Blog: How over-sampling of data can generate new business opportunities – and lower risk

How over-sampling of data can generate new business opportunities - and lower risk

Controlling a complex electrical, mechanical, chemical system (or a mix hereof) requires data sensors and processing for the control loop, system monitoring and set-point.

Traditionally Industrial Control Systems encompass optimised use of PLC or SCADA systems set up for managing exactly the task at hand: a stabilising control loop and data acquisition for operators to monitor the well-being of the facility.

Since the cost of installing control systems have been a significant part of the overall investment, choosing a cost-optimised range of components for sensors, signalling and processing elements has been essential. The theoretical principles for optimising control systems further suggest a modest sample rate calculated from the resonance frequency of the system controlled.

However, for many systems there is information of the behaviour hidden in the higher harmonics of the direct data. A rotating spindle would be measured for the direct control and stability of revolutions during a time interval, yet the higher harmonics indicate the vibration pattern which may change over time. A fluid’s pressure is measured to control e.g. a chemical process, but higher harmonics may reveal cavitation or damaging shock waves propagating throughout the system.

During the system design the development engineers were hopefully aware of individual contributors to vibrations or shock phenomena but during later operations this specialist knowledge may not be available – or – the daily wear and tear may instigate almost impossible to detect squeaks increasing over time until impact is costly or damaging.

Choose to add sensors able to record 2nd and higher harmonics of the direct signal used for the immediate control. For speed control of a spindle a simple rpm-meter is adequate for the feedback control loop but applying a microphone to record higher harmonics will yield details of the entire system’s well-being as the resonance pattern changes over time.

Additionally, choose an edge controller capable of recording and analysing the harmonics and transmit those for a Machine Learning tool to log and register firstly the regular as well as changing patterns over time. The price of sensorial and processing components as well as data transmission cost is steadily decreasing, whereas the overall cost of complex systems and productivity loss if failing is increasing.

Hence pre-emptive methods and processes for protection, control, and management are becoming ever more vital for companies and the society.

Generate new business and lower risk

Installing millions of sensors and analysing regular signals and disturbances will generate huge data sets for analysis to predict upcoming errors. The Swedish roller bearing manufacturer SKF has grown by 50% by forming a special entity of vibration experts who eventually know more than their customers of how bearings, the shafts they stabilise, and the surrounding machinery behave. By recording the higher harmonics of the machinery’s vibrations, they can predict when and how it will fail, and by dispatching service technicians prior to a critical point of failure, the customer’s equipment is saved from the effect of structural damage and loss of productivity.

Predictive Maintenance is an emerging skill acquired gradually by recording and analysing the hidden patterns of a continuous stream of data drifting slightly off-tune over time. And the cost of fixing the system – calculated as equipment cost, production loss or even any immeasurable human injury will greatly depend on the time in advance the defect was detected and mitigation initiated.

More information:

Senior Advisor Jesper Meulengracht, jesper.meulengracht@glaze.dk, +45 40 68 39 67