Peering into the Future of Machine Tools with Predictive Maintenance
Ian Wright posted on March 27, 2017 |
EMO experts discuss process monitoring, big data and more.
(Image courtesy of IPK.)
(Image courtesy of IPK.)
The phrase ‘the future of machine tools’ immediately conjures up visions of advanced manufacturing technologies housed in smart factories, perhaps with a dash of 3D printing and a liberal helping of the Industrial Internet of Things (IIoT) for good measure.

However, the future of machine tools can also be much more immediate:

“How soon will my 5-axis mill need an overhaul?”

The key to foreseeing this future is predictive maintenance (PM): it helps users identify the optimum junctures for maintenance work and avoid lost production time. In a recent interview with journalist Nikolaus Fecht, Eckhard Hohwieler, head of production machines and line management and Claudio Geisert shared their views on PM based on their experiences at the Fraunhofer Institute for Production Systems and Design Technology (IPK) in Berlin.

 

Eckhard Hohwieler, how does predictive maintenance differ from condition monitoring?

EH: Condition monitoring detects and monitors wear-and-tear status, whereas predictive maintenance forecasts the putative development of a machine’s future status and plans the appropriate maintenance work required.

Claudio Geisert, how does PM specifically benefit machine tool owners?

CG: Care and maintenance are governed by the condition of the machine. This means the staff concerned carry out precisely the care and maintenance work that is actually required. Effective PM reduces the number of maintenance routines needed and increases machine availability levels. It also enables line use to be more efficiently planned by enabling care and maintenance work to be carried out on pre-specified dates. 

 

One of your specialties is process monitoring and condition diagnostics. Can you give us a highlight from your research?

EH: For one machinery manufacturer, we created a tool monitoring feature without any additional sensors or other electronics. A software package integrated into the control system monitors tool wear-and tear and fracturing. On this basis, we developed further algorithms enabling the machine’s condition and behavior to be checked. This allows an employee to determine weak points with astonishing accuracy using the characteristic values of the drive shafts. Even textile flaws in belt drives have been discovered in this way.

 

Where is the data actually located – at the IPK or the company concerned? Who owns the data and who is entitled to use it?

CG: The data created during the utilization phase belong (unless something to the contrary has been contractually agreed) to the operating company. As a rule, the companies concerned will not reveal these data to outsiders, since they fear that sensitive information will be among them or can be derived from them. One common solution is to install an appropriate server inside the company’s own network. This, however, will deprive the manufacturer of an option for gaining additional insights into the behavior of his machines in the field. In order to overcome this problem, a relationship of mutual trust between the manufacturer and the operator is imperative, though contractual safeguarding for utilization of the data is indubitably helpful.

 

How has this new form of monitoring evolved into PM at your institute?

EH: While working on a project on e-maintenance, we took a long, close look at how information from condition monitoring can be utilized for planning maintenance work. As an aid, we used an electronic service checkbook that specifies the next required work steps. It also explains how users are supposed to prepare and carry out the maintenance routines and where they can order the requisite tools.  

 

Is there a practical example of a PM solution that you’ve developed together with a machine tool manufacturer?

CG: The automotive industry demands guarantees on availability from machinery manufacturers, plus particulars of the anticipated lifecycle costs. This necessitates complete-coverage monitoring of the machine involved. Our solution—developed in conjunction with the grinding machine manufacturer Schaudt-Mikrosa—was electronic monitoring of the drive elements using the machine’s own control system. This acquires and evaluates all messages and signals from the machine, which the system then uses to determine the dynamic behavior of the drive shafts and spindles over a lengthy time period.

 

How does the PM tool benefit the manufacturer?

EH: Schaudt-Mikrosa is already using it as an important tool for purposes of quality assurance – in machinery acceptance-testing, for example, or in the warranty phase to clarify the causes of damage, such as collisions between drive elements, tools and components.

CG: The service personnel use the tool for looking backwards into the past. Thanks to complete-coverage data acquisition and storage, they see when and under what circumstances a problem occurred for the first time, and can thus more easily identify how they can be remedied. 

 

Monitoring and condition diagnostics involve huge quantities of data. When acquiring 20 values (64 bits) per millisecond, the electronics will already be storing more than four gigabytes during an eight-hour shift, according to the Machine Tool Laboratory in Aachen. How do you evaluate this big data?

EH: It’s not a problem for us, because we don’t acquire the raw data. True to our motto of “smart data, not big data”, we determine and store only the typical characteristic values. My impression is that quite often, big data accumulates merely because it’s possible to store huge quantities and generate redundant data copies. It’s more sensible to make an intelligent pre-selection near the machine concerned before storage, so that a reduced data record is then transferred to the Cloud.

 

How can users ensure that a reduced data record does not overlook effects that can then no longer be reconstructed? What options are already available for making a selection?

CG: Compression of raw data to selected characteristic values always entails concomitant losses. The possibility that some effects will be overlooked cannot be ruled out entirely. However, this also applies to data acquisition: what physical variables will be acquired using what sensors with what accuracy? Without a certain amount of domain-specific expert knowledge, a monitoring concept cannot be translated into viable reality.

To make the correct selection, users can utilize machine learning processes in the development phase. This helps experts choose meaningful characteristic values. It must always be remembered that for developing a sustainable monitoring concept you need both knowledge of the fundamental theory involved and experiential knowledge as well.

 

What are the long-term benefits of predictive maintenance for the manufacturer?

EH: You get a fleet effect: huge amounts of information are created over the course of the production lines’ lifecycle at the customer’s facility, which enables the manufacturer’s service capabilities to be improved.

 

Please tell us about your personal vision: what – in conjunction with Industry 4.0 – might an optimum PM solution look like?

EH: It would be conceivable that the machine itself utilizes the information created to optimize its process or “call” the maintenance service.

 

What are you expecting (not least regarding your research work and predictive maintenance) from your visit to EMO Hannover 2017 trade fair for the metalworking sector?

EH: I’m keenly looking forward to seeing how the machinery manufacturers are responding to the issue and what apps they will be presenting. Perhaps, too, someone will even premiere the Machine Tool 4.0, one that tweets on Twitter and allows visitors to the EMO to contact it on their smartphones.

 

For more information on EMO Hannover 2017, click here.

Recommended For You