Fault detection logo

Fault detection using neural networks

By Pablo Martin, Artelnics.

Automatic fault detection concerns with monitoring a system, identifying when a fault has occurred, and pinpointing the type of fault and its location. This is always a big concern in industry production. The higher quality a production is required to have, the better fault diagnosis method the factories should apply.

Early discovery of system faults may ensure the reliability and safety of industrial systems and reduce the risk of unplanned breakdowns. Usual fault detection algorithms have limitations due to the growth of complexity of modern systems and the amount of information that is collected.

Artificial neural networks, due their capacity to lead with large amount of real-time information and to model complex relations between different variables, are a strong tool to find any anomaly that may arise.


Discover diseases in the historical data
Detect whether an item is working well or not by comparing the data received from it with the historical data.
Identify the causes of fault. This process should consider trends in health history and operational context.
Predict the state of the item in the future to detect any possible fault in advance.
Elaborate maintenance plans taking into account the previous predictions to reduce fault.

As industrial systems are increasingly equipped with a lot of sensors that collect information about the state of the items, being able to manage all this information to implement solutions in real time will allow industry to increase benefits through a more accurate maintenance process.

Industry picture

Neural Designer uses neural networks to analyze and build a predictive model based on the information provided by a dataset that contains the historical functioning of the industrial systems.

Neural Designer can establish which are the variables that characterize a fault, study the dataset attending to a set of statistical indicators, design and select the parameters of the predictive model, train it and check its prediction capacity. Finally, it also allows to export the mathematical expression that defines the trained model to use it in real cases.

The next image shows a representation of a neural network that could be used for this case. As inputs, it will receive information about the functioning of the system, environmental variables and other external variables that may be considered important for the analysis. As output, the neural network will response with the most likely fault.

Neural network

Related solutions:

> Predictive maintenance.
> Performance optimization.
> Quality improvement