Cisco Blogs
Share

Solving Manufacturing Complexities through Data Analytics: Part Two – Implementing Data Analytics


March 23, 2015 - 2 Comments

Data analytics has been an integral part of manufacturing management for most of its history. However, analytics has undergone both evolutionary and revolutionary changes over the decades with the advent of information technology and digital data gathering and analysis. In Part One of this series, I took a look at the evolution behind data analytics and applications in Manufacturing. Part Two provides insights into implementation of analytics in manufacturing.

Part Two: Implementing Data Analytics in Manufacturing

Acquiring Data

The first step for data analytics in manufacturing should be to implement solutions that connect manufacturing equipment, sensors and controllers to a converged network so data can be captured, moved and stored for analysis in an appropriate manner. While manual data entry is common and will probably continue to some extent, automation is critical to ensure that data is captured in real time, accurately and in the right format to enable analytics and decision making.

The amount of data available on the manufacturing plant floor has increased by many orders of magnitude over the past decade, however analysis and application of such data in decision making has not kept pace. It is this space of analytics that is now driving adoption of ‘Internet of Things’ (IoT) technologies such that IoT and analytics have now become intricately linked to each other.

Going beyond just analyzing data from IoT and expanding it to include the impact of this data on the people, skills, business processes and linking all of these disparate elements into a single business-focused system is referred to as the ‘Internet of Everything’ (IoE). Manufacturers are only now just starting to take this wider perspective to analytics and the application of analytics in manufacturing. As manufacturers begin to rely more on data for analysis into business processes, they must also consider some challenges that may arise during implementation.

Virtualizing the Data

Today’s manufacturers need the ability to integrate all data from various departments/locations, which has proven to be difficult in the past. The old approach consisted of building a data warehouse where data was extracted from multiple sources, transformed (normalized, processed, condensed) and loaded on a periodic basis into a central data warehouse. Today’s manufacturers need data that can be used in real-time to make decisions, not data stored in a warehouse for historical analysis. A steep increase in the use of cloud storage for such data warehouses has led to data being stored across different clouds (mix of public and private) on different platforms. Bringing all of this together to yield meaningful results without moving all the data physically into one data warehouse has been a challenge. Data virtualization solutions now enable accessing data that is physically in different databases and geographic locations as if it were physically in a single data warehouse. This has becomes even more critical with the large volumes of big that are typically unstructured and not easily amenable to traditional data warehousing approaches.

Integrating analytics into business processes

Data analytics cannot be a standalone activity done in a data center by a team of experts. It has to be integrated into the key business processes such that analytics are focused only in areas that provide business value and are available to decision makers at the right time in the right place. Important questions to be considered when implementing analytics solutions are:

  • How will the data be used?
  • Who will use it and how often?
  • What kind of analysis is needed?

Responses to these questions will define your strategy and dictate how analytics are integrated into the business. Implementation models could include

  • Data acquisition from sensors and analytics at the ‘Edge’ to feed-back to control system or human operator. The data is acquired and moved to a computing platform on the switch (in the manufacturing cell network) or to a data center in the manufacturing plant where it is processed and the result is used to drive the manufacturing process through control signals or visual / audio signals through the Human Machine Interface (HMI). Example would be a high definition camera taking 3D images of the product and comparing it to standards to identify quality defects in real time to eject the defective product or stop the machine or just sound an alarm via the HMI for the operator to take action.
  • Data capture from sensors and equipment for periodic reporting. The data is acquired, moved to a data center and analysis / reporting is done in conjunction with other databases on a periodic basis. Application would be machine uptime and speed data acquired in real time and used to report Overall Equipment Effectiveness (OEE) in conjunction with data like product mix, raw material / packaging source etc to identify performance issues and improve OEE.
  • Adhoc analysis of data acquired from sensors, done offline, after data has been normalized and moved to a data center. Typical use case would be analysis in support of six sigma/quality improvement projects where data gathered from the machine / production system is analyzed to support (or reject) hypotheses for problem resolution by shop floor employees.
  • Data capture and streaming out to equipment vendor in real time (machine as a service) where the machine vendor monitors performance of the machine parts and is able to take remote corrective action or schedule predictive maintenance or bring in appropriate spares just-in-time to ensure machine up-time and performance per contractually agreed levels. In such cases, security becomes a key issue too.

Implementation challenges

Implementation of data analytics should consider the following:

  1. Appropriate manufacturing cell and zone network to ensure high speed, quality of service and reliability. This is absolutely critical and is a huge challenge for manufacturers give the proliferation of standards and protocols in use on the shop floor and the lack of convergence of the networks.
  1. Moving and storage of data and location of the data center. This becomes very critical when handling big data in large volumes and high velocity and the decision on whether data center should be co-located in the manufacturing plant or remote/cloud can drive performance and cost of the solution.
  1. A comprehensive strategy and implementation approach focused on the entire data chain and not just on the final analytics and visualization. Typically analytics is seen as using algorithms on data and developing reports/visualization with little focus on acquisition, movement, storage and organization of the data. What appears in the user interface is the most visible but not necessarily the most important or most challenging aspect of implementation.

How can Cisco help your manufacturing organization improve efficiencies and gain valuable insight through data? Visit our solutions page to find out more and share your thoughts with us in the comments section below. Stayed tuned for Part Three of this series where I will share experiences in implementation and detail how analytics and IoT are working together to bring results in manufacturing.

In an effort to keep conversations fresh, Cisco Blogs closes comments after 60 days. Please visit the Cisco Blogs hub page for the latest content.

2 Comments

  1. As a QMS solution provider I agree with the analytics benefits you present. The challenge our clients encounter is linking the benefits to tangible business value. I would appreciate you insights on this topic. Regards, Don Jasurda DCS

    • Donald, indeed evaluating tangible business benefits before starting an implementation is critical. Typically such justification is required to gain funds for investment and also sustain momentum and measure benefits once implementation is completed. Some benefit areas that I have seen at clients are 1. Reduction quality defects / rework / scrap due to improved process control, early identification of out-of-control situations 2. Improved throughput due to reduced number of stoppages etc 3. Reduced cost of maintenance through implementing predictive maintenance 4. Reduced manpower through elimination of a number of data handling activities Cant get into details in this response but will try to write them up and post them in my next post if possible. Thank you for your question