IIoT’s role in production line diagnostics

The business proposition for the connected machine and new aftermarket services has merit when considering an isolated machine and a single machine builder. However, factories comprised of a heterogeneous array of machines integrated as a singular production line pose a whole new set of challenges – and potentially an even bigger opportunity. When trying to visualize how IIoT or connected machines can play a role in solving real manufacturing problems, one has to consider that the root cause can be hidden in operational setup data, machine performance characteristics, tooling variations, or even environmental characteristics.

One driver behind the Industrial Internet of Things (IIoT) is the ability to achieve higher levels of quality output in high variation production environments. As more production operations move toward shorter production runs and greater variation in products, the challenge will be to produce quality product on the first production run. Quality in many facilities is really a behavioral or cultural issue that requires a conscientious work force to respond to equipment maintenance issues. Moreover, some manufacturing organizations simply flout the use of sensors and instrumentation to determine whether intermediate production steps are within specification.   Either sensors don’t exist to detect certain variations, the cost of instrumentation is perceived to be too high, or the sensor is not reliable. This is where the conceptual framework of IIoT plays a critical role by connecting dots in sparse manufacturing data.

For example, “surface quality” on stamped metal parts cannot be easily measured with a sensor. Surface finish flaws not detected at this stage will result in an inferior painting finish. When deviations are detected at the end of the line, production lines are stopped to rectify the problem. Operators work back through the production process steps – which may include welding, stamping, shearing, bending, grinding, along with a host of other production steps – to identify and fix the problem.   This is difficult and time consuming because production inaccuracies may be incurred as a result of several of the process steps. In the stamping operation, they may have contaminants that would affect the surface quality. There are also deviations that are introduced as a result of the welding operation. The heat from the welding operation can distort the finished product. Operators making changes to the parameters of the weld control can introduce variations. There are a myriad causes of production variations, and the challenge is determining the root cause.

This is where IIoT can help. Every production parameter can’t be instrumented with a sensor, as it is too cost prohibitive. However production and performance data from connected machines is a source of information that can be mined to create a new domain of knowledge that is able to predict not machine degradation but production line degradation.   Line degradation is a result of multiple manufacturing steps that compound to create unacceptable quality parts. Factories that employ connected machines and sensors will be able to sample larger volumes of process production data using sophisticated analytic engines capable of identifying correlations that are not easily detected by the production workers or process engineers.   This can assist in solving production problems much faster than before. Furthermore, sensors capable of detecting quality problems at every stage of the production line may not be feasible, but incorporating sensors that measure process parameters can be an indirect source of information to characterize the production states. The disruptive aspect of this is the ability to initially use multiple interconnected machines as sensors to generate large data sets. Sophisticated analytics leveraging “machine learning” techniques can be used to identify relationships and patterns simply not visible using traditional Statistical Process Control production tools. Ideally, “machine learning “will be applied in ways that lead to further development of rules for cognitive systems that can assist production workers and process engineers on the production floor.

About ARC Advisory Group: Founded in 1986, ARC Advisory Group is a Boston based leading technology research and advisory firm for industry and infrastructure.

For further information or to provide feedback on this article, please contact nsingh@arcweb.com

Leave a Reply

Your email address will not be published. Required fields are marked *