Modern machines are built from thousands of components that constantly measure process values such as pressure, temperature, voltage, and communicate with each other which generates huge amount of data.
Artificial Intelligence algorithms give the opportunity of data analysis searching for dependencies, trends, anomalies, the interpretation of which enables to specify current technical condition of machines and predict failures long before they occur.
Knowing exact time of failure occurrence gives the opportunity of optimization exploitation, service, and logistic processes. It leads to minimizing the number of failures and unplanned stops and, as a consequence, it brings financial profit.
This approach to technical maintenance is called Predictive Maintenance. Our solutions will give you the opportunity to realize this remarkably efficient strategy. Imagine what profit it can bring!
The key factor of our algorithms’ work are data collected not only from machine sensors but also from sensors of the installation on which machine works, indication from Condition Monitoring Systems, data from technical maintenance process or any other valuable data. Our system does not involve any additional measuring tool. It integrates different sources of data and transfers huge amount of data in real time to find patterns foreshadowing incoming failure.
System dashboard visualizes current data of all monitored machines and displays estimated technical condition as a “health score” chart. User can shift between hierarchical structure of monitored machines, installations and even location and see data as KPI reports which helps to make conscious service, exploitation, and business decisions.
When system finds patterns foreshadowing incoming failures, it immediately informs the user about this fact. There are multiple different notification, e.g. via e-mail, SMS, phone call, ticket to CMMS/EAM/ERP. Along with the notification user finds report with detailed explanation on found patterns and suggested actions.
Our system is easily integrated with other IT solutions of a company. It does not contain functionalities of CMMS/ERP/EAM itself and it should be treated as a “analytical plug” that equips those applications with intelligence. As a result of integration, system automatically collects the data from different IT sources, analyze them and returns valuable results. In some cases, by usage of already existing software end user may even not be aware that there is a system exercising holistic control, ensuring safety.
Knowing the time of oncoming failure and knowing its causes enables user can take appropriate action in prior. System enables reproducing organization structure of the company and level of competencies of each user, therefore the information is spread immediately and precisely, documenting its full track. Highest permission user (e.g. maintenance manager) can easily check the information path and users’ reactions.
In many cases the reasons are more important than the notifications about the failure. Knowing the cause of failure one can change the approach from reactive (e.g. machine repair during planned technological stop after failure had been identified by the system) to preventive which identifies reasons and eliminates them before they lead to failure (e.g. a given method of explotation machines, faulty service)
Our system bases on newest achievements in the field of data analysis. For predicting failures we use deep learning algorithms and the computational architecture based on powerful graphics processing units by NVIDIA. Hence, we are able to generate precise predictive models forecasting the oncoming failure and process humongous amounts of data in real time. Combination of those two elements is beyond our wildest expectations.
Our system is designed to be easily expanded to monitor more machines and analyze more data. When the computational resources are not sufficient, we can easily add more computational units. Other solution is to use small, autonomous devices for direct data processing directly on the monitored machine, minimizing the need to send all the data to servers.
Used AI algorithms are equipped with self-learning mechanisms. That means that if there is an event identified by user as dangerous (e.g. failure, loss of productivity or loss of production quality) our algorithm will learn it and forecast it in future. Moreover, machines of similar type can share such knowledge, improving the quality and effectiveness of prediction in future.