HPC Workflow for Simulation and Optimization of Additive Manufacturing for Improving the Production of Gearboxes
Mastering Additive Manufacturing (AM) processes is rather an artisan art than a straight-forward engineering process, requiring a lot of experience to avoid trial-and-error-loops and to get it first-time-right, starting from design, covering the manufacturing engineering stages and ending with finding the appropriate AM machine process parameters for optimum product quality.
In this Application Experiment new simulation models for the Additive Manufacturing process of gearboxes are developed and cloudified to enable AM-based production of cycloidal gearboxes whose production costs are limiting its widespread use in automation, manufacturing machines and many other products.
Thus, the industrial relevance is twofold:
- improved simulation technology for the huge field of AM-based production of metal objects / parts (estimated market size for AM simulation technology: € 1 billion) and
- affordable AM-based production of cycloidal gearboxes with a multitude of applications (estimated market size for cycloidal gearboxes: € 600 million).
Currently, the complete engineering and production process of cycloidal gearboxes is based on subtractive manufacturing techniques and takes about four to six months and approximately € 10,000 of labour and material costs. The envisioned process uses cloud-based simulation and Additive Manufacturing techniques. While Additive Manufacturing will decrease the machining time including its preparation, the simulation of the additive process and especially the thermal effects will prevent costly trial-and-error-loops before usable AM-parts are produced. Thus, reducing the overall duration and render the production more flexibly and ‘real-time’.
The challenges of this experiment are manifold: the thermal effects of the AM process need to be understood, modelled into a simulation model which then is to be executed in reasonable time leveraging Cloud/HPC resources. The approach to master these challenges is to use existing simulation technology (FEMPAR - a multiscale, multilevel and multi-physics finite element solver), cloudify it to run on HPC resources via a Cloud-middleware-based interface to exploit the full power of scalable resources and augmenting it with a simulation model that captures the physics of the thermal process of a metal printing machine.
At the end of the experiment, the end user STAM has achieved the following technical impact:
STAM has now the possibility to take into consideration AM as a possible manufacturing approach for gearboxes, since the experiment results have demonstrated that the simulation time can be reduced by around 65 percent (18.5 hours for the simulation of a single part on standard workstation vs. 5.5 hours for simulating the same part using HPC), then the huge obstacle of long simulation time for the non-cloudified software is overcome.
The scrap material reduction was calculated via CAD model. It strongly depends from the shape of the object to be manufactured. Referring to the cycloidal gearbox parts (the case study), it was calculated that averagely more than 30 percent of scrap material is saved (e.g., 66 percent for the input shaft, 25 percent for the output shaft).
- The torque density can be increased thanks to AM approach, which allows the adoption of some solutions such as printing the output shaft as a one piece (the cost of such a method with traditional SM techniques would be huge). Indeed, reinforced connection between the pin and the plate can be printed, and the transmittable torque increased by 20 percent. This result was calculated via analytical formulas, taking into account the connection radius and the pin section radius
- The Time-to-Market of a new gearbox can be reduced by 30 percent: in the baseline situation, this time is around 6 months. The experiment showed how this can be reduced to by 2 months, mainly thanks to the engineering workflow time reduction (thanks to avoiding the trial-and-error approach) and production time reduction through AM.
Summarizing all the technical impacts, STAM can increase its competitiveness in the transmission systems industry for high tech applications due to the fact that STAM can become the number one source of mechanisms for special applications at a more affordable price in the cycloidal gearboxes market.
The ISV CIMNE will further advance FAMPAR-AM for additional industrial applications. The ambition is to become the reference simulation software provider for Additive Manufacturing. The experiment progresses FEMPAR- AM to become a Software-as-a-Service. Experience gathered in this process can be used to bring additional software tools to the CloudFlow infrastructure. CSUC as the HPC provider will benefit from the experiment by simplifying HPC usage, increasing the flexibility of solutions, acquiring new knowledge with respect to remote visualization and management of graphical nodes and provide them as powerful solutions to additional customers.
