Home 9 Resources 9 FAQs

Extensive Transparent AI Solutions

App FAQs

Here you will a number of frequently asked questions from our customers. We’re always happy to answer any questions you may have about our trusted and transparent AI solutions.

Why is the BASF Intelligent Mine relevant to the industrial world?

Processes with high-variability inputs, such as is found in the mining sector, are constantly battling to compensate for changing conditions and produce a consistent product. The Brains.app and its Optimization as a Service (OAAS) applications shed light on the unknown through data modelling and AI, so operators can make better decisions and processes can proactively adjust to changing conditions.

What are the industry standards that the brains.app platform complies with?

Brains.app supports the following standards:

  • Data Acquisition: we support a portfolio of data acquisition standards that range from:
    • IOT Devices: MQTT is the ubiquitous messaging standard generally supported by all IOT devices. brains.app platform supports XMLRPC which is the underlying technology for MQTT. To date no customers have asked for MQTT support.
    • Industrial Data: OPC DA and UA protocol is the ubiquitous data standard followed by all industrial systems that brains.app platform supports. All client situations to date has asked for this requirement. This is typically from OSI Soft Pi systems.
    • Web services: REST is the ubiquitous web data standard that is supported
    • Batch data: we support all types of batch data like XML, CSV, XL and manual data uploads.
  • Data Modelling: we support various data modelling libraries like Tensorflow: An Open Source Library used for numerical computation & large-scale machine learning. This can be used to create dataflow graphs and run on any target (cloud or local)
  • Deployment Architecture: we use Docker containers and Kubernetes as the container orchestration layer to deploy our platform + application in any environment: cloud, client data centre or local on-site deployment.
Can customers use brains.app with their own data platforms?

Yes, the brains.app platform is a modular platform. We work with various customers who have their own data platforms, and we have separate layers between algorithms, data lake and end-user interface.

All of these layers are modular, connected by micro services and standardised which allows us to be cloud (host) agnostic, data lake agnostic and deployment platform agnostic. brains.app can be deployed as an individual docker container on a different data platform or cloud host.

In addition, we are working with various customers who have internal data lakes built on MS Azure or other cloud hosts.

How does the brains.app help me better implement the value of IoT?
By creating a secure, multi-access interface to view, analyze, interpret and change functions across a range of industrial activities, the brains.app platform de-silos data and gives users all the power they need to improve productivity and efficiency.
How are your models different?
We combine Physics (first principle) models with machine learning / AI models that draw patterns from the data using Neural Networks (combination of generic black box and first principle/equipment models). Our models learn and adapt in real-time to evolving situations. Models are updated and upgraded regularly (quarterly at least) to adapt to a specific system e.g. site modifications.

We are one of the only companies in the mining industry to combine leading-edge machine learning / artificial intelligence based models with physical models to create digital equipment and process models. These models are used for process predictions that learn in real-time and adapt to changing conditions as are relevant for dynamic mining environments. We use simulation based optimization enabling our continuous optimization algorithms to be tested in a process simulator which also doubles up as a training simulator.

How do we achieve the efficiency?
We use the predictive control philosophy to predict process performance and then apply both financial and technical optimization constraints to recommend optimum set points on a continuous basis. By incorporating upstream and downstream process information, our optimizer is constantly looking for the optimum set point ranges for specific processes and equipment factoring in geometallurgical properties of materials and residence time of the materials.
How open are the models? Can my team open them and inspect why a prediction was made?
Our platform comes with API’s and SDK’s that will allow staff from your team to be able to work with the models and also understand some of the decisions. They can retrain the models with new data as well as build on it. However, before any model changes are implemented on our platform they need to go through a testing and certification process before our platform would deploy and run them. IntelliSense.io would prefer to train your staff to understand the models, retrain them and use the platform to build any new models they want using our API and SDK’s. The models that have been developed and any new ones we develop, always remain our property. If you build any new models that address site specific problems, then you would own that model. But in order to run that model in real time in our platform, you need to use our SDK and API’s. You are not permitted to copy any of our existing models.
How do you know that your recommendations and/or predictions are correct?
The Intellisese.io mining application portfolio considers the material properties being fed by the material transport and influence models as well as current operational status of key equipment to accurately predict future performance. The apps use a combination of Machine Learning and Physical approaches to give full confidence and verification of outputs. The optimization engine is generated with and constantly reviewed with input from operations personnel on site, with full transparency of the uncertainty levels of all recommendations to build trust. In addition, a process simulation environment is provided which allows engineers to test various operational scenarios and for training purposes.

We continually calibrate our products including the models during the commissioning phase to ensure predictions are matched to actuals, giving confidence on the predictions once the system is live.

Does the app work on iOS or android, and can it be accessed remotely?

The brains.app runs on any browser type and on any screen resolution and size. Screens are fully reactive and can be manipulated (zoomed) to change the amount of content shown. All that is needed to access brains.app is an active internet connection.

Implementation FAQs

Which areas of mining operations do you cover?

Our focus is on the mine to market value chain with applications segmented across

  • Digital Mine
  • Digital Plant
  • Digital Markets

Our Applications include:

  • Digital Stockpile
  • Grinding Circuit
  • Floatation Circuit
  • Thickener Circuit
  • Heap Leach
  • Pipeline Pumping
You mentioned that your solution does not require additional instrumentation, what happens if we are missing critical devices?
Depending on the application, some of the instrumentation/data that we need can be substituted with virtual sensors to give a best approximation of the missing data stream.

