrity, data quality, data protection, sampling rate. The design process starts at the customer. The requirements of the customer are the core element.

That approach is value / customer-driven during the design phase.

Refining of Data to Value
The data & system architecture has to support the chosen data approach. It rests in a smart manufacturing environment on 4 pillars:
• Data storage and processing
• Data quality
• Interoperability
• Human factor

Data Storage and Processing
In a smart manufacturing line in operation since 2010, I have used the stack device, cell, backbone / enterprise for data storage and for data processing. In that smart production line, processing and storage of the data is defined by real-time requirements, computing power, and accessibility of data. Structured, semi-structure and unstructured data is stored in a type of data lake. Applications extract data from that area for post processing. Available bandwidth and latency are key drivers of the architecture as well as governance and regulations. Data processing is done on the layer that has the best access to the required data.

Different processing dimension unleash the value of data.

Figure 1: Dimensions of processing

There is a huge amount of time-series data in manufacturing.

Life happens is the present. Information enables humans and machines to control the process. The guiding question for the present is: What is happing?

Whenever a problem appears in manufacturing, it is important to know the history of a part. Predictive and prescriptive models often need high quality history data for training and validation of the models. The guiding for the past is: What happened? How and why did it happen?

Based on patterns found in the data of the past, events in the future are predicted, e.g. the tool has to be changed before the planned service life due to higher tool wear. The guiding questions for the future are: What is about to happen? (predictive) What shall happen? What is required to make it happen? (prescriptive).

The higher the number of humans users, the higher the importance the user interface becomes. On the one end is an ad hoc analysis, few user high user interaction, on the other end is an automated process monitoring in real-time.

Data Quality
If data is perceived as a product, it will get an infrastructure as a product. A product has a product architecture and its quality is assured by a quality system. Result: Data is regarded as a valuable company asset. With that mindset, data-driven processes are possible.

A free flow of information is a precondition of smart manufacturing. Interoperability is a pivotal.

There are two different customers requiring data: Humans and machines. Interoperability enables machine-to-machine communication and access to data.

Human Factor
The key issue in a data-driven and data-enabled manufacturing environment is that transparency for the personnel is kept. Personnel shall use data to improve performance in manufacturing.

A human and a system interact via a user interface. The better the system and the user understand each other, the better the user experience is. It has a high impact how reliable the system runs and how much training a user need. The spreadsheet has enabled a lot of user to do calculations with a computer. Both items influence productivity of the manufacturing unit. The user interface has to be appropriate for different user groups.

Value and Results
Successfully completed actions lead to operation excellence, i.e. data is used to optimize the value engine “machining”, e.g. by data-driven continuous improvement, to operational insurance, i.e. a risk reduction via transparency which enables a problem fixed based on facts. A new value stream has been created: Data-based products. These may be new products based on data can be sold.