The total costs of engineering and production of a new gearbox can be reduced thanks to CloudFlow method by 30 percent. This was calculated taking into account the engineering time reduction (33 percent), the material reduction (30 percent), the avoidance of trial-and-error approach. Given that the € 10,000 cost of the baseline solution is mainly due to the heavy engineering workflow, the 33 percent time reduction is estimated in € 2,000 of cost saving. The remainder € 1,000 is due to the material saving and the avoided discarded parts thanks to the optimized production. This reduction will allow for lowering the price of the innovative cycloidal gearboxes, consequently increasing the competitiveness of STAM. Increased flexibility provided by AM as the manufacturing method will allow for opening new markets. Market expansion and new manufacturing opportunities will contribute to creating new jobs to support the growth brought by these opportunities. Existing resources will work to foster the follow-up activities of the project and allow STAM to use AM and simulation as the standard manufacturing workflow for cycloid gearboxes. Three years after the experiment completion it is expected to increase number of jobs by three employees (1 person hired as AM manufacturing and simulation specialist, 1 person hired as application manager and 1 person hired as sales and business development specialist).
The typical customer of cycloidal gearboxes is an SME or Large Enterprise developing machines for industrial automation, robotics and space. The market size of such companies is estimated € 600 Mio. One year after the experiment STAM aims at having 10 companies as customers in this market (approx. 0,015 percent of the market size), three years after experiment completion, STAM wants to triple their market share from 10 to 30 companies (approx. 0,05 percent of the market size). The type of product and associated service STAM provides in the field of cycloidal gearboxes is for special high-tech applications that cannot rely on series production suppliers. For this reason, each client usually requires the production of 1 to 2 gearboxes so we expect to sell roughly 12 to 15 gearboxes after one year and 35 to 40 after 3 years. STAM estimates the additional revenues of the company due to exploiting the experiment’s results to be € 100,000 after one year and € 320,000 after 3 years, allowing them to create three new jobs in the focus area of the experiment.
From STAM’s customers’ perspective, cycloidal gearboxes are a key technology for a number of applications, since such devices allow them to obtain a high ratio between the input speed (e.g., an electric motor) and output (e.g., an end effector), which is translated into a high increase of the transmitted torque. Moreover, the high efficiency and compactness of cycloidal gearboxes make them suitable for some particular applications, such as aerospace and robotics, while the resistance to torque peaks and shocks makes them appealing for heavy industries such as wood working and steel making. If AM is successfully adopted thanks to CF simulation, the customers of STAM will benefit from a higher product quality (see increased rated torque), which is sold at a lower price (see cost savings).
The software provider CIMNE will offer virtualization of the manufacturing process permitting lower design costs through numerical simulation. The exploitation of parallel computing resources of HPC providers in a user-friendly and flexible way reduces computing time for users who cannot afford buying such resources otherwise. CINME wants to increase their user community and become the reference as software providers for AM (increased market share, by 20 percent in the first year after the experiment and by to 50 percent after 3 years). The expected total income is € 14,000 for the first year and € 64,000 on a 3-year perspective which is enough to cover the production and commercial related costs estimated in € 12,000 for the first year (0,25 person) and € 40,000 on a 3-year perspective.
CSUC as being the HPC centre in this experiment will offer new and better services contributing to the strategic goal of CSUC to increase the use of HPC by SMEs. Easy access via the CloudFlow portal allow CSUC to implement fast payper- use access to their HPC services with an expanded service portfolio thanks to the remote visualization service implemented in this experiment. CSUC is estimating the local (Catalonia) market size for the experiment results to be 10 customers (one year after the project) and 20 customers 3 years after the experiment, when (a) market is more mature in terms of HPC simulation services usage and (2) developed tools have been improved through the use by industrial customers, who at those first stages will help to increase the usability of the platform, to make it easier for new customers that therefore will have less barriers to become users of HPC simulation services. After 1 year, CSUC estimates to gather 3 new companies (30 percent of the market) and growing up to 10 companies (50 percent of the market) after three years. The income (revenue) is estimated to be € 90,000 after 1 year and € 180,000 after 3 years: the envisioned income per customer at first stages is higher than on late stages, as the services need to be automatized and so the cost will become more affordable (and attractive) for users. This will allow for creating one additional job position after 3 years.