Whilst not perfect and will result in models that have a larger uncertainty band, the result is still better than no data/no instrument. Clearly there are limits to this approach, but you would be surprised how much can be achieved. These virtual sensors can be regularly calibrated by using physical sensors.

What happens if our data is poor or there isn’t much data available?
IntelliSense.io approach is to use a combination of Machine Learning (AI) that relies on data and Physics (first principle) to create, test, and calibrate the models. This allows us to ensure even if any of the site-specific data is of poor quality, we can deliver value. This value can be in terms of Virtual Sensors which provide hidden states in your process that you don’t know today.
How is this different to advanced process control (APC)?
Predictive and Prescriptive AI control improves operational stability by predicting upstream feed changes before impacting downstream processes, while at the same time channelling back the optimized state of these processes to reduce plant feedstock variability, all in real-time. This is something APC cannot achieve.

General FAQs

Why did BASF partner with IntelliSense.io?

Artificial Intelligence (AI) is transforming the mining industry. We are seeing mining processes change significantly when computer-driven technology is used alongside equipment. Our partnership has combined mining process experts, software engineers, chemists, and data scientists to find targeted solutions to support your process or known problems. BASF’s Intelligent Mine, powered by IntelliSense.io, was established to bring people and technology together to improve planning and to enable mine and plant operations to become more efficient, sustainable, and safe. We offer a well-rounded approach through AI solutions that provide you real-time prediction, simulation, and optimization recommendations. Unlike anything in the market today, our ‘Ready-to-Go’ Applications generate value within the first few weeks of implementation.

What are the challenges of the Internet of Things?

There are three primary challenges that the Internet of Things faces:

  • Data tends to be stored in silos, which makes broad-based access and overview challenging
  • Security: by its nature, the Internet of Things requires connecting multiple data points and IT functions together to produce an inclusive and non-silo capability or functionality. This challenges numerous standard security protocols and requires a complex permissions-orientated security system that ringfences the correct data points and IT functionalities without compromising efficiencies and capabilities.
  • Accessibility via numerous devices, multiple users requiring varying types of functionality, data, reporting etc. creates another challenge for IoT. The system needs to be able to adapt to a constantly fluctuating user landscape.

BASF solves these challenges by harnessing the power of Cloud-based technologies and Machine Learning to provide a robust real-time decision making AI Applications platform.

What is OaaS (Optimization as a Service)?

Our solutions, including software, pre-build AI models and support sold as a service through an annual subscription. This negates the need to buy any hardware, software licenses or services to develop custom models, therein reducing upfront (CAPEX cost) and eliminating the need for expensive service costs for developing and deploying AI projects.

The applications are delivered out of the box and kept up to date on a continuous basis (including the retraining of the models) to ensure the application delivers value to users and scales with them. The standard name for this business model is ‘Software as a Service’

Benefits of Business Model ‘OaaS’

  • Quick to deploy and scale without requiring upfront CAPEX
  • Faster benefit realization
  • Guaranteed levels of service
  • Long term relationship with models adapted and software updated to site needs and changes.
  • Accessible anywhere and no additional fees for upgrades.
How can you optimize a system when you are not experts on the system?

We are not equipment experts but we have process expertise in the team in both mine and plant. We also use data and models to learn about the system and, by using algorithms to evaluate performance, predict future performance and make recommendations.

We have expertise on the specific mining process and use data to derive non-linear correlations across the end to end processes. This allows us to get a better understanding of the entire system, this system knowledge is used to optimize the process. Leveraging this system knowledge we develop prediction models that predict process behavior, we then use this process prediction and apply financial as well as technical constraints to deduce the optimum control set point for the equipment. These control set points are delivered on a continuous basis.

We see the fundamental breakthrough is in providing a system-based optimization by combining process expertise and insights generated from data, then deploying them through a real-time decision support system to support both manual and automated decision making.

How are you different to the traditional Industrial automation companies?

We are not a controls company, our focus is completely on delivering software-driven continuous optimizationmodeling for processes by incorporating upstream and downstream process information. We undertake this by modelling the entire system across the pit to port process and then using our prediction models to predict future process behavior across the system.

Traditional process control models degrade over time, aren’t flexible to add new variables and equipment and aren’t dynamic requiring plant intervention to be calibrated while IntelliSense.io models are self-learning, flexible to add new equipment and variables without requiring any process intervention.

Honeywell systems require you to upgrade to new versions of the software to get new features, also requiring hardware while IntelliSense.io follows a NO CAPEX policy requiring no upfront hardware cost and also provides new features free of charge as part of the annual subscription.

We are also control system agnostic and can work with any existing controls platform (Honeywell, ABB etc). This allows us to ensure your existing control system can be used to run the process at optimum efficiency. We add new functionality to our software every quarter that is included as a part of our annual subscription whereas another company would most likely want to sell you new upgrades for new functionality.

BASF sells Chemicals, is your digital offering linked to the sale of your chemistry?
No, the BASF Intelligent Mine is agnostic to the brand of chemistry used. Our Applications use Scientific AI, a fusion of physics models and machine learning techniques, resulting in products that can be applied to any mineral commodity, process flowsheet or OEM equipment. Our products are designed to get the best out of what is available. Having said that, we know the importance of chemistry in mining operations, allowing us to incorporate unique know-how into our digital model library for maximum value.