feedability analysis and optimisation driven by casting...
TRANSCRIPT
April 2007
Technical paper submitted to the Indian Foundry Journal
Feedability Analysis and Optimisation
Driven by Casting Simulation
Dr. B Ravi, Professor,
Mechanical Engineering Department, Indian Institute of Technology, Bombay, Mumbai 400076.
E-mail: [email protected]
Durgesh Joshi, Lecturer,
Department of Industrial and Production Engineering, Shri GS Institute of Technology and Science, Indore
E-mail: [email protected]
Abstract
Casting simulation programs can be used by foundry engineers for quality assurance
and yield optimisation without shop-floor trials, as well as by product engineers for
analysing and optimising the feedability of a casting during design phase itself. For
widespread application, the simulation programs should require little domain
experience, and they should be fast, reliable, easy-to-use, and economical. These
goals can be achieved by automating some of the tasks involved: identifying the
location of a feeder, calculating its minimum size, creating its solid model, attaching
the feeder model to the casting, carrying out solidification simulation, predicting
quality and estimating the yield. This paper describes the underlying algorithms, and
their implementation in a software program. Two industrial cases of large steel
castings show how simulation facilitates troubleshooting and elimination of defects by
minor changes to part design coupled with improved methoding of casting.
Keywords: Casting simulation, Design for Manufacture, Feeder, Solid Modelling,
Quality Assurance.
1. INTRODUCTION
There is an increasing penetration of computer applications in Indian metal casting
industry, as seen in a 5-year survey covering 200 participants from 145 organisations
including foundries, OEMs, tool rooms, R&D institutes, and CAD/CAM service
providers [1]. The survey was carried out during a continuing education programme at
IIT Bombay conducted every September during 2002-2006. The survey showed that
Internet, CAD/CAM, and computer-aided planning are being used in 90% or more
organisations (Figure 1). Nearly 25% of the organisations also reported using casting
simulation, either in-house or through the services of a CAD/CAM agency.
Fig. 1. Penetration of computer applications in Indian casting industry
Casting simulation essentially replaces or minimises shop-floor trials to achieve the
desired internal quality at the highest possible yield. A number of casting simulation
programs are available today, such as CastCAE, MAGMA, Novacast, ProCAST, and
SolidCAST. Most of them use Finite Element Method to discretise the domain and
solve the heat transfer and/or fluid flow equations [2]. The main inputs include the
geometry of the mould cavity (including the part cavity, feeders, and gating channels),
thermo-physical properties (density, specific heat, and thermal conductivity of the cast
metal as well as the mould material, as a function of temperature), boundary
conditions (such as the metal-mould heat transfer coefficient, for normal mould as
well as feed-aids including chills, insulation and exothermic materials), and process
parameters (such as pouring rate, time and temperature). The results of solidification
simulation include colour-coded freezing contours at different instants of time starting
from beginning to end of solidification. This provides a much better insight into the
phenomenon compared to shop-floor trials (real moulds being opaque). The user can
verify if the location and size of feeders are adequate, and carry out iterations of
design modification and simulation until satisfactory results are obtained.
Sometimes, it is not possible to achieve the desired quality by changes to method
(mainly feeding and gating) alone. In such an event, it may become necessary to
modify the part design. For example, the wall thickness of the part can be increased
(referred to as padding) at locations that choke the flow of feed metal. Another typical
modification is adding or increasing taper to promote directional solidification. These
modifications however, imply additional machining cost. If feedability analysis is
carried out at the product design stage itself in a systematic manner, it can potentially
lead to superior product-process compatibility. This implies ‘foundry-friendly’
castings, making it easier to achieve the desired quality with high yield [3].
For wider application of casting simulation in foundry as well as OEM firms, the
programs should require little domain knowledge, and must be fast, reliable, easy-to-
use, and economical. For example, the location and size of the feeders is an important
input for solidification simulation. This decision requires considerable methoding
experience from the user. Further, the engineer has to create or modify the solid
model of the feeder, attach it to the casting model using a CAD program, and import it
into the casting simulation program for each iteration. These tasks require computer
skills. The accuracy of results (such as solidification time and location of shrinkage
defects) is influenced by metallurgical models and availability of temperature-
dependent material property database. The simulation of intricate castings may
involve more time and cost than shop-floor trials, and any error in program inputs
implies further delay and expenses.
