The value of a machine learning platform
Aging qualified workforce and more stringent environmental legislation are just a few important reasons why organisations today are looking to automate and optimise their operations. By learning from recent asset data, machine learning models can identify derailing processes before they become real issues. This enables your organisation to transition from reactive to proactive operations and maintenance. In addition, machine learning models can learn to directly control processes by sending setpoints to your PLC/SCADA systems.
Our machine learning platform was developed with a digital twin mindset. This means that machine learning models on our platform can be developed to target a common asset or process. Specific model instances are then trained on data from specific assets or processes. This gives our platform the power to apply machine learning across your organisation, without the need for manual configuration per use-case or asset. Underneath this is the power of your data model, the ontology. By defining inputs to the model in the language of the ontology, the model can scale to many locations and can work with varying numbers of datastream across the locations.
Today, various services exist that can help you deploy a machine learning model. However, a single model can not always be used to monitor a large number of assets. Each asset or process has its own pattern of behaviour that requires a separate model instance. This means that the number of models that need to be deployed scales with the number of assets or processes you want to monitor. This poses unique challenges to a machine learning platform, such as training or making predictions for large numbers of models, but also for rolling out releases, and monitoring model performance and data quality. Twinn's machine learning platform was developed to tackle exactly these challenges.
To unlock the power of Twinn`s machine learning platform, all you need is Twinn`s model interface (a Python protocol). All models that comply with this interface can benefit from automated re-training and predictions, rolling out releases, model performance monitoring and data quality tracking.
You can either develop the machine learning model yourself, or with the support of our team of experienced data scientists. We also have an open-sourced Python package that can supercharge the development of your time-series models.
Data Science and Data Engineering Leading Professional