Artificial neural networks
Artificial Neural Networks (ANN) are mathematical algorithms. They simulate specific functions of a nervous system: ANNs can learn; “reach conclusions” based on data (observations) and detect patterns, as well as make predictions about their environment. ANNs are networks made up of simple equations which act like model neurons.
A neural network is therefore a mathematical equation which uses data records to learn typical patterns that cannot be detected (in the data records) by normal observers.
For example, if neural networks are trained with data from a process, they learn how specific process variables lead to specific efficiency levels or qualities of a product. This knowledge can be queried in a targeted manner in order to anticipate quality analyses (soft sensors), learn more about a process (process analysis) or even to optimally control the process.
There are hundreds of different neural network types. Each type requires the operator to have expert knowledge. Only very few experts are able to optimally select and use neural networks. Configuration errors result in failures. In the past, these limitations have lead to a bad image for this type of technology.
atlan-tec already developed a revolutionary new concept as early as 1996: the automatic and self-configuring NeuroModel® neural network. It can do automatically what must be elaborately redeveloped every time for other tools in the market.