Dr. Jay Lee on IoT, Big Data and Successful Industry 4.0 Case Studies – Part 4

I recently introduced a series of posts inspired by a webinar I attended on April 2, 2014, entitled Enabling Smart Factories – The Transformation of Manufacturing Systems & Factory Facilities, hosted by Advantech and presented by Dr. Jay Lee. I introduced Dr. Lee and his Center for Intelligent Maintenance Systems in my first post, and the second covered the first of his five presentation topics, Mega Trends and Unmet Needs. My third post covered the second of his topics, Driving Forces and Emerging Technologies (Big Data, Agent, Industry 4.0 and Cyber-Physical Systems), and this post, Part 4, is dedicated to Dr. Lee’s discussion of Successful Case Studies.

 

Harley-Davidson

 

Harley-Davidson’s production facilities include CNC machining centers used to produce transmission components and other critical parts. These are a high-capacity machines of which the continuous operation is critical to the entire manufacturing line. Spindle health is, in turn, key to the machining centers’ continuous operation.

 

Dr. Lee recommended the installation of wireless monitoring sensors for collection of spindle data during both idle and machining operations. The installed sensors included accelerometers that monitor vibration. A wireless solution was chosen because it eased installation – wired sensors interfere with machining processes. Other advantages of this solution included high sampling frequency.

 

Dr. Lee’s IMS Watchdog Agent collects the data from the monitoring sensors and uses IMS algorithms to convert this raw spindle data into health-condition data. Algorithms include statistical pattern recognition that compares baseline data gathered under normal operating conditions to formulate a hypothesis for calculation of the confidence value of testing data. The resulting analyses allow timely deployment of preventative measures to avoid costly, protracted downtime.

 

 

Nissan Smyrna Manufacturing Plant

 

 Nissan’s Smyrna, Tennessee plant employs numerous robots. The system recommended by Dr. Lee integrated sensors into the robot controllers for gathering of such real-time data as RMS torque profiles. Current profiles are compared to baseline profiles to evaluate the relative health of components such as motor drives. While profile signatures of outright failure conditions are visually distinctive, advanced analytical methods are required for optimally-early detection of invisible problems.

For example, optimal evaluation of the health of a critical servo-motor robot joint involved application of an IMS logistic regression algorithm that included using moving averages and RMS torque values to model unacceptable versus acceptable states. As a result, evidence of degradation is detected well before the occurrence of an actual failure – three weeks earlier, in the provided example – allowing focused, time-efficient preventative measures as opposed to ad-hoc maintenance procedures or, worse, repair after catastrophic failure.