The Path to Predictive


The Industrial Internet of Things, IIoT, is creating profoundly positive impacts for the machinery industry.  Solutions are ready and available that reduce service costs, increase machine performance, save energy and improve quality. For leading machine companies the IIoT has become a significant enabler that can get them to market quickly, create additional savings, and provide profitable revenues. With so much at stake, machine builders are thinking deeply about their strategy to deliver predictive services.

Machine builders want to monitor equipment performance and create new service programs to predict key outcomes. They know that value is created by predicting when a consumable is needed, or by scheduling a part replacement during a routine maintenance session. Predictions like these foster new business models that alert people and systems to take action before something goes wrong. Predictive is a major value driver of the Industrial Internet of Things and as such, many companies are searching for practical ways to provide predictive services to their customers. At ei³ we have a solution.

Since our start in 1999, ei³ has provided predictive solutions based using the Internet and data analytics.  From the vantage point of experience, we have created a roadmap for machine builders to reach the promised value of predictive services. We call this the “Path to Predictive” . This article outlines that path with some reasons why it can be a useful guide to a successful IIoT strategy.

The Path to Predictive

1: Connect All Machines

Putting a global install base of machines online is a daunting challenge that’s achievable by adhering to a broad vision of having all your machines online. Once a commitment is made to this vision, the first step is to standardize machine designs and install equipment so that every machine is ready to connect.

Don’t make IIoT an option, ship every machine ready for connection.

Add secure internet communications equipment to your machine bill of materials. This step “future-proofs” machines and builds an addressable market for your new business of IIoT services. Every machine shipped contributes to a growing install base. Your customers will appreciate the ability to put their machines online and benefit from your predictive cloud-based IIoT services.

Choose a hardware solution that works with new and existing machines. Your customers enjoy the benefits of the IIoT and they will want to put it on all equipment. Make it easy by choosing a flexible connection device.

Connect Machines

Establishing a secure connection should be an integral step of commissioning a new machine.  Technicians can use the connection to check out and calibrate machines from their desk. Doing this usually reduces travel costs in warranty reserve so much that it creates a ROI for the hardware and connection simply by reducing the startup and warranty costs. Network security is an important concern so the machine connection solution must have features to enable controlled, audited remote access for authorized and authenticated technicians and engineers.

OEMs should share benefits and cost savings with their customers. This will provide incentives to allow machine connections.

One ei3 customer has measured an 80% reduction of service trips for start-up and warranties. Machine owners are aware that the IIoT brings value by reducing unplanned downtime and this awareness has changed attitudes and policies about connecting equipment. If you deliver a solution with world class security, your customers, the machine owners, will allow connections to be made to plant floor equipment. And they know that services through secure connections provide a solid return on investment.

Step 2: Monitor Machines and Collect Data

Today everything that is important to the productivity and quality of a machine is controlled by a Programmable Logic Controller, industrial PC, or embedded computer. Internet communications using standard protocols make instantaneous readings of key values from these controllers readily available to be collected as data.

In fact, there is a growing list of IIoT standards that make it easier to gather machine data. Examples include OPC-UA, PackML and others. But don’t wait. There is no need to redesign machine controls around these standards. Valuable data monitoring can already be done using existing legacy communication methods.

Machine data is collected online by cloud servers that gather and store key values in a database. This forms the foundation for analytic-driven insights that ultimately improve machines and production. Six-sigma practitioners appreciate having convenient access to a secure source of the machine data that they need to understand their process and operations. Cloud stored machine data can be used for compliance reporting because it provides traceability and is safely backed up off-site.

Store values shown on the machine HMI. It is better to start the IIoT journey than wait to complete a complex definition project.

A common question is what data to collect? There is no real answer to this question but a good place to start is to capture everything shown on the machine’s HMI or operator interface – set points, feedbacks, states, and faults.

Step 3: Track Key Performance Indicators

Organizations relentlessly strive to maximize the output of their shop floor machines.  Sharing real time machine performance across the shop floor is a proven way to boost production.  Studies show that displaying machine KPIs creates a ROI by improving plant floor OEE by up to 5% per year.

Displaying OEE performance can be challenging for manufacturing plants. The IIoT provides a perfect solution to capture, store, analyze, and display machine Key Performance Indicators.  Monitored data from securely connected machines is analyzed in the cloud to generate informative performance numbers e.g. OEE. This information is delivered to plant people via web pages, dashboards, reports, and mobile apps. Performance information is also shared with computing systems, ERP, and business software using RESTful APIs to exchange data.

Machine owners will pay for Software-as-a-Service subscriptions that provide these real time measurements.  This is one way the machine builder is rewarded for delivering IIoT services.  Performance KPI data becomes a global machine standard measurement that helps product development engineers understand machine performance in different settings at different customers.  The aggregation of performance information is a key requirement for analytics to be able to understand– and find – the golden nuggets of information that drive predictive services.

