US20090254316A1 - Sector meshing and neighbor searching for object interaction simulation - Google Patents
Sector meshing and neighbor searching for object interaction simulation Download PDFInfo
- Publication number
- US20090254316A1 US20090254316A1 US11/336,885 US33688506A US2009254316A1 US 20090254316 A1 US20090254316 A1 US 20090254316A1 US 33688506 A US33688506 A US 33688506A US 2009254316 A1 US2009254316 A1 US 2009254316A1
- Authority
- US
- United States
- Prior art keywords
- particles
- sector
- computer
- sectors
- select
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/10—Numerical modelling
Definitions
- the present invention relates to computer simulations of physical phenomena, and more particularly to methods for reducing the amount of computations necessary during a particle-system based computer simulation of the engagement of two or more objects.
- Computer simulations that model the interaction or engagement of two or more objects typically execute numerous computations at each time step through the engagement.
- the computations may be based on equations for the laws of conservation of mass, energy and momentum.
- One type of modeling technique uses data for particles that represent each of the objects involved in the engagement.
- the particles representing the objects interact with each other according to the laws of conservation of mass, energy and momentum.
- the system of particles also referred to as a computational mesh, is updated at each of a plurality of time steps through the engagement of the objects.
- the reality is that at any given time step, not all of the particles in the system may have an active role in the engagement.
- each particle is affected by its neighboring particles, not all particles have a significant influence on other particles in the system.
- the run time of the computer simulation can be greatly reduced with little or no expense to the fidelity of the results. Further, if each particle's significant neighboring particles can be efficiently determined at each time step, the run time of the simulation can be reduced.
- Data describing particles that represent each of the objects is generated from geometric data for objects.
- the data for each particle describes a mass density, velocity and energy at a position of the corresponding object.
- the particles are grouped into sectors to define a computational mesh comprising a plurality of sectors, wherein each sector is a volume region at a position in space in which particles associated with the objects may reside.
- so called neighboring particles are determined that are within a region of influence with respect to a select particle.
- Computations are performed based on laws of conservation of mass, energy and momentum to produce updated values for mass, velocity, energy and position for the particles at each of a plurality of time steps.
- a search is made through a limited or bounded volume region with respect to the select particle that consists of the region of influence for the select particle at the previous time step and within those sectors in contact with or bordering the region of influence at the previous time step.
- the plurality of select particles are identified as those particles that reside in an active sector, wherein an active sector is a sector that contains, or is adjacent to a sector that contains, particles that is actively involved in the engagement between the two objects.
- an active sector is a sector that contains, or is adjacent to a sector that contains, at least one particle that has a velocity or stress greater than a corresponding predetermined amount.
- the particles are regrouped and neighboring particles are identified at each time step.
- the number of computations performed at a given time step is greatly reduced.
- the computer simulation run time is significantly reduced by focusing computational resources on the particles that are most relevant to the engagement of the objects at a particular time step, therefore reducing the number of particles in the computational mesh at a given time step.
- FIG. 1 is a diagram depicting interaction of two objects, where data representing the objects are processed for computer simulation of the object interaction according to an embodiment of the present invention.
- FIG. 2 is a flow chart generally depicting a process for generating data describing collections of particles representing the objects for computer simulation of the object interaction according to an embodiment of the present invention.
- FIG. 3 is a flow chart depicting a process for generating a computational mesh comprising particles grouped in sectors and in which neighboring particles are determined for each particle in the computational mesh, according to an embodiment of the present invention.
- FIG. 4 is a screen shot showing how particles representing two objects are grouped into sectors in the process of FIG. 3 according to an embodiment of the present invention.
- FIG. 5 is a diagram showing how neighbor particles are identified for a particle in the process of FIG. 3 according to according to an embodiment of the present invention.
- FIG. 6 is a screen shot of an initial material plot of an exemplary engagement of two objects according to an embodiment of the present invention.
- FIGS. 7-12 are screen shots of the computation sector mesh at time steps of for the exemplary engagement of the two objects shown in FIG. 6 according to an embodiment of the present invention.
- a first object 10 is to collide with a second object 20 .
- the objects 10 and 20 could be any two objects that may collide with each other, or one of which may explode or detonate near or on the other, etc. Either or both objects may be moving, or one object may explode in or near the other object.
- Non-limiting examples of the experiment include: object 20 is stationary (e.g., a building structure) and object 10 is moving and collides or explodes near object 20 , where object 10 is a moving vehicle such as a land vehicle, air vehicle (airplane, missile, etc.); object 20 is moving and object 10 is moving and the two objects collide with each other, one of which may or may not set off an explosion upon or near impact, where object 20 is an air vehicle and object 10 is an air vehicle; objects 10 and 20 are both stationary and one explodes inside or near the other object. It should be understood that while only two objects are shown in FIG. 1 , the experiment may involve more than two objects.
- simulation algorithms have been, and are being, developed to predict the possible outcomes of such events using computations that represent the various physical phenomena occurring.
- Some simulation algorithms use specialized data describing the objects in the experiment.
- One type of specialized data is data for a collection of “particles” that represent each object, where each particle is defined generally by mass, density, velocity, energy, pressure and stress (so called “particle parameters”) at a position of the object at a given point in time during the experiment.
- Computations are then made on each of the particles representing each object using equations describing the physical phenomena occurring during the experiment to ultimately determine possible outcomes.
- These computations are very complex and can therefore take a significant amount of time to complete throughout time steps of an experiment. Therefore, according to the present invention, techniques are provided to reduce the number of the computations made at each time step for the particles in the system, also referred to as the computational mesh, comprised of the collection of particles representing each object in the simulation.
- FIG. 2 a process 100 is generally shown for such an improved computer simulation.
- Data describing the two or more objects involved in the computer simulation are received as input.
- a discretization algorithm is then executed at 110 to convert the object data for each object to a collection of particles. This may involve separately each distinct component of each object.
- FIG. 1 shows, as an example, that object 10 may consist of distinct components or portions A, B, C, D and E.
