Seite wählen

Leider ist der Eintrag nur auf Amerikanisches Englisch verfügbar. Der Inhalt wird unten in einer verfügbaren Sprache angezeigt. Klicken Sie auf den Link, um die aktuelle Sprache zu ändern.

Leider ist der Eintrag nur auf Amerikanisches Englisch verfügbar. Der Inhalt wird unten in einer verfügbaren Sprache angezeigt. Klicken Sie auf den Link, um die aktuelle Sprache zu ändern.

According to some industry analysts, global companies will spend over $900 billion per year on Industry 4.0 digitalization.  In an era of disruptive technologies, it often takes discipline to overlook hype and delay investments.  In this article, I review the cost of industrial analytics and explore ways to minimize investments.

Let’s start with the three elements of Maintenance 4.0.   The first component is industrial analytics – bringing Artificial Intelligence and Machine Learning to the production line.  By analyzing Big Data generated by sensors embedded in industrial machinery, signs of evolving asset failure can be detected before degradation occurs.  A second aspect is the use of automated failure reporting, repair scheduling and inventory management. Third, the actual repair and inspection functions of OEM can be (partially) performed by robots and drones.

Four Options for Industrial Analytics

Because the application of Machine Learning to the maintenance discipline is a relatively new phenomena, there is no definitive approach to industrial analytics. Traditionally, technicians have monitored SCADA data for breaches of manually set control thresholds.

There are four approaches that are typically used for Maintenance 4.0 Industrial Analytics:

The Digital Twin:  The Digital Twin is a virtual clone of the physical asset.  Although the Digital Twin has not been widely adopted, it is a powerful tool because it provides real time visualization of the performance of the asset.

From a cost perspective, when the Digital Twin is built for a new solution, it requires significant input from technicians within the industrial plant. Furthermore, external Big Data Scientists and Engineers are required to train the Digital Twin on the behavior of the underlying machine asset.  The direct cost of the software coupled with the labor-intensive deployment makes the Digital Twin an expensive option.

It should be noted that if the Digital Twin is bundled with the purchase of new equipment, then the one-time development of the virtual clone is eliminated.  However, the ongoing labor cost associated with monitoring the Digital Twin should be accounted for.

Manually Statistical Modeling:  There are both custom and off-the-shelf software packages to perform statistical modelling using Big Data samples.  Apart from the application (software and hardware), it requires the skills of trained experts in statistical analysis.  In today’s market, there is a shortage of skilled Big Data professionals thereby driving the cost of this workstream.

Software as a Service:  Cloud-based industrial intelligence solutions do not require the installation of hardware or software at the plant.  Also, because analysis is performed offsite by the vendor, there is no need to hire Big Data Scientists or Engineers.

In the case of Presenso, the Automated Machine Learning (or AutoML) methodology is used to reduce much of the manual work associated with Machine Learning. Based on labor costs (and shortages), Software as a Service is cost advantageous relative to the Digital Twin and Manually Statistical Modelling.

Hardware as a Service:  OEMs have recognized that there is a financial opportunity associated with bundling their equipment with industrial analytics capabilities.   With the Hardware as a Service (HaaS) model, OEMs can lease their equipment to industrial plants and monitor asset health from a remote location.  When evolving failure is detected, the OEM can dispatch technicians to remediate.

In the case of HaaS, the cost implication does not merely relate to the cost of the industrial analytics solution, but the change in how OEMs charge for the usage of their equipment.


When considering the cost of Industrial Analytics, it is important to recognize both the cost of the solution (hardware and software) and labor (internal and external resources).  The adoption of Hardware as a Service whereby Industrial Analytics is provided as part of an overall service model adds a further complexity to the evaluation process.

The good news for industrial plants is that because of innovations in the Machine Learning displacing such as AutoML there are more purchasing options for Industrial Analytics.