To overcome the limitations mentioned above, an integrated software program called
AutoCAST has been developed, by combining automated methoding and casting
simulation in a single environment. It is a collaborative effort between the Indian
Institute of Technology, Bombay, and Advanced Reasoning Technologies Pvt. Ltd.,
Mumbai [4]. The overall framework is described first, following by two major
algorithms in the program: automated feeder design and feedability analysis. Two
industrial examples of casting troubleshooting and optimisation are provided at the
end to illustrate the technology and methodology involved.
2. CASTING METHOD AND SIMULATION FRAMEWORK
The main inputs to the casting method and simulation program include: (1) CAD
model of the cast part in standard STL format, (2) cast metal or alloy name, and (3)
type of casting process. The method involves three major decisions: (1) orientation
and parting line, (2) core print design, (3) feeder design, and (4) gating design. The
major results include: (1) method data, including solid models of cores, feeders and
gating channels, (2) internal quality, in terms of shrinkage porosity, inclusions, etc.,
and (3) casting yield and cost estimations.
This paper focuses on feedability analysis and optimisation driven by solidification
simulation (Fig. 2). It starts with the user importing the CAD model of the as-cast
part. The feeder design is carried out by identifying a suitable location to connect the
feeder, computing its dimensions, creating its solid model and attaching to the part
model. Then feedability analysis is carried out by solidification simulation followed
by checking for the presence of hot spots inside the casting as well as connections
between the feed paths. If internal defects are predicted, then feeder design is
modified. If it is not possible to improve the quality by feeder design alone, then the
part design is modified. Finally, yield is computed and compared with the yield
obtained by other layouts with acceptable quality, and the one giving the highest yield
is selected for implementation. The steps are described in detail next.
Fig. 2. Framework for feeder design and optimisation
3. AUTOMATED FEEDER DESIGN
Automated feeder design provides a good-first starting point for method iterations,
and is especially useful for product and tooling engineers. It also saves valuable time
for foundry engineers, since feeder design and modelling requires a combination of
both foundry knowledge and CAD skills, and becomes a bottleneck in methoding and
simulation iterations.
There are three major functions in automated feeder design. The first is finding a
suitable location on the casting surface to connect the feeder. The second involves
computing the appropriate dimensions (diameter, height, neck, etc.) of the feeder. The
third function is to create a solid model of the feeder and connect it to the casting,
ensuring a proper matching between the surfaces involved. These functions are
described here.
The feeder location is driven by the following algorithm derived from the knowledge
of experienced foundry engineers:
1. Identify the faces of the cast part around the hottest spot.
2. Discard bottom-facing faces (that is, keep only top and side faces).
3. Sort the faces in a list according to their distance from the hot spot.
4. Choose the nearest face if it satisfies the selection criteria.
5. If not, continue checking with other faces in the list.
The three criteria are given below. All of them can be used simultaneously after
providing weights to them to facilitate evaluation.
(i) Flatness of the face, since it is difficult to fettle the feeder connected to a
curved surface of casting.
(ii) Orientation of the face: either top faces (feeding is aided by gravity), or
side faces (feeder can be connected to gating or multiple castings).
(iii) Wall thickness, since a larger wall thickness of casting at the feeder
connection point prevents damage during fettling.
The algorithm is implemented as follows. First, the solidification simulation of the
casting is carried out without any feeder (referred to as Layout 0), to identify the last
freezing point (hottest spot). Then rays are projected from the hottest spot in all
directions, to intersect the visible faces of the casting. The faces are sorted depending
on the ray length, and the most suitable face to connect the feeder is identified using
the algorithm given above.
Some castings may contain multiple hot spots, and may require more than one feeder.
In such cases, the first feeder is designed for the hottest spot. Then solidification
simulation is carried out to identify the next hottest spot, for which the second feeder
is designed, and so on. In the case of symmetrical castings (such as a wheel), the
number of feeders can be estimated by analysing the feeding distance (described
later).
The feeder dimensions are computed using geometric modulus method. The default
feeder shape (for example, cylinder) is used for this purpose. The geometric modulus
of the casting around the hot spot is computed by the same ray-shooting technique, by
computing the volume enclosed between the rays, and the cooling area of the surfaces
enclosed between the points of intersection with the casting surface. The ratio gives
the geometric modulus. This is multiplied by a suitable factor to obtain the geometric
modulus of the feeder. The factor mainly depends on the metal (default factor for steel
is 1.3, ductile iron is 1.15, grey iron is 1.0. Then the feeder aspect ratio (height to
diameter) is set depending on the metal (H/D=2.0 for steel, lower for other metals).
Using the aspect ratio and geometric modulus, the diameter and height of the feeder
are computed.
