Autonomous Optimization of Uncertainties in the High Pressure Die Casting Process

The High Pressure Die Casting Process

In theory, the high-pressure die casting process is simple: Molten metal is injected into the mould, solidifies a few seconds later and is then ejected as a casting. The mould gets treated with a lubricant to avoid having the casting stick to the mould then closed for the next “shot”.

In reality, this simple-sounding process is highly unstable. For example, a nozzle intended to spray lubricant on the mould surface may loosen and spray over a wider area than intended – or may clog partially or completely, restricting the amount and area of the spray. Missing a sufficient layer of lubrication, castings may stick in the mould, crack during ejection then have to be found within the production batch and scrapped.Similarly, melt volume in the holding furnace at the die-casting machine can change the melt volume dosing into the machine chamber. This in turn changes the metal fill pattern into the cavity by increasing turbulences and the amounts of air that are entrapped in the casting creating porosities. Or consider a basic variable: the temperature of the mould. In order to eliminate pre-solidification during cavity filling the mould has to be brought to working temperature of approximately 400° F / 200°C.

However, the liquid melt injected into the mould heats it beyond this temperature, which, if left unaddressed, would affect the functional life of the mould. In an attempt to regulate this variable, liquid cooling medium containing flowing water or oil is forced through channels in the die steel. Even so, the mould itself changes over the course of a production run; as it heats from room to production temperature, it increases in size; components that held the mould closed in an unheated state may no longer function correctly or completely. Even the condition of the die-casting machine itself changes during a production week; adjustments made on a cold machine yield different scale readings when the machine reaches production temperatures. Variations are present when unexpected breaks occur; the longer the machine stops, the more difficult it becomes for the operator to equilibrate the temperatures again. Furthermore, mould aging affects casting results: the mould wears at different speeds in different areas, affected by melt flow patterns, location of clamps, scale build-up in cooling lines – all of which require the operator to constantly strive to equilibrate the mould temperature.

In short, the high-pressure die casting process is in continuous flux – and the outcome is in the hands of the shop floor personal.

Process Simulation

Blaming only the shop floor personnel for bad castings would be as easy as it is unfair. In the most cases, poor quality results not from the efforts of the floor personnel, but from far earlier in the engineering process: the production process may have been insufficiently developed; the casting shape may have been poor designed. In either case, a good quality casting will never be achieved. This is where process simulation can be of the greatest help.

In the first stage of process planning, casting simulations can be performed, long before cutting die steel or the release of the final casting design. Using CAD files of the early casting design ideas in combination with theoretical process parameters, simulations directly point to potential problems. At this point of development, the casting design and manufacturing process can be changed easily, quickly and inexpensively. With experience, a well-trained die casting engineer using simulation tools can create a process that yields good quality castings during the first die trials.

In the most cases this success is achieved by designing a mould using one set of possible process parameters only. Given the volume of parameters possible in the casting process, and the range of variation within those parameters, the number of potential interactions that the engineer could consider approach infinity – as would the time needed to research those possibilities.With limited time and resources – and having achieved these good castings through single parameter simulation – the engineer may stop at this point, and turn his attention to another project, leaving shop floor personnel to resolve any further production variations.

Autonomous Optimization and Uncertainties

Autonomous optimization is mainly used to find a good process set but can also easily identify dependencies and sensitivities between these process parameters; such as casting and cavity designs, fill and cycle times, mould and melt temperatures, and their levels of variations can be defined within the program and simulated. The optimization software autonomously selects parameter, simulates the set and evaluates the results. By smart selection based on genetic algorithms one of the best parameter set will be found out of the thousands possible variations.

Autonomous optimization should not simulate all possibilities and define the ‘best”; rather, the target is to find an optimum in the shortest time and the least amounts of simulations – and without requiring more than two hours of the engineer’s time (one hour for set up, one hour for result analysis.) As the simulations are done on an office computer and do not influence production, the engineer can simultaneously continue to work on other projects.

In addition, as parameter sets are not lost after the simulation has been completed, they can also be further analyzed at a later time and in greater detail of sensitivities to each other. To start an optimization, parameter values are selected by random or based on designs of experiments. Doing so should provide enough results to calculate sensitivities between the defined parameter even before starting the optimization. Sensitivity analyses are a by-product of the autonomous optimization; using a small design of experiments, this starting array can be increased and analyzed by its own.

Producing good castings in a stable process is an achievable goal; it should not, however, preclude developing further improvements. Admittedly, it is very difficult to search for uncertainties under production conditions, let alone to get consent from management to stop production in order to search for uncertainties, which may not exist at all. Yet what is almost impossible to achieve on the shop floor could be an easy target for simulation tools. For example, controlling a single variable, such as mould temperature, might be feasible, if difficult, on the shop floor – but experimenting with changes to the mould through welding and grinding might be impractical in a working situation. Time and personnel would be required for each alteration, and the cumulative effect of the welding and grinding would reduce die life to the extent that this experiment would not make much sense. Using a computer simulation of the changes, however, allows the engineer to analyze the efficacy without threatening the life of the die. Rather than actually welding the mould, a CAD file is replaced autonomous and the selection of a temperature profile is done with the push of a button.

The use of simulation software is also more efficient than ‘real world’ experiments; rather than shutting down a production line for ‘trial and error’ experiments, one set of parameters can be analyzed through a simulation even as the production of other casting continues, resulting in more efficient production of higher quality castings. Even so, the engineer should consider the time frame involved as well as computer hardware availability, and select the most important variables. Consider the task facing an engineer who wants to optimize a casting process by considering the following variables: mould temperature ranging from 250° F and 450° F in 50° F increments gives five levels of variation; three levels for the use of a lubrication nozzle; four levels for pouring temperature; three for pouring volumes and two for different casting designs results in 360 iterations to simulate. Depending on the casting design and use of a standard desktop computer one simulation could take up to one hour; to process all 360 permutations could take 15 days – still far faster than attempting to do so on a production machine, providing that such an examination is possible at all. Using more powerful computer this advantage leans even more towards simulations. Not to stretch the releases of faster machines in the future. Even today by investing in high-end equipment the discussed calculations could be done in less than 3 days.


Using state of the art computer hardware and the right simulation program, engineers can provide much more information about their high-pressure die casting process even before process development than during production ignoring simulation at all.Before implementing fabrication, processes can be optimized, influencing parameter can be found, dependencies can be defined and tolerances can be implemented based on these dependencies and variations in influencing parameter. With this knowledge at hand and the right monitoring systems, production processes are well understood and able to produce good castings under stable conditions.

Source by Ralf Kind

Disclaimer: Active links in article are done by The article author did not include or endorse these active links.

#Autonomous #Optimization #Uncertainties #High #Pressure #Die #Casting #Process

Scroll to Top