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.

Can IIoT and Machine Learning save the auto industry?  First, let me state what you may be thinking.  The auto industry does not need to be saved.

It is the largest contributor to growth in the global economy and is responsible for 5% of world’s manufacturing jobs. In 2016, 72 million cars and 22 million commercial vehicles were manufactured worldwide.  This represents a 4.5% improvement from 2015.

Anemic growth is projected in the industrialized world. As a result, automobile manufacturers are looking to innovation in the following areas:

1) Autonomous Vehicles:  The race for fully-autonomous vehicles is still in its infancy and there are too many regulatory and insurance issues that remain unresolved. The market for semi-autonomous is more promising.  According to BI Intelligence, 10 million cars will have autonomous features such as the ability to brake with limited or no driver interaction by 2020. 

2) Connected Cars: General Motors, Ford and Tesla are at the forefront of the move to the Connected Car in North America as the US is beginning to pay more attention to the development of an Intelligent Transportation System (ITS).

Are Autonomous and Connected Cars inevitable? Perhaps. The problem for the industry is the price: the significant investment in R&D and manufacturing processes.  Furthermore, PwC suggests that even if there will be new revenue streams (digital services, shared mobility etc.), OEM P&L’s will be hurt by a reduction in revenue from legacy features such as entertainment, navigation and safety systems. 

The bottom line is that with all the excitement about digitalization, there are still concerns about securing new revenue to offset the increase in R&D expenditures from uncertain income sources.

A Non-Technical Explanation for Machine Learning for Asset Maintenance and Industrial Analytics

Up until recently, Machine Learning was firmly ensconced in academia.   The commercial application of Big Data and Artificial Intelligence to Industry 4.0 and the automotive industry is now possible due to the significant reduction in the costs of network connectivity, data storage, access and transportation, and computational power.  Sensor prices have also fallen dramatically in the last few years and are expected to continue to do so.

The result is that machine sensor data can be accessed and processed in real time using advanced Artificial Intelligence.   All factory sensor data can be analyzed without the need to prioritize critical sensors.  Using Machine Learning, the algorithms detect abnormal sensor data behavior and then find patterns of anomalous behavior.

Machine Learning is particularly relevant to the automobile industry.  The trade publication, Reliable Plant gave an example of a major North American automotive manufacturer with maintenance staff of over 15,000 in all combined factories.   This company reported that “85 percent to 90 percent [of their maintenance work] is crisis work.”

With Industrial Analytics, this can change. A facilities’ sensor data is transferred to the cloud and analyzed in real time.  Patterns of abnormal behavior are used to detect emerging machine failure. Rather than waiting for the inevitable crises work, Machine Learning identifies the emerging threats.  This allows the maintenance staff to work proactively on fixing machines before breakdown.

Asset Maintenance and Industrial Analytics

What is the Difference between Predictive Maintenance and Machine Learning for Asset Maintenance?

Machine Learning for Asset Maintenance and Industrial Analytics exceed the capabilities of Predictive Maintenance (PdM) or Condition Monitoring that have typically been associated with cost savings.

The fundamental difference between the IIoT approach described in this blog and the traditional methodologies is that the former applies to all the assets in an automotive plant and the latter is sensor or asset specific.

The traditional approach to PdM is to analyze data for a specific asset.  For example, sensors can record the temperature of a factory robot part.   Manual control limits are set based on past experience and if these control limits are breached, an alert is triggered.

There are two challenges with PdM.  First, the system only identifies breaches of the control limits but does not detect any other abnormal behavior patterns.   Second, only a limited number of assets can be monitored baas a result of bandwidth constraints.

IIoT based Industrial Analytics tools such as Presenso analyze all the sensor data from a production facility and not merely the high priority sensors.  As a result, the unknown and unexpected root causes of machine downtime can be identified when they are still emerging threats.

What is the Difference between Preventive Maintenance and Machine Learning for Asset Maintenance?

One of the biggest challenges for factory owners is to create the optimal investment level in Preventive Maintenance.  While regularly scheduled maintenance prevents some asset failure, there are several limitations.   According to research, only 13% of Preventive Maintenance is considered worthwhile and almost 20% should be eliminated (See Review of Preventive Maintenance Activities) below.

What are the drawbacks of Preventive Maintenance?  Finding the optimal Preventive Maintenance balance is extremely difficult.   Preventive Maintenance is labor intensive and often unnecessary.  By definition, scheduled machine shutdown reduces factory yields.  In fact, 30 to 40% of Preventive Maintenance costs are spent on assets with negligible failure impact.

The question is not whether machine assets require regularly scheduled maintenance, but how often this should occur.  As shown in the chart below, 30% of Preventive Maintenance activities occur too often whereas 25% do not occur often enough.  There is a high expense associated with over-maintenance.  Similarly, assets degrade over time and Reactive Maintenance results in extended downtime.

Review of Preventive Maintenance Activities


Preventive Maintenance is about estimating when a critical asset is most likely to fail and remediating before a problem can occur.  With Machine Learning for Asset Maintenance, all the sensor data of all the factory assets are analyzed in real time and abnormal behavior is detected.   One of the outcomes of detecting the failure origin is that maintenance can be prioritized on those assets that are most likely to break.

Industry 4.0 and the use of Big Data will not eliminate the need for Preventive Maintenance.  At the same time, automobile manufacturers can rationalize their spending and move towards the goal of a reaching the optimal level of investment in this activity.

Machine Learning for Asset Maintenance: Improved Uptime and the Bottom Line

When comparing Machine Learning to Predictive and Preventive Maintenance, the metric we reviewed was cost related.  However, there is an even bigger opportunity:  additional revenue.

Let’s start with a powerful statistic:  $2,200 a minute or $1.3 million dollars per hour.

Over a decade ago, a study conducted by Nielsen Research of automobile manufacturing executives found that the value of each minute of production downtime is $22,000.  In some cases, the number is as high as $50,000 a minute or $3 million an hour.    According to one study, the number of lost revenue producing-hours is 661 hours a year.

Even if there is an overhead cost included in the $3 million an hour, the single largest contribution factor is the opportunity cost from lost revenue.

Machine Learning for Asset Maintenance applies to all the sensor data in a factory.  Until now, preventive maintenance was limited to the obvious causes of downtime or highly critical machinery.   With IIoT Industrial Analytics based on Machine Leaning, maintenance staff can focus their attention on remediating specific machinery that is showing signs of degradation or failure.   Even using conservative estimates, if a factory can increase uptime by 5% per annum, this would translate into $40 million at a single facility.


Machine Learning and Industrial Analytics are not theoretical exercises.  They are a way for companies to realize unrealized potential from increased uptime.   The automotive industry can now tap into the power of big data and machine learning and apply advanced Artificial Intelligence to the biggest problem it faces:  how to generate more income.

Does Machine Learning for Asset Maintenance apply to your organization?  To learn more about Presenso or to schedule a demo, please contact us here.