Step 4: Use Data Science and Discover Relationships

The largest improvement opportunities usually come from the largest sources of unplanned machine downtime. So the most fruitful efforts start by finding these largest sources. This requires capturing every downtime event and the reason behind it. Pareto analysis and tree maps of these unplanned downtime events separates the vital few from the trivial many and quickly reveals problems that hold promise for machine improvement.

Making changes and solving the root causes of the top culprits behind unplanned downtime across an entire installed base yields a strong return on investment for both the machine owner and the machine builder. Downtime data is best analyzed using cloud services because data from many machines can be aggregated. By bringing in data from multiple machines across multiple plant sites – and for the OEM even across customer company boundaries, – makes it possible to find the most common major causes of problems.

Make sure all machine stop reasons described as “Faults” in your machine control program are collected in the cloud.

It is a perfect application for the IIoT, because unlike an in-the-walls solution, the cloud solution can measure the aggregate stop reasons for an entire global install base. This powerful technique reveals the top reasons for machine stoppage – independent of customer use, product, geography and many more factors. The resulting information leads to better performance and better machine designs.

Further, most digitally controlled machines have logic that automatically stops the machine if something goes wrong. These stop reasons are known and categorized by the machine controller as “stop codes” or “fault codes”. The communications channel that monitors machine data also monitors these stop codes. This makes it easy to deploy solutions that capture, analyze and display machine stops. It is not unreasonable to expect a 10-50% reduction of downtime by pursuing this methodology.

Step 5: Run Prediction Models and Send Alerts

It seems like every discussion about predicting things in machines ends up trying to predict absolutely everything that could possibly happen on a piece of equipment. It is a great goal and one that may be achieved one day, but it can lead to a project too large to approach. We recommend a stepwise approach. For example, consider the value of being able to reasonably predict the few things that cause up to 80% of unplanned downtime. Putting a prediction system like this online immediately creates value for the machine owner and gives your engineering team the experience necessary for taking further steps towards building ever more reliable equipment.

Start by creating prediction models for the important few, the top reasons for downtime

This step requires data – lots of it. In most cases many machine-years of operational and service data is needed to perform an analysis that produces good prediction models. This is why taking steps 1-3 of our Path to Prediction right away are so important. A data scientist can perform deep analysis and produce practical, predictive models.

Models are developed by studying machine events and their operational impact. Using multiple analytic methods, the data scientist discovers key relationships in the monitored data that leads up to the most common events. These relationships are compiled into predictive models which can be programmed into cloud-based applications to set up alerts that deliver the promised value.

Predictive Models look for characteristics, i.e. patterns in machine operational data that lead to costly unplanned machine downtimes and service. Predictions delivered by the machine builder give the machine owner the opportunity to proactively address these items and thus reduce unplanned downtime and its related expense.

At ei³ we believe that the resultant prediction models should be proprietary Intellectual Property of the machine builder.  These models are based on deep understanding of the dynamics of their equipment. By providing their market with proprietary predictions, the machine builder can create a differentiating competitive advantage for their equipment. Our Data Science team can provide this analytic work for machine builders to get started with their predictive business.

Step 6: Deliver a Predictive Business

Once a machine’s set of predictive models are defined, they are ready to be run in the cloud. Incoming real time data from monitored machines is compared to the predictive models. If the evaluation of a model indicates probable trouble, then an alert is sent to the machine owner, the machine builder or both. These alerts can take many forms; a status report, a web page, or an immediate email or mobile alert. The form is determined by the customer and the business model behind the prediction.

Predictive information can drive spare parts and consumables sales, help plan maintenance, reduce quality failures, or all of the above. The Path to Predictive is a journey. By putting predictive measurements in place, the models are continually improved and refined to develop increasingly sophisticated, reliable and valuable predictions. One benefit of having the performance measurements in place from the earlier steps is that progress is easy to measure, and it shows a very definitive return on investment.

A place to start is predictive models for the most common spare parts. Alert your aftermarket team when machines in the field need replacement parts at defined intervals. It will create immediate business value to your customer and boost your spare parts sales.

The ei³proposition to Machinery Builders

Founded in 1999 with the vision to use the Internet to manage and monitor machines, ei³ has delivers OEM solutions for leading machine builders serving a diverse range of industries, including Plastics, Paper, Converting, Printing, Packaging, Processing, and others. The ei³ IIoT platform can be added to a machine with minimal customization, it is easy to pilot and instantly demonstrates value to machine builders and their end customers. Our white label solution is comprised of four pillars: gateway hardware, a secure global network, a private cloud, and a suite of SaaS applications. All of these together provide a single source for a complete and comprehensive OEM offering.

In the spring of 2017 ei³ added Data Science services. At our office in Zurich, Switzerland, the ei³ team focuses on analyzing and building machine predictive models. The team was created to help machinery companies by providing access to data scientists who are skilled and experienced at applying the latest big data analytical methods to understand and predict behavior of industrial machines.

Machine builders who want to deliver predictive solutions to their customers can benefit from ei³’s years of experience, mature multi-tenant IIoT platform, and proven road map. If you are interested in learning more, contact us to learn more about the opportunity that follows the Path to Predictive.