- a collection of particles is generated at 110 for each of these components for object 10 .
- a collection of particles is also generated for object 20 .
- a so-called computational mesh is generated from the collection of particles and the neighboring particles are located and identified for each particle in the computational sectorized mesh.
- the purpose of this step is also to determine which particles are “active” in the interaction among the objects and to limit the computations to only those active particles and a limited number of neighboring particles for the active particles.
- the computations are performed at 140 for the active particles based on the identified neighbors according to the governing equations associated with the physical phenomena being modeled by the computer simulation.
- the results of the computations are examined at 150 to determine if a desired outcome has occurred (or can even possibly occur). For example, a desired outcome may be destruction of a portion of all of one or all of the objects involved in the interaction.
- a desired outcome may be destruction of a portion of all of one or all of the objects involved in the interaction.
- the energy and other parameters associated with the particles in the system are examined (after the computations made at 140 ) to determine whether it meets certain criteria indicative of the desired outcome.
- the object discretization involves generating data describing particles that represent each of the objects (at least first and second objects) from geometric data for the objects, wherein the data for each particle describes a mass, density, velocity, energy, pressure and stress at a position of the corresponding object, to produce a first collection of particles representing the first object and a second collection of particles representing the second object.
- the time step is incremented at 160 and the process 100 is repeated from 120 .
- the processes referred to in the flow chart of FIG. 2 may be performed by one or more computers 200 ( 1 ) to 200 (N). Moreover, the techniques described herein may be embodied by instructions stored on a computer readable medium that, when executed by a computer, cause the computer to perform the various steps described herein.
- SPH smooth particle hydrodynamics
- r j is a new independent variable and W is an appropriate weighting function usually chosen to have the following properties:
- h is a so-called smoothing length that defines the region of compact support known as the Kernel and ⁇ is the Delta function.
- the first and second properties ensure compact support while the third property ensures convergence. Although many different functions satisfy the above three properties, one such function is the 3 rd order B-spline function.
- Equation (1) may be converted to a summation if the function, f(r), is only known at discrete points of corresponding volume (m j / ⁇ j ) where m j and ⁇ j are the mass and density of interpolation point, j, respectively. Details of the derivation are known in the literature and are not repeated here. The resulting summation is:
- equations (5) and (6) form the basis of the SPH method and allow the partial differential equations of the Newtonian conservation laws to be transformed into discrete summations of neighboring interpolation points.
- the artificial viscosity is included in both the energy and momentum conservation equations.
- the artificial viscosity contains a bulk viscosity to suppress post-shock velocity oscillations and a Neumann-Richtmeyer viscosity to dissipate shock energy. Weighting functions of the two forms, ⁇ and ⁇ , are taken to be unity.
- Each particle's sound speed is updated at every time step via the following relationship:
- the stress tensor, strain rate tensor, rotation rate tensor, constitutive relationship and equation of state employed are listed in Table 2 along with accompanying particle relationships.
- the time rate of change of the stress deviator tensor is provided via the well known Jaumann rate equation. All materials are assumed to behave elastic-plastic with material failure checked against the von Mises yield criteria. For plastic deformations, an appropriate flow rule is used to relax stresses to the yield surface.
- the Mie-Gruneisen equation of state relates pressure to material density and internal energy.
- the main computational expense of the SPH computer simulation technique is searching for neighbor particles for each particle in the computational mesh. Location and identification of each particle's neighbor particles is necessary when executing the computations for the conservation equations in order to precisely model the interaction of the objects. For example, in a typical calculation, the neighbor search process can account for 50 to 60 percent of total computational time. Obviously, increasing the efficiency of this aspect of the computer simulation process can significantly boost run time statistics.
- the geometric extents of the system of particles representing the objects is determined at 122 .
- this is simply determining the geometric extents of the two objects at the point of closest approach, just prior the interaction of the objects.
- the geometrical extents may be defined by the movements of the outermost particles.
- the particles are binned or grouped into sectors based on their spatial positions.
- the sectors are volume regions that may have any shape.
- the sectors have the shape of parallelepipeds, e.g., rectangular volume regions and more specifically cube-shaped regions.
- the x-, y- and z- dimensions of the sectors are a function of the overall geometric extents determined at 122 .
- the dimensions of the sectors are the same size for a given time step.
- the dimensions of the sectors can be different across time steps, but generally are proportional to the geometric extents so that the larger the geometric extents, the bigger the dimensions are for the sectors.
- sectors that have no particles in them are ignored, whereas sectors having one or more particles having a non-trivial velocity or stress (pressure), e.g., a velocity and/or stress value greater than a corresponding predetermined amount or value, are considered “active sectors”.
- Active sectors may be further classified as primary or secondary.
- the criteria for calling a sector “active”, referred to herein as the “activation criteria”, is a sector that has at least one particle with a non-trivial amount of velocity, stress or pressure. Such a “non-trivial” amount depends on the particular physical experiment being modeled but is generally an amount greater than a certain threshold or amount.
- a primary active sector has non-trivial velocity or stress (pressure).
- Secondary active sector are those sectors “touched” by a primary active sector.
- a primary sector can spawn secondary active sectors, but a secondary active sector cannot spawn further secondary active sectors.
- the active sectors (primary and secondary) are identified as well in 124 and stored in a linked list.
- a screen shot showing a three-dimensional view of a sectorized computational mesh is shown in FIG. 4 , where one object, an aluminum sphere, impacts another object, an aluminum plate.
- the neighbor searching process in the next step 126 is limited to finding neighboring particles only for particles that are in the active sectors determined in 124 . However, the neighboring particles themselves may be in an inactive sector. Particles residing in an active sector at a given time step are referred to hereinafter as select particles.
- Each particle in the computational mesh has a smoothing length, h, which defines a region (e.g., having a sphere shape) of influence around the particle. This sphere of influence for a particle is called the Kernel.