The dimensions of the feeder neck are very critical, and depend on the relative
geometric modulus of the hot spot, feeder, and type of metal. For this purpose, the
geometric modulus of the feeder is first computed from its dimensions as computed
above. The neck modulus is given by the average of the modulus of hot spot and
feeder, applicable for most alloys. For grey and ductile iron, the neck modulus is
lower, to prevent ‘reverse feeding’ during graphitic expansion phase.
Finally, a solid model of the feeder is generated by creating its vertices, edges, and
faces based on the connection point and dimensions. The model of the neck is also
created in a similar manner. The shape of the feeder neck is modified to match the
curvature of the connected portion of the casting (Fig.3). The volume and weight of
the feeder along with its neck are also computed for yield estimation.
Fig. 3. Automated feeder design, solid modelling, and attachment
The user may modify the connection point of the feeder before its design, or
dimensions of the feeder before its solid modelling. More feeders can be added, if
necessary. Feed-aids such as insulating and exothermic sleeves and covers, and chills
are semi-automatically designed and modelled based on the location suggested by the
user. This is driven by the computation of amount of heat to be transferred from the
local volume of casting, and the thermo-physical properties of the feedaid material.
4. FEEDABILITY ANALYSIS BY SOLIDIFICATION SIMULATION
The feedability (ease of feeding) of a location inside the casting is characterized by
low temperature, coupled with high gradient and low cooling rate. Conversely, the
most probable locations for shrinkage porosity inside a casting have high temperature,
low gradient and high cooling rate. High temperature (could be a peak, a ridge or even
a plateau) signifies fewer directions from where liquid metal can flow in to
compensate for solidification shrinkage. Low gradient implies that even if liquid
metal is available at a neighbouring region, there is insufficient thermal ‘pressure’ for
the flow to actually take place. High cooling rate implies that even if liquid metal and
sufficient gradients are available, the time available is too short and the liquid metal
freezes before reaching the hot spot. This is also supported by the criteria proposed by
Niyama and others, which state that for good feeding (in steel castings), the ratio
G/R0.5 must be greater than one.
The feedability of the entire casting, for a given methoding layout, can be determined
by combining the results of temperature mapping as well as feed metal flow tracing.
Feedability is considered to be high (low probability of shrinkage defects inside
casting), if the following two conditions are satisfied:
1. The casting does not contain any isolated hot spots. The hot spots show up as
islands in temperature maps.
2. All feed metal paths connect with each other and lead inside a feeder. Any
breaks in feed paths, especially if the local temperature is high, imply poor
feeding at that location.
Casting solidification is simulated using a hybrid method combining Vector Element
Method (VEM) and Finite Difference Method (FDM). The VEM traces the feed metal
paths in reverse to pinpoint the location and extent of shrinkage defects such as cavity,
porosity and centreline shrinkage [5]. It is based on the principle that the direction of
the highest temperature gradient (feed metal path) at any point inside the casting is
given by the vector sum of individual thermal flux vectors in all directions around the
point (Fig.4). Multiple hot spots, if present, are detected by starting from several
directions. Ideal feeding implies that all feed metal paths meet and converge inside a
feeder. The FDM relies on discretizing the domain of interest (mould, cast part,
feeders, gating channels, feedaids, etc.) into small elements, applying the boundary
conditions corresponding to each material and interfaces between them, and solving
the differential equations for heat transfer. In the hybrid method, the mould domain
along with other elements mentioned above is divided into at least one million cubic
elements (100 per axis), boundary conditions are generated, and the VEM is used for
computing temperatures (to identify hot spots), gradients (to identify feed paths), and
solidification time.
Fig. 4. Feed metal path tracing using Vector Element Method
Since feedability is influenced by part design as well as methoding, the above
approach to feedability analysis driven by casting simulation is also very useful for
evaluating and improving castability by minor changes to part design. The following
case studies show how part design without considering castability aspects led to
internal defects, and how they were eliminated by a combination of changes in
methoding as well as part design.
5. INDUSTRIAL CASE STUDIES
The case study involved a steel casting of magnet frame tube of overall size 750 mm
weighing 780 Kg. It is the base part of the housing of traction motor, manufactured by
a public sector unit. The castings were produced with feeders on the top face of the
casting, and with bottom gating. During machining of internal face of the three
vertical walls (side and back walls with large feeders seen in the figure), shrinkage
defects of size up to 6-8 mm were observed 5-10 mm beneath the cast surface, about
150-250 mm below the top face (Fig. 5).
Fig. 5. Magnet frame tube casting revealed shrinkage porosity during machining
Solid modelling of the casting along with the current method layout followed by
casting simulation showed that a thin vertical section of casting just below the feeders
is choking the supply of feed metal to the casting wall, leading to shrinkage porosity.
