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.

At a future point in time, how will historians look back at the first few years of the fourth industrial revolution? Although the far-reaching potential of IIoT has been widely recognized, the reality is that we are still in the early stages of deployment. In fact, according to the World Economic Forum, 85% of potential assets remain unconnected.
At Presenso, we work with organizations that view Industry 4.0 as a strategic competitive advantage and that have committed resources to its adoption. This article offers five predictions for the maintenance and reliability discipline based on anecdotal evidence and engagement with early adapters in Europe.

1. Over time, the Digital Twin will be implemented selectively

For the last three years, Gartner has included the Digital Twin in its list of top 10 strategic trends. Despite the positive publicity, the hype has not yet penetrated the Operations and Maintenance (O&M) arena.

The slow implementation of the Digital Twin can be explained by two factors.

First, a surprising number of plant-level O&M employees are not aware of the Digital Twin. In a research study conducted by Emory University and Presenso, 28% of respondents were not even familiar with the Digital Twin concept.

A second factor is that the Digital Twin is both expensive and time-consuming to deploy on existing industrial equipment. To develop a virtual clone of physical equipment, a 3D model based on the blueprints for the equipment is required. For the Digital Twin to be accurate, the blueprints must be up to date, which is often not the case. Furthermore, plant-level technicians must work with Big Data experts to train the Digital Twin on the production processes and machine behavior.Digital Twin

2. A shift from Centralized to Federated Manufacturing

Federated Manufacturing refers to an emerging trend in which micro-factories replace large centralized manufacturing sites. In the past, drivers of manufacturing decisions were based on factors including economies of scale and local labor costs.

Industry 4.0 has changed this model. First, Big Data and Machine Learning can be applied in new ways to predict demand and customize production. Second, because of new manufacturing techniques such as 3D printing, production can move closer to the end customer.

What does this mean from a maintenance perspective? Although manufacturing will likely become more decentralized, we expect that offsite monitoring centers will support multiple manufacturing plants. With access to real-time sensor-generated data and cloud-based Machine Learning predictive maintenance, evolving asset failure can be detected in one location and technicians dispatched from another location.

3. A new OEM model: Hardware as a Service

Hardware as a Service is a concept that Rolls Royce pioneered in the 1960s. Instead of selling jet engines to its aviation customers, Rolls Royce leased the jet engines and charged for usage (Power by the Hour).

Advances in Machine Learning are allowing OEMs to move from a sales model to a service model. This is how it works: The OEM maintains ownership of the underlying industrial asset and is responsible for maintenance activities. The OEM embeds sensors within its industrial equipment which generate Big Data that is analyzed in real time. Machine Learning algorithms detect patterns of anomalous behavior and technicians can be alerted to changes in asset health.

4. The operationalization of Machine Learning

Due to advances in Machine Learning, the ability to scale AI-driven Predictive Maintenance has significantly improved. Here are two examples.

First, Automated Machine Learning (AutoML) automates many of the repetitive and laborious data science tasks, such as data preprocessing and model selection, that data scientists are currently performing. Through the replacement of human data scientists, the deployment of Predictive Maintenance is accelerated and solutions can be scaled across a greater number of industrial machines.

Second, the application of Reinforcement Learning to Predictive Maintenance is improving algorithms’ robustness and accuracy.

How does Reinforcement Learning work? The traditional approach to training a learning algorithm is to by using historical data. The problem is that over time, the algorithm becomes dated because the training was based on historic data and not on more recent incidents. With Reinforcement Learning, current data is used to continuously improve the performance of the algorithm.
In the case of Predictive Maintenance, user feedback about the accuracy of the predictive alert is provided back to the algorithm for it to re-train on the updated data.

When an algorithm correctly makes a prediction, it receives a reward (reinforcement). If the prediction of failure is accurate, the algorithm is adjusted accordingly, thereby improving its future performance.

Advances such as AutoML and Reinforcement Learning will increase the adoption and accuracy of Machine-Learning-based Predictive Maintenance tools.

5. Replacement of human labor

Today, robotics plays an important role in the production process. In the future, we expect that many labor-intensive reliability and maintenance tasks will be handled by machines. For instance, drones can replace humans for pre- and post-repair inspection tasks.

One unexpected consequence is an overall improvement in Occupational Safety and Health (OSH) due to a decline in work-related accidents among O&M employees.

Summary and conclusion

Although the practices outlined in this article are already taking place, we are not yet witnessing widespread adoption. I believe that in the coming years, the benefits from advances in Machine Learning and technology will result in more efficient and cost-effective maintenance and reliability practices.