Related solutions:> Performance optimization.
> Fault detection.
> Predictive maintenance.
Traditionally, the way in which quality have been measured and improved has been inspecting the product after manufacturing. The result of this process is a report of the number of possible problems that appeared, that is, a description of past problem frequencies.
Although this technique is still valid, the industry is moving to another methods that not only describe the problems that appeared in the final products in the past a make recommendations based on them but also that aim to predict any possible default beforehand in other to avoid it.
Due to the increase in the number of sensors that collect information about the manufacturing of the product during all the phases, machine learning, and specifically neural networks, is getting more importance in the management of this information to obtain predictions about the quality of the product even in the early phases of production, such as the design phase.
These new methods have a lot of benefits for industry. They speed up the detection of any quality problem and, as a consequence, they allow early detection and a more adequate preventive maintenance which reduces the costs of repairing them.
Neural Designer, a software that implements neural networks, is a tool that is able to study the historical data and provide information about the quality problems that appeared and study the relation that they have with different factors.
In addition, its clear interface allows to easily set the parameters to design the predictive model and to train it using the configuration of the neural network that fits your problem at best. This software also provides a set of testing methods to analyze the performance of the trained model on new data.
The next image shows a summary of the process that would be followed to obtain a predictive model for this case. Firstly, the data about past quality problems are arranged in a database. This database is analyzed by Neural Designer to find patterns. As a result, we obtain a predictive model.