Add math function from file #692 #701

ezrayst · 2020-11-27T07:16:42Z

Describe the PR
Math function is modified to read acceleration time history of an earthquake, thus better to have it from a file instead of simply a json entry.

Related Issues/PRs
Part 1 of #692

Explanation
This design utilizes the available class structures already available. It mirrors the velocity boundary condition, with input being

node_id    dir    velocity_value / acceleration_value

with the variation over time defined in the math_function. With that, the acceleration will be kept at 1.0 in node and math_function defines the ground motion. The advantage is the robustness as we now have acceleration boundary condition as input, and it can be modified easily to allow for dynamic inputs. This method is preferred, but there may be extra double at each constrained node.

math_function can be defined through a csv file as shown here: https://ezrayst.github.io/mpm-doc/#/user/preprocess/input?id=math-functions

Alternative
To have another class, timehistory.h to define acceleration and stress over time. Those can have public functions such as apply_acceleration_time() that can be called in our solver (mpm) and applied directly instead of having a static value in node. This will mean that we have separate json entry called dynamic_input such as follow:

"dynamic_input": {
    "type": "rigid" (or "flexible"),
    "rigid_acceleration_time": "acc_time_filename.csv",
    "flexible_stress_time": "stress_time_filename.csv"
}

bodhinandach

Looks okay to me, @ezrayst, one minor comment only. A small question: could you let me know which lines assign these math_functions to nodal variables acceleration/velocity/displacement? Or is it coming in the near future? Thanks btw!

include/io/io_mesh_ascii.tcc

codecov · 2020-11-27T07:34:05Z

Codecov Report

Merging #701 (322e0f5) into develop (c9630b3) will increase coverage by 12.82%.
The diff coverage is 41.18%.

@@             Coverage Diff              @@
##           develop     #701       +/-   ##
============================================
+ Coverage    83.91%   96.73%   +12.82%     
============================================
  Files          169      130       -39     
  Lines        34176    25849     -8327     
============================================
- Hits         28676    25004     -3672     
+ Misses        5500      845     -4655

Impacted Files	Coverage Δ
include/solvers/mpm_base.h	`50.00% <ø> (ø)`
tests/io/io_mesh_ascii_test.cc	`99.58% <ø> (ø)`
include/solvers/mpm_base.tcc	`75.64% <37.50%> (ø)`
include/io/io_mesh_ascii.tcc	`74.80% <100.00%> (ø)`
...e/cbgeo/project/external/spdlog/details/registry.h
/home/cbgeo/project/src/io/io.cc
/home/cbgeo/project/tests/test_main.cc
...cbgeo/project/external/spdlog/details/null_mutex.h
...home/cbgeo/project/tests/materials/norsand_test.cc
...ome/cbgeo/project/tests/solvers/mpm_scheme_test.cc
... and 293 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c9630b3...322e0f5. Read the comment docs.

ezrayst · 2020-11-27T07:36:04Z

It should be coming next, but it is briefly described here #692, @bodhinandach

jgiven100 · 2020-11-27T07:40:41Z

@ezrayst are acceleration time histories typically only 1d?

ezrayst · 2020-11-27T08:05:37Z

@jgiven100, in 2D we usually only use the horizontal acceleration (not vertical), so earthquake input will only be in one direction. In 3D, we have two lateral directions.

But maybe that is not your question (I might be reading too much into this). If we need to run horizontal and vertical acceleration time history, we will have math_function_0 for the x axis and math_function_1 for the y axis.

jgiven100 · 2020-11-27T08:10:51Z

we will have math_function_0 for the x axis and math_function_1 for the y axis

@ezrayst Right, I was thinking about the 3D case. Seperate math functions for the 2 directions understood 👍

kks32

Instead of writing a custom csv reader, it would be better to use https://github.com/ben-strasser/fast-cpp-csv-parser

Also, the input JSON file structure is needed. Please add more details to the PR describing the design and input file structure.

ezrayst · 2020-11-30T20:23:03Z

Hi @kks32, two things:

