Using Machine Learning When Pouring Molds

Every day foundries are faced with loss of knowledge due to the retirement of experienced operators and the difficulty of replacing them with new talent. Automation technologies that reduce operator intervention in critical process is more necessary now than ever. The first step toward that goal is to find a reliable automatic pouring system.

The pourTECH™ System is considered the best pouring process by foundrymen all around the world. Once the job parameters have been set up, pourTECH™ uses real-time laser feedback to control the flow of metal into the mold, minimizing short pours and over-pours.

In addition to pouring, the pourTECH™ system performs a number of functions, such as automatic positioning of the nozzle over the pour cup, in-stream inoculation, inoTECH™ inoculation verification (monitoring the inoculation as it hits the iron stream), and non-contact pour-stream temperature measurements using non-contact pyrometers.

The user sets up limits for inoculation hit rates and temperature ranges, and the system will alert the operator if a limit is being approached or stop the pouring if it is exceeded.

As with any automatic system, the pourTECH™ system parameters need to be tweaked when the pouring conditions changes due to slag build-up, bath level and temperature changes, etc. And because the system requires only minimal operator involvement these slow changes can go unnoticed, especially when new operators are introduced to the system.

Viking Technologies (USA) and pour-tech AB (Sweden) have developed a new software module based on machine learning to take control of the pouring once the basic parameters have been set up. The EASYpour™ software allows the operator to focus less to the pouring process and devote more attention to supervising the overall pouring process, such as taking samples, filling inoculants, and performing other important tasks.

With EASYpour™, the operator has a small number of parameters to set before the system takes over.

EASYpour™ uses data stored in the pourTECH™ database to analyze previous pours, in order to optimize the next pour. The database stores a wide variety of pouring and molding machine data for each pour, also making it invaluable when analyzing past production results.

One critical piece of EASYpour™ information comes from a secondary device, the “Final Level Laser”, placed a short distance after the pouring station to determine the de facto final level in the mold after all cavities have filled and the metal level in the cup has stabilized. This information is used by EASYpour™ to ensure complete filling of the mold. This feedback allows EASYpour™ to fill the mold in the shortest time possible, without overpouring.

The graph shows a pour-time reduction of around 1 second, for one particular mold. For this job, the mold rate went from 313 to 343 molds/hr. - a productivity increase of nearly 10%.  

The pour-time reduction will vary from job to job, with some jobs already filling at close to optimum time before EASYpour™ is activated, and others benefitting tremendously from EASYpour™.

Adding the Final Level Laser will provide still more quality-improving functions. It will detect a short-filled mold caused by an aborted pour, a mold run-out, or any other unwanted result. Together with data such as the inoculation setting, inoculation hit-rate and pouring temperature, the final level information completes the quality data for each poured mold.

With more and more new recruits operating the pouring systems, it is important to provide a tool that helps the foundry meet its goals. By constantly adapting the system control parameters, EASYpour™ provides this tool.