- the size of a particle's Kernel is adjusted by varying the smoothing length, h, in order to maintain a constant number of neighbor particles in the Kernel for a given particle.
- the search region for neighboring particles is limited to sectors in the Kernel from the previous time step and those sectors in contact with or bordering the particle's Kernel from the previous time step.
- a bounding volume region that limits this search region is determined as shown at reference numeral 200 in the two-dimensional representation in FIG. 5 .
- the bordering sectors are those sectors 300 ( 1 ) to 300 (N) within the bounding volume region 200 .
- a sort routine identifies X number of neighbor particles, such as 55, that are closest to the particle in question and the smoothing length is chosen such that:
- the search and identification of neighbor particles j for particle i is performed at 128 in the bounding volume region. This aspect of the process greatly reduces the run time of the computer simulation.
- This process is repeated for each particle in the active sector, as shown at 129 until the neighbors are determined for all particles in all active sectors.
- the data representing the neighbor particles for all particles in the active sectors are then returned to the computations at 140 ( FIG. 2 ).
- smoothing lengths for individual particles are not allowed to exceed a predetermined size, such as three times the average initial smoothing length of all particles in the computational domain.
- a particle j may be considered for purposes of a neighbor search for particle i, particle j may be in a sector that is not an active sector and thus a neighbor search is not performed for particle j.
- a new sector mesh is defined based on the extents of the active particles within the system and on an expanded zone around the geometry of a particular portion of one of the object, a so called target zone. Additionally, sectors that are not active (primary or secondary) sectors are excluded from kernelling (steps 126 and 128 in FIG. 3 ). This allows the sector mesh to concentrate in the area of need.
- FIG. 8 shows that there are some particles for object 400 that are not part of the computational mesh because they do not have sufficient velocity or pressure associated with them at that point in time in the simulation.
- the particles representing the end of the object 400 remote from the engagement with object 500 that are initially outside of the computational mesh in FIGS. 8 and 9 become part of the computational mesh as the simulation proceeds and those particles become involved in the engagement.
- the dimensions of the sector bins grow proportionally to the geometric limits of the computational mesh.
- the computation mesh becomes larger as the first object 400 makes further progress into the second object 500 and all of the particles from the first object 400 become active and more particles of the second object 500 become active, as shown in FIGS. 10 , 11 and 12 .
- the computer simulation run time is significantly reduced by employing a refined sector mesh that focuses computational resources in the highest need areas (those particles that are most relevant to the engagement of the objects at a particular time step) and therefore reducing the number of particles in the computational mesh at a given time step.
Abstract
Description
- The present invention relates to computer simulations of physical phenomena, and more particularly to methods for reducing the amount of computations necessary during a particle-system based computer simulation of the engagement of two or more objects.
- Computer simulations for experiments involving the impact of one object with another object have widespread applications. For example, automobile manufacturers use such simulations in designing safer vehicles. In a totally different technology field, scientist uses such simulations to study the effectiveness of a missile destroying a moving or stationary target. Regardless of the particular application, it is an overall goal to design a computer simulation that can accurately produce data concerning possible outcomes of the physical phenomena of interest pertaining to two or more objects. However, there is a tradeoff between accuracy and simulation run time. Generally, the more complex a simulation is in order to achieve better accuracy, the longer it takes for that simulation to run to completion. In fact, very complex computer simulations, such as so-called “hydrocodes” can take several days or longer to execute on highly sophisticated models of certain physical events.
- Computer simulations that model the interaction or engagement of two or more objects typically execute numerous computations at each time step through the engagement. For example, the computations may be based on equations for the laws of conservation of mass, energy and momentum. One type of modeling technique uses data for particles that represent each of the objects involved in the engagement. The particles representing the objects interact with each other according to the laws of conservation of mass, energy and momentum. The system of particles, also referred to as a computational mesh, is updated at each of a plurality of time steps through the engagement of the objects. At each of the time steps, it is necessary to consider the influence that particles from within the same object and from other objects have on each other. However, the reality is that at any given time step, not all of the particles in the system may have an active role in the engagement. Moreover, while each particle is affected by its neighboring particles, not all particles have a significant influence on other particles in the system.
- If, at each time step, the number of particles involved in the computations is limited to those particles determined to have an active role in the engagement, the run time of the computer simulation can be greatly reduced with little or no expense to the fidelity of the results. Further, if each particle's significant neighboring particles can be efficiently determined at each time step, the run time of the simulation can be reduced.
- Briefly, methods for computer-implemented simulation for the interaction of two or more objects are provided. Data describing particles that represent each of the objects is generated from geometric data for objects. The data for each particle describes a mass density, velocity and energy at a position of the corresponding object. The particles are grouped into sectors to define a computational mesh comprising a plurality of sectors, wherein each sector is a volume region at a position in space in which particles associated with the objects may reside. For each of a plurality of select particles, so called neighboring particles are determined that are within a region of influence with respect to a select particle. Computations are performed based on laws of conservation of mass, energy and momentum to produce updated values for mass, velocity, energy and position for the particles at each of a plurality of time steps. According to one aspect, when determining neighboring particles for a given select particle, a search is made through a limited or bounded volume region with respect to the select particle that consists of the region of influence for the select particle at the previous time step and within those sectors in contact with or bordering the region of influence at the previous time step.
- According to another aspect, the plurality of select particles are identified as those particles that reside in an active sector, wherein an active sector is a sector that contains, or is adjacent to a sector that contains, particles that is actively involved in the engagement between the two objects. For example, an active sector is a sector that contains, or is adjacent to a sector that contains, at least one particle that has a velocity or stress greater than a corresponding predetermined amount.
- The particles are regrouped and neighboring particles are identified at each time step. By limiting the volume region which a search is made for neighboring particles, and by performing the neighbor search step only for particles in sectors that are “actively” involved in the engagement, the number of computations performed at a given time step is greatly reduced.
- The computer simulation run time is significantly reduced by focusing computational resources on the particles that are most relevant to the engagement of the objects at a particular time step, therefore reducing the number of particles in the computational mesh at a given time step.