The input file and json input are available here: Add documentation for math function mpm-doc#53, so please take a look. The overall design is described in Earthquake MPM #692, thus this PR is just the first point (make it easier and quicker for review).
For the parser, do you want to change it for the whole of our code? This PR is consistent currently with the rest of the code. If you want to change the io to use ben-strasser, I could do that too. Would that be a separate PR?

kks32 · 2020-11-30T21:05:13Z

@ezrayst

It's good to have PRs which are self-contained, so adding input details to the PR description helps to understand how the file would look like and the JSON configuration too since mpm-doc is a separate repository. Additionally, RFC Earthquake MPM #692 is inconsistent with the PR:

A time_history.h class is derived from the abstract function_base.h. Usually this math function is used to yield a value between 0 and 1, which will be multiplied by the prescribed traction or velocity. However, here, the prescribed acceleration is 1 where the time_history.h defines the acceleration given the time. In addition, another alternative input file will be implemented to read a file containing time and value (in this case it is acceleration).

No time_history.h is defined in this PR, so adding input files and design here helps.

I'd suggest using the csv parser in io, you can just do it for the time-history files.

ezrayst · 2020-12-01T16:33:32Z

@kks32 I updated the initial comment, and I will change to CSV within today, hopefully. Thanks!

ezrayst · 2020-12-03T08:25:51Z

@kks32, please let me know if you would like to discuss the structure of the code, but currently I have changed to csv file input. I think I see you are concerned about the size of the file - but I think typical earthquake ground motion would be in the order of thousands of entries, not millions. Thanks anyways!

ezrayst · 2020-12-10T09:18:31Z

Hi @kks32, do you still have more comments?

kks32

create a folder inside external for csv and then add their license as well. Follow the convention.

kks32 · 2020-12-10T12:25:36Z

include/solvers/mpm_base.h

@@ -27,6 +27,9 @@
 #include "particle.h"
 #include "vector.h"

+// CSV-parser
+#include "csv.h"


Should be included in io class not mpm base

Should we move json and catch to their own respective folders?

kks32

Could you please add a test that shows you can read the files and get the correct x and fx values?

kks32 · 2020-12-10T12:27:00Z

include/solvers/mpm_base.tcc

@@ -762,16 +762,44 @@ bool mpm::MPMBase<Tdim>::initialise_math_functions(const Json& math_functions) {
      const std::string function_type =
          function_props["type"].template get<std::string>();

+      // Initiate another function_prop to be passed
+      auto function_props_update = function_props;


Why do we need this?

kks32 · 2020-12-10T12:29:43Z

include/solvers/mpm_base.tcc

+          io_->json_object("mesh")["io_type"].template get<std::string>();
+      auto reader = Factory<mpm::IOMesh<Tdim>>::instance()->create(io_type);
+
+      // Math function is specified in a file, replace function_props_update


Read files in io class not in mpm_base. Create a separate class in io or use the read_mesh_ascii to read these files. I'd prefer for now if you use read_mesh_ascii to read the input file.

That's what I have tried to use in the beginning, and you requested CSV. Can we agree on which one to use now before I make more changes?

I'm not asking you to revert back, this is an io operation and reading the CSV should be in the io class as an additional function and not in the mpm_base

I see, so still CSV but within read_mesh_ascii. But doesn't it refer to ascii not csv?

kks32 · 2020-12-10T12:30:53Z

include/solvers/mpm_base.tcc

      // Create a new function from JSON object
      auto function =
          Factory<mpm::FunctionBase, unsigned, const Json&>::instance()->create(
-              function_type, std::move(function_id), function_props);
+              function_type, std::move(function_id), function_props_update);


why is it called function_props_update?

kks32 · 2020-12-10T12:32:32Z

include/solvers/mpm_base.tcc

+      // Math function is specified in a file, replace function_props_update
+      if (function_props.find("file") != function_props.end()) {
+        // Make separate arrays
+        std::vector<double> xvalues;


Having separate vectors makes it decoupled between x and y values, use a vector of arrays instead.

