Data-Based Maintenance: Intelligently Planned and Implemented with a Focus on Saving Resources
Digitally planning maintenance of machines and systems and using artificial intelligence to calculate service cycles: that is the future of mechanical engineering.
Understanding why machines fail. Carrying out maintenance early, but in a targeted manner. Providing service technicians with information on the causes of failures and the necessary documents and spare parts on time and saving resources. Avoiding unnecessary interventions. Extending the machine service life. Avoiding abrupt disruptions and unplanned downtimes.
A scenario that would seem to represent the natural state of things and is of course precisely what machine system suppliers wish for. However, the reality of maintenance is different. When maintenance takes place is determined by expert knowledge amassed over many years, and even cyclical maintenance intervals are stipulated. Performance is then left entirely to the customer but can also be an integral part of the service contract during the warranty period. Predictive maintenance based on evaluations and correlations of sensor data is still the exception. The reasons are complex: On the one hand, the increasing complexity and diversity of interfaces in their machinery plays a role for machine manufacturers, and on the other, a lack of resources or willingness to invest acts as a disincentive. Because carrying out truly data-based maintenance and servicing, you need much more than just sensors:
- Preprocessed machine data in real time if possible
- Attractive visualization and end-to-end analytics
- Information on context to enable classification of correlations
- Intelligent and rule-based creation & planning of maintenance
Even before installing the sensors, it is important to consider which process values or alarms indicate imminent machine failure and to prioritize potential faults and their effects as well as the risk. As a rule, many historical evaluations are available because today's machines have sophisticated controls. However, processing this data is very time-consuming and labor-intensive.
The preprocessing, bundling, buffering, and translation of interfaces are handled by IoT devices, which therefore ensure that data is processed directly on site. In a next step, the IoT devices forge a link to the cloud application, where analytics and visualization work their magic together. The aha effect and added value of this data emerge when it is visualized meaningfully and clearly and evaluations are easily comprehensible, comparable, and intuitively available with one click. Dashboards, interactive diagrams, filters, correlations of process value data, status, alarms, and statistical evaluations are of considerable help, since immediate dependencies and causes of downtimes can be localized more quickly and reliably.
Additional added value is offered by contextual information such as logbooks, descriptions of faults gained from tickets, master data, status, or progress of maintenance carried out. The service life of spare parts, which provide valuable clues when searching for causes in correlation with machine data, can also provide information on downtimes.
Cyclical, time-based maintenance intervals are not the right choice for planning predictive maintenance. What is needed are rule-based and intelligent maintenance intervals that “actively contribute” based on data. This can include simple alarms or events, but also process value threshold limits exceeded (temperature, pressures, gases, acceleration, etc.) or meters as well as the operating hours of machines which considers the actual productive times and can anticipate from the historical data when the next maintenance cycle will occur. Notifications are then sent beforehand by email or as push notification, so that the service employee concerned can schedule resources and procure necessary spare parts in advance.
The machine manufacturer’s service technician receives additional support from checklists which are already available for the maintenance case in hand and enable immediate performance because of unambiguous instructions. Maintenance planning is made easier for the administrator or person responsible in the sense that they can define advance maintenance plan templates consisting of checklists and parameters that trigger the maintenance intervals, and place them “on standby”, not only for individual machines, but also for several machines or entire machine types at the same time. Warnings enabling automatic alignment of the system of machine-specific process values or alarms provide additional support, allowing the administrator to make manual changes in the event of discrepancies (missing temperature value) and prevent error messages. This reduces the administrator’s workload enormously, especially when maintenance is to be planned for many machines and customers and there is a strong desire for dynamic maintenance intervals.
Would you like to learn more about implementing predictive maintenance?