-
FIG. 1 is a diagram depicting interaction of two objects, where data representing the objects are processed for computer simulation of the object interaction according to an embodiment of the present invention. -
FIG. 2 is a flow chart generally depicting a process for generating data describing collections of particles representing the objects for computer simulation of the object interaction according to an embodiment of the present invention. -
FIG. 3 is a flow chart depicting a process for generating a computational mesh comprising particles grouped in sectors and in which neighboring particles are determined for each particle in the computational mesh, according to an embodiment of the present invention. -
FIG. 4 is a screen shot showing how particles representing two objects are grouped into sectors in the process ofFIG. 3 according to an embodiment of the present invention. -
FIG. 5 is a diagram showing how neighbor particles are identified for a particle in the process ofFIG. 3 according to according to an embodiment of the present invention. -
FIG. 6 is a screen shot of an initial material plot of an exemplary engagement of two objects according to an embodiment of the present invention. -
FIGS. 7-12 are screen shots of the computation sector mesh at time steps of for the exemplary engagement of the two objects shown inFIG. 6 according to an embodiment of the present invention. - Referring first to
FIG. 1 , an experiment is depicted in which afirst object 10 is to collide with asecond object 20. Theobjects object 20 is stationary (e.g., a building structure) andobject 10 is moving and collides or explodes nearobject 20, whereobject 10 is a moving vehicle such as a land vehicle, air vehicle (airplane, missile, etc.);object 20 is moving andobject 10 is moving and the two objects collide with each other, one of which may or may not set off an explosion upon or near impact, whereobject 20 is an air vehicle andobject 10 is an air vehicle;objects FIG. 1 , the experiment may involve more than two objects. - For these types of experiments, simulation algorithms have been, and are being, developed to predict the possible outcomes of such events using computations that represent the various physical phenomena occurring. Some simulation algorithms use specialized data describing the objects in the experiment. One type of specialized data is data for a collection of “particles” that represent each object, where each particle is defined generally by mass, density, velocity, energy, pressure and stress (so called “particle parameters”) at a position of the object at a given point in time during the experiment. Computations are then made on each of the particles representing each object using equations describing the physical phenomena occurring during the experiment to ultimately determine possible outcomes. These computations are very complex and can therefore take a significant amount of time to complete throughout time steps of an experiment. Therefore, according to the present invention, techniques are provided to reduce the number of the computations made at each time step for the particles in the system, also referred to as the computational mesh, comprised of the collection of particles representing each object in the simulation.
- With reference to
FIG. 2 , aprocess 100 is generally shown for such an improved computer simulation. Data describing the two or more objects involved in the computer simulation are received as input. A discretization algorithm is then executed at 110 to convert the object data for each object to a collection of particles. This may involve separately each distinct component of each object.FIG. 1 shows, as an example, thatobject 10 may consist of distinct components or portions A, B, C, D and E. A collection of particles is generated at 110 for each of these components forobject 10. A collection of particles is also generated forobject 20. - Next, at 120, a so-called computational mesh is generated from the collection of particles and the neighboring particles are located and identified for each particle in the computational sectorized mesh. The purpose of this step is also to determine which particles are “active” in the interaction among the objects and to limit the computations to only those active particles and a limited number of neighboring particles for the active particles.
- After the sector mesh and neighboring identification processing performed at 120, the computations are performed at 140 for the active particles based on the identified neighbors according to the governing equations associated with the physical phenomena being modeled by the computer simulation.
- The results of the computations are examined at 150 to determine if a desired outcome has occurred (or can even possibly occur). For example, a desired outcome may be destruction of a portion of all of one or all of the objects involved in the interaction. Thus, in 150, the energy and other parameters associated with the particles in the system are examined (after the computations made at 140) to determine whether it meets certain criteria indicative of the desired outcome. The object discretization involves generating data describing particles that represent each of the objects (at least first and second objects) from geometric data for the objects, wherein the data for each particle describes a mass, density, velocity, energy, pressure and stress at a position of the corresponding object, to produce a first collection of particles representing the first object and a second collection of particles representing the second object.
- If the desired outcome has not yet occurred (and there are not indications determined at 150 that it will not likely occur) then the time step is incremented at 160 and the
process 100 is repeated from 120. Thus, theobject discretization process 110 is performed once to establish the data for the collection of particles representing the objects in the computations for time t=0, just prior to actual interaction of the collection of particles representing the objects. - The processes referred to in the flow chart of
FIG. 2 may be performed by one or more computers 200(1) to 200(N). Moreover, the techniques described herein may be embodied by instructions stored on a computer readable medium that, when executed by a computer, cause the computer to perform the various steps described herein. - One computer simulation algorithm that uses the particles-based representation of the objects employs a so-called physics model using smooth particle hydrodynamics (SPH) to approximate variables over finite domains of compact support. SPH is a LaGrangian technique originally formulated to solve astrophysics problems, but has been expanded and enhanced to include material strength effects making the method attractive for hypervelocity impact problems.
- SPH does not rely on a traditional grid to quantify node relationships but rather uses interpolation theory to compute smooth field variables at discrete spatial locations throughout the computational domain. Using the theory, the function f at the spatial location, r, may be approximated as:
- where rj is a new independent variable and W is an appropriate weighting function usually chosen to have the following properties:
-
- and where h is a so-called smoothing length that defines the region of compact support known as the Kernel and δ is the Delta function. The first and second properties ensure compact support while the third property ensures convergence. Although many different functions satisfy the above three properties, one such function is the 3rd order B-spline function.
- Equation (1) may be converted to a summation if the function, f(r), is only known at discrete points of corresponding volume (mj/ρj) where mj and ρj are the mass and density of interpolation point, j, respectively. Details of the derivation are known in the literature and are not repeated here. The resulting summation is:
-
- Gradients of the function, f(r), may also be converted to summations. The resulting expression is given below.
