Automated monitoring of factory equipment isn't a new idea -- manufacturing machines have long included sensors...
or software tools that continuously monitor their operations and store status data in logs for later review. However, without any standard mechanisms for transmitting that data, it typically was left in place on the factory floor instead of being moved to a higher-level manufacturing management system, let alone a centralized data warehouse.
Fortunately, this "language barrier" is becoming a thing of the past as the internet of things challenges the status quo on analyzing manufacturing data. Via an IoT architecture, networks in manufacturing environments are rapidly being reconfigured to not only connect the machines in a particular factory, but to also pull together data from equipment throughout an organization's facilities.
Such connections enable continuous streaming of device and sensor data that can be processed and analyzed to help drive the following operational improvements:
Better product quality. In any given production run, it's not uncommon for there to be some number of manufactured items that fail to meet specifications for use or are otherwise deemed unsuitable for regular sale. In the best case, these items can be repaired or sold at a discount as seconds; in the worst case, they get thrown out.
If data analysis identifies premonitory signals indicating potential product flaws, a company could take steps to minimize defects and achieve higher and more consistent quality levels. In turn, that could reduce the financial hit caused by product-quality issues, including the sales impact of production delays and the cost of repairing defective items and buying more raw materials to produce additional goods.
Increased production uptime. Unexpected manufacturing stoppages due to equipment failures are the bane of a factory's existence -- they can sharply reduce the overall productivity and efficiency of production lines as factory managers wait (and wait) for repairs to be made. Analyzing operational data can flag possible failures before they occur, allowing preventive measures to help avoid outages and maintain continuous production, thereby increasing overall manufacturing capacity in the bargain.
Optimized production cycles. There are going to be planned stoppages in plants -- scheduled times when production lines are shut down for maintenance, cleaning, upgrades or setup changes required when the items being built on a line are switched. Analyzing data generated by manufacturing devices can aid in better scheduling downtime and minimizing its length in order to reduce the impact on operations.
In each of those cases, running predictive models against the data being streamed from factory devices can have a positive effect on manufacturing productivity and output. The opportunities for operational improvement are driven by the ability to identify the root causes of undesirable occurrences before they happen and address them right away, before their negative consequences are felt.
That's where an IoT architecture and big data analytics applications come in. The massive volumes of data generated by manufacturing machines provide a rich vein of information that can be mined by data scientists and other analysts, as long as they can actually access it. The IoT makes that feasible, in many cases for the first time.
Advance warnings through IoT analytics
Combining an IoT network that can transmit data from the factory floor to a Hadoop cluster or other back-end platform with predictive analytics, machine learning and event stream processing tools creates an early-warning system for manufacturers. When data analysts or automated algorithms detect imminent issues that pose a threat to production, alerts can be sent to factory managers or technicians so they can take actions to head off the problems. Or else an analytics system can be set up to trigger automated actions, such as shutting down a machine that's about to overheat or ordering a needed replacement part.
Data processing and analysis can also be done at different points in an IoT architecture. Predictive models typically are run at the central analytics server against data sets collected from various devices and plants. But parts of the analytics process can be pushed down to local points on a network. That type of edge-analytics approach allows for faster filtering and assessment of smaller data sets. In addition, it can reduce the overall amount of data that needs to be transmitted to the central system, keeping processing and storage resource requirements in check.
Giving data analysts armed with predictive analytics applications access to streaming data from a multitude of factory-floor machines is a new concept for many manufacturers. But it's one that could be hard to ignore, especially as business rivals take advantage of IoT analytics to reduce costs and improve manufacturing quality and productivity.
The importance of an IoT roadmap
How to prepare for the internet of things
Why a common approach to IoT data is needed