Discover how predictive maintenance is revolutionizing plant equipment management!
Let's compare plants A and B's approach to bearing replacement. Plant A follows a standard schedule, replacing bearings every every 10,000 hours, without precise optimization. Plant B, on the other hand, thanks to the collection of relevant data, evaluates the actual condition of the bearings, the operating conditions and the overall condition of the machine.
Result - Optimized replacements, substantial savings and trouble-free production.
"Transform your company's long-term profitability with predictive maintenance (PdM) "
By maximizing equipment availability and eliminating most breakdowns, PdM optimizes the use of material, human and time resources. resources. According to a study by the U.S. Department of Energy (page 52), the benefits are impressive:
10-fold return on investment - Maintenance costs reduced by 25 to 30%. - Elimination of breakdowns by 70 to 75%. - Unplanned downtime reduced by 35-45%. - Increase production by 20 to 25%.
Of course, these advantages require initial investment and a learning phase. Like machine learning PdM thrives on data. - Opt for intelligent investment today to reap the rewards.
Opt for intelligent investment today to reap the rewards of optimum performance tomorrow.
To successfully implement your predictive maintenance solution, you need to think strategically in two key stages. necessary!
1. Define the ideal scope of action.
- Identify the operations with the greatest impact on machine downtime.
- Find out which machines are frequently subject to breakdowns.
- Analyze the operations influencing quality defects.
- Evaluate which operations generate the highest costs.
2. Encourage essential human involvement.
- Actively involve maintenance personnel in the use and exploitation of data.
- Transform the predictive maintenance solution into an intuitive interface, linking sophisticated analysis models to a language that everyone can understand. that everyone can understand.
- During the learning phase, use their expertise to characterize events, improving the accuracy of fault predictions and reduce false alarms.
Make your predictive maintenance deployment a fluid, collaborative experience, where every member of your team becomes a key player in performance.
Psst! Here are some useful tips!
For immediate, quantifiable impact, opt for a pilot project targeting a few key machines. Then, make it easy to measure by choosing equipment where the benefits are clearly demonstrable. Because yes, data is the cornerstone cornerstone of success, and the right choice of sensors is essential!
You can maximize efficiency with targeted instrumentation, limiting the number of sensors required. For a rotating machine, an optimal configuration might monitor vibration, temperature or even acoustic emissions.
Your predictive maintenance project starts with strategic choices, supported by relevant data and optimized sensors. sensors. Dare to revolutionize your operations with a proactive, measurable approach.
1/ Identifying and reporting relevant data
Sensor Selection
Data is at the heart of any predictive maintenance project. Depending on the sensors installed, it is possible to set up efficient monitoring by limiting the number of sensors, or conversely, to push analysis using a variety of sensors. For example, for a rotating machine, efficient instrumentation that limits the number of sensors to a minimum will monitor : vibration, temperature or even acoustic emissions.
Depending on the type of machine, other sensors can be installed to collect data on (non-exhaustive list): pressure, electrical currents, torque for rotating parts, oil/lubrication quality, motor circuit analysis, ultrasonic and acoustic analysis, electromagnetic measurements.
Sensor Communication Mode
There are a variety of possible architectures that can enable sensors to transmit collected data to a third-party cloud or ERP. The choice depends on : the size of the plant, its configuration: whether or not it has several floors, for example, the presence or absence of Faraday cages, the type of WAN network available (4G/5G? LPWAN?...) limitations/restrictions linked to internal security rules concerning data, its transfer and storage (confidentiality)
As IOT sensors are wireless, they offer both an economic advantage and ease of deployment: 5 minutes on average.
There are two classic deployment methods: Cloud: data is centralized and accessible on any connected device. It's a digital platform that can be integrated with an ERP or CMMS thanks to its APIs. On-premise: data is stored on the plant's own servers. Nevertheless, it does not depend on an Internet connection, and ensures the highest possible level of confidentiality.
2/ Data analysis and learning predictive models
This is the most important step in predictive maintenance, and certainly the most important criterion for choosing a predictive maintenance solution.
The creation of predictive (or prognostic) algorithms is a difficult and time-consuming step. The aim is to build a model that takes into account numerous variables and the way they influence each other.
Initially, the installed sensors generate data that can be used as is, by setting thresholds. The sensors, then in standby mode, wake up and perform a complete data collection and analysis according to a programmed and configurable period: 10 minutes for vibration analysis, for example, is a good compromise, since the higher the frequency, the greater the strain on the sensor battery.
If one of the readings exceeds the threshold, an alert is sent. This is known as condition-based maintenance. It enables real-time monitoring of the machine's condition. It is generally also possible to monitor drift: a series of readings which do not exceed the threshold, but which show an upward or downward trend, is a sign of potential minor degradation.
Thresholds can be set in 3 ways: the standard communicated by the manufacturer experience accumulated by maintenance teams learning from predictive models
On the other hand, condition-based maintenance does not identify the source of the fault, and forces maintenance teams to check the machine to understand the nature of the fault. By operating via threshold crossings, sources external to machine operation are taken into account. As a result, an impact on the machine or a nearby noise source, for example, can produce false alarms. This is where the involvement of maintenance teams is important. The data collected, analyzed and used to monitor the crossing of thresholds thus enables the models previously acquired by the chosen maintenance solution to be improved.
Ultimately, the aim is to predict breakdowns by taking into account the reality in the field: the behavior of the machine being monitored.
The algorithms follow a set of predetermined rules that compare the machine's current behavior with its expected behavior. Measured deviations enable monitoring of the progressive deterioration that will lead to failure. On the basis of the deviations, current operating conditions, past failure data and all other variables integrated into the data model, the algorithms attempt to predict failure points.
The experience of the chosen predictive maintenance solution is therefore crucial. It is this previously acquired experience that enables the algorithms to rapidly gain in maturity, or in the best case, to be able to apply a plug&play predictive maintenance solution to common machine types such as motors, pumps or AHUs.
3/ Transition from pilot to plant-wide implementation
A few weeks/months have passed, the solution seems promising or is already delivering satisfactory maintenance recommendations detailing the source of the problem, its severity, as well as its evolution over time and future estimates: It's decided, deployment for critical machines or for almost all equipment will take place. This requires a certain amount of organization.
However, if the scale is not the same, the steps are. Installing the sensors and learning the predictive algorithms are very similar to the pilot project set up a few weeks/months earlier. However, particular attention needs to be paid to training the teams who now have access to the predictive maintenance tools.
Predictive maintenance does not replace maintenance teams, but it is an additional element in the technician's toolbox. It will enable them to concentrate on the more technical aspects of their job. Better still, in some companies, the deployment of IOT sensors is helping to improve safety in certain areas.
ROI will obviously be at the heart of the project. It can be easily estimated before a roll-out project if the pilot has been carried out correctly. The business case thus established makes it possible to estimate the gains at plant level, or even for a group of sites within the same group.