-
- Together equations (5) and (6) form the basis of the SPH method and allow the partial differential equations of the Newtonian conservation laws to be transformed into discrete summations of neighboring interpolation points.
- These interpolation points are the particles referred to above. At every time step, the density, velocity, and energy of each particle are updated by solving appropriate conservation of mass, momentum, and energy equations. Particle positions are updated by integrating particle velocities over time steps selected to satisfy stability criteria. The conservation laws and the resulting particle relationships employed are provided in Table 1.
-
TABLE 1 Conservation Laws Property Governing Relationship Particle Relationship Mass Momentum Energy Position
where subscripts i and j refer to individual and neighbor particles, respectively, and vji=vj−vi. - The artificial viscosity is included in both the energy and momentum conservation equations. The artificial viscosity contains a bulk viscosity to suppress post-shock velocity oscillations and a Neumann-Richtmeyer viscosity to dissipate shock energy. Weighting functions of the two forms, α and β, are taken to be unity.
- Each particle's sound speed is updated at every time step via the following relationship:
-
- where the partial derivative of pressure with respect to density is evaluated with respect to constant entropy.
- The stress tensor, strain rate tensor, rotation rate tensor, constitutive relationship and equation of state employed are listed in Table 2 along with accompanying particle relationships. The time rate of change of the stress deviator tensor is provided via the well known Jaumann rate equation. All materials are assumed to behave elastic-plastic with material failure checked against the von Mises yield criteria. For plastic deformations, an appropriate flow rule is used to relax stresses to the yield surface. The Mie-Gruneisen equation of state relates pressure to material density and internal energy.
-
TABLE 2 Constitutive Relationships and Equation of State Property Governing Relationship Particle Relationship Stress σaβ = Saβ − δaβ P σi aβ = Si aβ − δaβ Pi Tensor Strain Rate Tensor Rotation Rate Tensor Constitutive Relationship Equation of State
where PH refers to the Hugoniot pressure and η is the compression ratio (1−ρ0/ρ). - Numerical stability is assured by satisfaction of the Courant condition. One technique selects appropriate time steps based on the minimum ratio of smoothing length to sound speed, smoothing length to particle velocity, and square root of the ratio of smoothing length to particle acceleration for all particles within the computational domain at every time step.
-
- Any number of schemes known in the art may be used for integrating the system of equations. However, one technique is a scheme that is accurate to the order of (Δt)2. To help boost run time statistics, the total number of particles in the computational mesh representing the two or more objects is minimized using sectorization and
neighbor generation process 120 described in more detail in conjunction withFIGS. 3-12 . - The main computational expense of the SPH computer simulation technique is searching for neighbor particles for each particle in the computational mesh. Location and identification of each particle's neighbor particles is necessary when executing the computations for the conservation equations in order to precisely model the interaction of the objects. For example, in a typical calculation, the neighbor search process can account for 50 to 60 percent of total computational time. Obviously, increasing the efficiency of this aspect of the computer simulation process can significantly boost run time statistics.
- With reference to
FIG. 3 , at the current time step (tn), the geometric extents of the system of particles representing the objects is determined at 122. At time t=0, this is simply determining the geometric extents of the two objects at the point of closest approach, just prior the interaction of the objects. At some time step beyond t=0, the geometrical extents may be defined by the movements of the outermost particles. - Next, at 124, the particles are binned or grouped into sectors based on their spatial positions. The sectors are volume regions that may have any shape. In one example, the sectors have the shape of parallelepipeds, e.g., rectangular volume regions and more specifically cube-shaped regions. The x-, y- and z- dimensions of the sectors are a function of the overall geometric extents determined at 122. The dimensions of the sectors are the same size for a given time step. The dimensions of the sectors can be different across time steps, but generally are proportional to the geometric extents so that the larger the geometric extents, the bigger the dimensions are for the sectors. In addition, at 124, sectors that have no particles in them, so called empty sectors, are ignored, whereas sectors having one or more particles having a non-trivial velocity or stress (pressure), e.g., a velocity and/or stress value greater than a corresponding predetermined amount or value, are considered “active sectors”. Active sectors may be further classified as primary or secondary. The criteria for calling a sector “active”, referred to herein as the “activation criteria”, is a sector that has at least one particle with a non-trivial amount of velocity, stress or pressure. Such a “non-trivial” amount depends on the particular physical experiment being modeled but is generally an amount greater than a certain threshold or amount. A primary active sector has non-trivial velocity or stress (pressure). Secondary active sector are those sectors “touched” by a primary active sector. A primary sector can spawn secondary active sectors, but a secondary active sector cannot spawn further secondary active sectors. The active sectors (primary and secondary) are identified as well in 124 and stored in a linked list. A screen shot showing a three-dimensional view of a sectorized computational mesh is shown in
FIG. 4 , where one object, an aluminum sphere, impacts another object, an aluminum plate. The neighbor searching process in thenext step 126 is limited to finding neighboring particles only for particles that are in the active sectors determined in 124. However, the neighboring particles themselves may be in an inactive sector. Particles residing in an active sector at a given time step are referred to hereinafter as select particles. - Each particle in the computational mesh has a smoothing length, h, which defines a region (e.g., having a sphere shape) of influence around the particle. This sphere of influence for a particle is called the Kernel. At 126, the size of a particle's Kernel is adjusted by varying the smoothing length, h, in order to maintain a constant number of neighbor particles in the Kernel for a given particle. To expedite the re-sizing process, the search region for neighboring particles is limited to sectors in the Kernel from the previous time step and those sectors in contact with or bordering the particle's Kernel from the previous time step. Thus, a bounding volume region that limits this search region is determined as shown at
reference numeral 200 in the two-dimensional representation inFIG. 5 . The bordering sectors are those sectors 300(1) to 300(N) within the boundingvolume region 200. A sort routine identifies X number of neighbor particles, such as 55, that are closest to the particle in question and the smoothing length is chosen such that: -
- Thus, once the search region limits are determined by the bounding
volume region 200, the search and identification of neighbor particles j for particle i is performed at 128 in the bounding volume region. This aspect of the process greatly reduces the run time of the computer simulation. - This process is repeated for each particle in the active sector, as shown at 129 until the neighbors are determined for all particles in all active sectors. The data representing the neighbor particles for all particles in the active sectors are then returned to the computations at 140 (
FIG. 2 ). - To avoid excessively large Kernels, smoothing lengths for individual particles are not allowed to exceed a predetermined size, such as three times the average initial smoothing length of all particles in the computational domain. Even though a particle j may be considered for purposes of a neighbor search for particle i, particle j may be in a sector that is not an active sector and thus a neighbor search is not performed for particle j.