Since there was a constraint to place the feeders on the top face only, it was decided to
explore product design modification to solve the feedability problem. Accordingly,
three different solutions were proposed:
1. Adding cooling fins on the wall affected by shrinkage porosity
2. Increasing thickness of the narrow section below the feeders (padding)
3. Adding one degree taper to the vertical wall as well as padding to the narrow
section.
All three modifications were incorporated in the product design on the three adjacent
walls. The solidification simulation was carried out to verify the three solutions (Fig.
6). All showed improved feeding characteristics, and hot spots are now seen in the
feeders instead of inside the casting.
Fig. 6. All three design modifications eliminated shrinkage porosity
Based on the above analysis, two product design changes: padding and fins were
implemented on two different castings. The trial casting with fins showed completely
porosity free machined surface (Fig. 7), as predicted by simulation, and was finally
implemented since it was also easier to machine the fins.
Fig. 7. Trial castings with fins revealed defect free machined faces
The second case study involved an extrusion press cylinder casting of steel of about
3.1 m height and weight 23 tons. After a few years of continuous use, the casting
tended to crack or leak near a square projection on the side surface (Fig. 8). Initial
investigation included analysis of the cylinder for stresses developed in working
condition under cyclic loads assuming the failure was caused by fatigue. It was
however, observed that the projection (location of leakage) was subjected to low
stresses, ruling out this hypothesis. The foundry on their part experimented with
different methoding layouts, including use of local chill, feeder at the projection, and
providing padding (taper on inside part of cylinder). The problem however, persisted.
Fig. 8. The defect-prone square projection is moved up in the modified design
The foundry and the OEM collaborated to analyse the problem supported by casting
simulation. Their inputs from the foundry included the current method and all major
process parameters. The design engineers from OEM provided a 3D model of the
press cylinder and possible directions for product design modifications. The part
model, methoding and process plan data was used for simulating the casting. The
results of analysis showed that the section connecting the feeder with the junction
solidifies much before the projection, chocking the flow of feed metal and thereby
leading to potential micro porosity under the projection. It was concluded that after a
few years of service, these micro porosities get connected leading to leakage in few
cases. Different solutions were modeled and simulated to alleviate the problem. This
included use of padding (internal taper on the thin section to increase its thickness),
but was discarded as it involved increased machining costs. After several virtual
tryouts and analysis it was concluded that the micro porosity can not be completely
eliminated by methoding or process parameters alone. Therefore, in consultation with
the product designer, it was decided to shift the square projection to a new location
closer to the feeder (Fig. 8). This did not affect the product functionality, and solved
the shrinkage porosity problem since the top feeder was now able to feed the junction
area (Fig.9). The casting was successfully produced, and is performing satisfactorily.
Fig. 9. Modified press-cylinder design shows connected feed paths and hot spots
inside feeder; actual casting is defect-free.
6. CONCLUSION
The bottlenecks and non-value added time in casting development can be minimised
by adopting computer-aided methoding, solid modelling, and casting simulation
technologies. An integrated system combining these functions has been developed,
and successfully demonstrated on industrial castings. Several innovative algorithms,
including geometric reasoning, domain knowledge modelling, automatic solid
modelling, and Vector Element Method helped in reducing the iteration time for
methoding modification and simulation to less than one hour for typical castings. It
has also proven to be very useful for verifying the manufacturability of a casting and
improving it by minor modifications to part geometry, before freezing the design in
early stages of product lifecycle. We hope that these developments will lower the
entry barrier for CAD and simulation technologies to even SME foundries in remote
areas, and lead to increased cooperation between foundry and OEM engineers.
REFERENCES
1. B. Ravi, Durgesh Joshi, Rahul Chougule, “Survery of Computer Applications in
Indian Foundry Industry: Benefits and Concerns,” 53rd Indian Foundry Congress,
Institute of Indian Foundrymen, Kolkata, Jan 2005.
2. B. Ravi, Metal Casting: Computer-Aided Design and Analysis, Prentice-Hall
India, New Delhi, 2005.
3. B. Ravi, R.C. Creese and D. Ramesh, “Design for Casting – A New Paradigm to
Prevent Potential Problems,” Transactions of the AFS, 107, 1999.
4. AutoCAST information and case studies, Advanced Reasoning Technologies,
http://www.adva-reason.com.
5. B. Ravi and M.N. Srinivasan, “Casting Solidification Analysis by Modulus Vector
Method,” International Cast Metals Journal, 9(1), 1-7, 1996.