- At each time step, a new sector mesh is defined based on the extents of the active particles within the system and on an expanded zone around the geometry of a particular portion of one of the object, a so called target zone. Additionally, sectors that are not active (primary or secondary) sectors are excluded from kernelling (
steps FIG. 3 ). This allows the sector mesh to concentrate in the area of need. -
FIGS. 7 through 12 show a time progression of sector mesh growth from t=0 to t=139.7 μsec for the engagement of twoobjects FIG. 6 .FIG. 7 shows the initial sectorization of the particles representobjects 400 and 500 (at time t=0), and indicates that just before engagement all of the particles for both objects are considered part of the computational mesh.FIG. 8 shows that there are some particles forobject 400 that are not part of the computational mesh because they do not have sufficient velocity or pressure associated with them at that point in time in the simulation. However, it is noteworthy that the particles representing the end of theobject 400 remote from the engagement withobject 500 that are initially outside of the computational mesh inFIGS. 8 and 9 become part of the computational mesh as the simulation proceeds and those particles become involved in the engagement. Also noteworthy is that the dimensions of the sector bins grow proportionally to the geometric limits of the computational mesh. The computation mesh becomes larger as thefirst object 400 makes further progress into thesecond object 500 and all of the particles from thefirst object 400 become active and more particles of thesecond object 500 become active, as shown inFIGS. 10 , 11 and 12. - In sum, the computer simulation run time is significantly reduced by employing a refined sector mesh that focuses computational resources in the highest need areas (those particles that are most relevant to the engagement of the objects at a particular time step) and therefore reducing the number of particles in the computational mesh at a given time step.
- The system and methods described herein may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The foregoing embodiments are therefore to be considered in all respects illustrative and not meant to be limiting.
Claims (31)
Priority Applications (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US11/336,885 US7610184B1 (en) | 2006-01-23 | 2006-01-23 | Sector meshing and neighbor searching for object interaction simulation |
JP2008551527A JP5280863B2 (en) | 2006-01-23 | 2007-01-18 | Piecewise meshing and neighborhood search for object interaction simulation |
EP07710192A EP1979725A4 (en) | 2006-01-23 | 2007-01-18 | Sector meshing and neighbor searching for object interaction simulation |
PCT/US2007/060685 WO2007087490A2 (en) | 2006-01-23 | 2007-01-18 | Sector meshing and neighbor searching for object interaction simulation |
AU2007208111A AU2007208111B2 (en) | 2006-01-23 | 2007-01-18 | Sector meshing and neighbor searching for object interaction simulation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US11/336,885 US7610184B1 (en) | 2006-01-23 | 2006-01-23 | Sector meshing and neighbor searching for object interaction simulation |
Publications (2)
Publication Number | Publication Date |
---|---|
US20090254316A1 true US20090254316A1 (en) | 2009-10-08 |
US7610184B1 US7610184B1 (en) | 2009-10-27 |
Family
ID=38309900
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US11/336,885 Active 2028-02-03 US7610184B1 (en) | 2006-01-23 | 2006-01-23 | Sector meshing and neighbor searching for object interaction simulation |
Country Status (5)
Country | Link |
---|---|
US (1) | US7610184B1 (en) |
EP (1) | EP1979725A4 (en) |
JP (1) | JP5280863B2 (en) |
AU (1) | AU2007208111B2 (en) |
WO (1) | WO2007087490A2 (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120116737A1 (en) * | 2005-04-19 | 2012-05-10 | D.E. Shaw Research LLC | Zonal methods for computation of particle interactions |
US20120284002A1 (en) * | 2011-05-05 | 2012-11-08 | Siemens Corporation | Simplified Smoothed Particle Hydrodynamics |
US20140297624A1 (en) * | 2012-06-01 | 2014-10-02 | Sas Ip, Inc. | Systems and Methods for Context Based Search of Simulation Objects |
WO2015013524A1 (en) * | 2013-07-25 | 2015-01-29 | Sas Ip, Inc. | Systems and methods for context based search of simulation objects |
US9280838B2 (en) | 2011-09-20 | 2016-03-08 | Fujitsu Limited | Computer product, search apparatus, and search method |
US20160306773A1 (en) * | 2011-08-09 | 2016-10-20 | Yale University | Quantitative analysis and visualization of spatial points |
US11449202B1 (en) | 2012-06-01 | 2022-09-20 | Ansys, Inc. | User interface and method of data navigation in the user interface of engineering analysis applications |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8180605B1 (en) * | 2005-11-23 | 2012-05-15 | Livermore Software Technology Corporation | Methods and systems for creating a smooth contact-impact interface in finite element analysis |
US9256713B2 (en) * | 2007-08-30 | 2016-02-09 | Exelis Inc. | Library generation for detection and identification of shielded radioisotopes |
US8200464B2 (en) * | 2009-04-01 | 2012-06-12 | Livermore Software Technology Corporation | Explosion simulation in finite element analysis |
GB2489526A (en) | 2011-04-01 | 2012-10-03 | Schlumberger Holdings | Representing and calculating with sparse matrixes in simulating incompressible fluid flows. |
JP6176029B2 (en) * | 2013-09-27 | 2017-08-09 | 富士通株式会社 | Simulation device, simulation program, and simulation method |
JP6940752B2 (en) * | 2017-06-01 | 2021-09-29 | 富士通株式会社 | Probe molecule placement method and placement device, target molecule binding site search method, search device, and program |
KR20230090208A (en) | 2021-12-14 | 2023-06-21 | 고려대학교 산학협력단 | Energy-momentum conservation method and apparatus for real-time simulation of deformable objects |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6195625B1 (en) * | 1999-02-26 | 2001-02-27 | Engineering Dynamics Corporation | Method for simulating collisions |
US6420698B1 (en) * | 1997-04-24 | 2002-07-16 | Cyra Technologies, Inc. | Integrated system for quickly and accurately imaging and modeling three-dimensional objects |
US20020181751A1 (en) * | 2000-12-08 | 2002-12-05 | Hale Ira David | Method for aligning a lattice of points in response to features in a digital image |
US6678642B1 (en) * | 1998-10-08 | 2004-01-13 | Sandia Corporation | Method of and apparatus for modeling interactions |
US20060235659A1 (en) * | 2005-04-13 | 2006-10-19 | Alias Systems Corp. | Fixed time step dynamical solver for interacting particle systems |
US20070242269A1 (en) * | 2004-03-06 | 2007-10-18 | Michael Trainer | Methods and apparatus for determining characteristics of particles |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH03204758A (en) * | 1990-01-06 | 1991-09-06 | Fujitsu Ltd | Particle motion simulation processing system |
JPH0477853A (en) * | 1990-07-13 | 1992-03-11 | Fujitsu Ltd | Particle motion simulation system |
-
2006
- 2006-01-23 US US11/336,885 patent/US7610184B1/en active Active
-
2007
- 2007-01-18 JP JP2008551527A patent/JP5280863B2/en not_active Expired - Fee Related
- 2007-01-18 WO PCT/US2007/060685 patent/WO2007087490A2/en active Application Filing
- 2007-01-18 AU AU2007208111A patent/AU2007208111B2/en not_active Ceased
- 2007-01-18 EP EP07710192A patent/EP1979725A4/en not_active Withdrawn
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6420698B1 (en) * | 1997-04-24 | 2002-07-16 | Cyra Technologies, Inc. | Integrated system for quickly and accurately imaging and modeling three-dimensional objects |
US6678642B1 (en) * | 1998-10-08 | 2004-01-13 | Sandia Corporation | Method of and apparatus for modeling interactions |
US6195625B1 (en) * | 1999-02-26 | 2001-02-27 | Engineering Dynamics Corporation | Method for simulating collisions |
US20020181751A1 (en) * | 2000-12-08 | 2002-12-05 | Hale Ira David | Method for aligning a lattice of points in response to features in a digital image |
US20040006450A1 (en) * | 2000-12-08 | 2004-01-08 | Landmark Graphics Corporation | Method for aligning a lattice of points in response to features in a digital image |
US20070242269A1 (en) * | 2004-03-06 | 2007-10-18 | Michael Trainer | Methods and apparatus for determining characteristics of particles |
US20060235659A1 (en) * | 2005-04-13 | 2006-10-19 | Alias Systems Corp. | Fixed time step dynamical solver for interacting particle systems |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9612832B2 (en) * | 2004-06-30 | 2017-04-04 | D.E. Shaw Research LLC | Parallel processing system for computing particle interactions |
US10824422B2 (en) * | 2005-04-19 | 2020-11-03 | D.E. Shaw Research, Llc | Zonal methods for computation of particle interactions |
US20130091341A1 (en) * | 2005-04-19 | 2013-04-11 | D.E. Shaw Research LLC | Parallel computer architecture for computation of particle interactions |
US20120116737A1 (en) * | 2005-04-19 | 2012-05-10 | D.E. Shaw Research LLC | Zonal methods for computation of particle interactions |
US9747099B2 (en) * | 2005-04-19 | 2017-08-29 | D.E. Shaw Research LLC | Parallel computer architecture for computation of particle interactions |
US8831916B2 (en) * | 2011-05-05 | 2014-09-09 | Siemens Aktiengesellschaft | Simplified smoothed particle hydrodynamics |
US20120284002A1 (en) * | 2011-05-05 | 2012-11-08 | Siemens Corporation | Simplified Smoothed Particle Hydrodynamics |
US20160306773A1 (en) * | 2011-08-09 | 2016-10-20 | Yale University | Quantitative analysis and visualization of spatial points |
US10303738B2 (en) * | 2011-08-09 | 2019-05-28 | Yale University | Quantitative analysis and visualization of spatial points |
US9280838B2 (en) | 2011-09-20 | 2016-03-08 | Fujitsu Limited | Computer product, search apparatus, and search method |
US20140297624A1 (en) * | 2012-06-01 | 2014-10-02 | Sas Ip, Inc. | Systems and Methods for Context Based Search of Simulation Objects |
US10002164B2 (en) * | 2012-06-01 | 2018-06-19 | Ansys, Inc. | Systems and methods for context based search of simulation objects |
US11449202B1 (en) | 2012-06-01 | 2022-09-20 | Ansys, Inc. | User interface and method of data navigation in the user interface of engineering analysis applications |
WO2015013524A1 (en) * | 2013-07-25 | 2015-01-29 | Sas Ip, Inc. | Systems and methods for context based search of simulation objects |
Also Published As
Publication number | Publication date |
---|---|
US7610184B1 (en) | 2009-10-27 |
AU2007208111B2 (en) | 2010-08-12 |
WO2007087490A3 (en) | 2008-06-19 |
AU2007208111A1 (en) | 2007-08-02 |
EP1979725A2 (en) | 2008-10-15 |
EP1979725A4 (en) | 2012-01-04 |
WO2007087490A2 (en) | 2007-08-02 |
JP2009524161A (en) | 2009-06-25 |
JP5280863B2 (en) | 2013-09-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7610184B1 (en) | Sector meshing and neighbor searching for object interaction simulation | |
Lindsay et al. | A particle method and adaptive treecode for vortex sheet motion in three-dimensional flow | |
Sun et al. | An accurate SPH Volume Adaptive Scheme for modeling strongly-compressible multiphase flows. Part 1: Numerical scheme and validations with basic 1D and 2D benchmarks | |
US8666713B2 (en) | Method of simulating deformable object using geometrically motivated model | |
US10354099B2 (en) | Particle simulation device, particle simulation method, and particle simulation program | |
US20120284002A1 (en) | Simplified Smoothed Particle Hydrodynamics | |
US11281820B2 (en) | Systems and methods for growth-based design | |
US7620532B2 (en) | Object discretization to particles for computer simulation and analysis | |
Ghisu et al. | An improved cellular automata for wildfire spread | |
US20190294754A1 (en) | Systems and methods for growth-based design | |
Holmberg et al. | Global and clustered approaches for stress constrained topology optimization and deactivation of design variables | |
US7684972B2 (en) | Termination assessment of a computer simulation | |
Štoller et al. | Design and assessment of shape of protective structure by usage of CFD software environment Ansys Fluent | |
Wei et al. | Mass-conservative network model for brittle fracture | |
Feuillet et al. | Mesh adaptation for the embedded boundary method in CFD | |
Isik | Accuracy and Stability of Smoothed Particle Hydrodynamics in Fluid-Structure Interaction Problems | |
Montanari et al. | Isogeometric models for impact analysis with LS-DYNA | |
Zhang et al. | A dynamic mesh method for unstructured flow solvers | |
CN117953127A (en) | Impact breaking special effect simulation method using PhysX engine | |
Rubio et al. | Generalized elastic staggered grids on multi-GPU platforms | |
Arnout et al. | Optimal design of shells with FE based parameterization | |
Hidajat et al. | An NSE ordering for plausible cloth simulations | |
Katz et al. | Convergence Acceleration Techniques for Coupled Adaptive Cartesian-Strand Grid Solutions | |
Strofylas et al. | On the Use of Improved Radial Basis Functions Methods in Fluid-Structure Interaction Simulations | |
Nguyen et al. | A feature-based mesh adaptation for the unsteady high speed compressible flows in complex three-dimensional domains |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: ITT MANUFACTURING ENTERPRISES, INC., DELAWARE Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:TILLMAN, STEVEN T.;WITZIG, ANDREW J.;REEL/FRAME:021136/0610 Effective date: 20060117 |
|
STCF | Information on status: patent grant |
Free format text: PATENTED CASE |
|
AS | Assignment |
Owner name: EXELIS INC., VIRGINIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:ITT MANUFACTURING ENTERPRISES LLC (FORMERLY KNOWN AS ITT MANUFACTURING ENTERPRISES, INC.);REEL/FRAME:027604/0136 Effective date: 20111221 |
|
FPAY | Fee payment |
Year of fee payment: 4 |
|
AS | Assignment |
Owner name: HARRIS CORPORATION, FLORIDA Free format text: MERGER;ASSIGNOR:EXELIS INC.;REEL/FRAME:039362/0534 Effective date: 20151223 |
|
AS | Assignment |
Owner name: HARRIS INTERNATIONAL, INC., FLORIDA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:HARRIS CORPORATION;REEL/FRAME:040945/0267 Effective date: 20161213 Owner name: EAGLE TECHNOLOGY, LLC, DELAWARE Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:HARRIS INTERNATIONAL, INC.;REEL/FRAME:040981/0138 Effective date: 20161213 |
|
FPAY | Fee payment |
Year of fee payment: 8 |
|
AS | Assignment |
Owner name: HARRIS IT SERVICES CORPORATION, VIRGINIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:EAGLE TECHNOLOGY, LLC;REEL/FRAME:042415/0432 Effective date: 20170428 |
|
AS | Assignment |
Owner name: MACQUARIE CAPITAL FUNDING LLC, AS COLLATERAL AGENT Free format text: SECURITY INTEREST;ASSIGNOR:HARRIS IT SERVICES CORPORATION;REEL/FRAME:042419/0527 Effective date: 20170428 Owner name: HPS INVESTMENT PARTNERS, LLC, NEW YORK Free format text: SECOND LIEN PATENT SECURITY AGREEMENT;ASSIGNOR:HARRIS IT SERVICES CORPORATION;REEL/FRAME:042419/0795 Effective date: 20170428 |
|
AS | Assignment |
Owner name: PERATON INC., VIRGINIA Free format text: CHANGE OF NAME;ASSIGNOR:HARRIS IT SERVICES CORPORATION;REEL/FRAME:043482/0524 Effective date: 20170725 |
|
AS | Assignment |
Owner name: ALTER DOMUS (US) LLC, AS ADMINISTRATIVE AGENT, ILLINOIS Free format text: SECOND LIEN PATENT SECURITY AGREEMENT;ASSIGNOR:PERATON INC.;REEL/FRAME:055192/0315 Effective date: 20210201 |
|
AS | Assignment |
Owner name: JPMORGAN CHASE BANK, N.A., NEW YORK Free format text: FIRST LIEN SECURITY AGREEMENT;ASSIGNOR:PERATON INC.;REEL/FRAME:055194/0474 Effective date: 20210201 Owner name: PERATON INC. (F/K/A HARRIS IT SERVICES CORPORATION), VIRGINIA Free format text: RELEASE OF FIRST LIEN SECURITY INTEREST;ASSIGNOR:MACQUARIE CAPITAL FUNDING LLC;REEL/FRAME:055194/0021 Effective date: 20210201 Owner name: HARRIS IT SERVICES CORPORATION, VIRGINIA Free format text: RELEASE OF SECOND LIEN SECURITY INTEREST;ASSIGNOR:HPS INVESTMENT PARTNERS, LLC;REEL/FRAME:055194/0034 Effective date: 20210201 |
|
MAFP | Maintenance fee payment |
Free format text: PAYMENT OF MAINTENANCE FEE, 12TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1553); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY Year of fee payment: 12 |