kks32 · 2020-12-10T12:40:24Z

@ezrayst This PR is a generic PR for reading x and fx values from a CSV file and has nothing to do with acceleration time histories per se. Please update your PR details accordingly, you may see it will be used for acceleration time histories, but there is no need to discuss how it will be implemented at the node level because that's not done here. Provide details specific to this PR, which is missing (an input JSON file and an example CSV file would help). These are the core details and should be included rather than a link to mpm-doc PR.

ezrayst · 2020-12-16T07:39:00Z

Hi @kks32. I will update the PR accordingly then, to just simply update the math_function. I will make a separate PR regarding Earthquake one, but can we discuss this quickly sometime this week?

kks32 · 2020-12-16T12:25:03Z

include/solvers/mpm_base.tcc

+          io_->json_object("mesh")["io_type"].template get<std::string>();
+      auto reader = Factory<mpm::IOMesh<Tdim>>::instance()->create(io_type);
+
+      // Math function is specified in a file, replace function_props_update


I'm not asking you to revert back, this is an io operation and reading the CSV should be in the io class as an additional function and not in the mpm_base

kks32 · 2021-07-15T21:05:25Z

#715 supersedes this PR. Closing

🎨 add math function from file

13f5194

ezrayst added Priority: Medium Status: Review needed Type: Feature request labels Nov 27, 2020

ezrayst requested a review from kks32 November 27, 2020 07:16

ezrayst self-assigned this Nov 27, 2020

ezrayst requested review from bodhinandach and jgiven100 November 27, 2020 07:16

bodhinandach approved these changes Nov 27, 2020

View reviewed changes

include/io/io_mesh_ascii.tcc Outdated Show resolved Hide resolved

🔧 refactor declaration of variables

a413321

jgiven100 approved these changes Nov 27, 2020

View reviewed changes

kks32 requested changes Nov 30, 2020

View reviewed changes

Ezra Tjung added 2 commits December 3, 2020 00:22

🎨 refactor and change input data to csv

46853d0

🧑‍🔧 add csv.h

3d10bf6

kks32 requested changes Dec 10, 2020

View reviewed changes

🔧 move csv and add license

322e0f5

kks32 requested changes Dec 16, 2020

View reviewed changes

thiagordonho mentioned this pull request Jun 22, 2021

Apply ground motion input as nodal acceleration constraint #715

Open

kks32 closed this Jul 15, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add math function from file #692 #701

Add math function from file #692 #701

ezrayst commented Nov 27, 2020 •

edited

Loading

bodhinandach left a comment

codecov bot commented Nov 27, 2020 •

edited

Loading

ezrayst commented Nov 27, 2020

jgiven100 commented Nov 27, 2020

ezrayst commented Nov 27, 2020

jgiven100 commented Nov 27, 2020 •

edited

Loading

kks32 left a comment

ezrayst commented Nov 30, 2020

kks32 commented Nov 30, 2020

ezrayst commented Dec 1, 2020

ezrayst commented Dec 3, 2020

ezrayst commented Dec 10, 2020

kks32 left a comment

kks32 Dec 10, 2020

ezrayst Dec 16, 2020

ezrayst Dec 16, 2020

kks32 left a comment

kks32 Dec 10, 2020

kks32 Dec 10, 2020

ezrayst Dec 16, 2020

kks32 Dec 16, 2020

ezrayst Dec 17, 2020

kks32 Dec 10, 2020

kks32 Dec 10, 2020

ezrayst Dec 16, 2020

kks32 commented Dec 10, 2020

ezrayst commented Dec 16, 2020

kks32 Dec 16, 2020

kks32 commented Jul 15, 2021

Add math function from file #692 #701

Add math function from file #692 #701

Conversation

ezrayst commented Nov 27, 2020 • edited Loading

bodhinandach left a comment

Choose a reason for hiding this comment

codecov bot commented Nov 27, 2020 • edited Loading

Codecov Report

ezrayst commented Nov 27, 2020

jgiven100 commented Nov 27, 2020

ezrayst commented Nov 27, 2020

jgiven100 commented Nov 27, 2020 • edited Loading

kks32 left a comment

Choose a reason for hiding this comment

ezrayst commented Nov 30, 2020

kks32 commented Nov 30, 2020

ezrayst commented Dec 1, 2020

ezrayst commented Dec 3, 2020

ezrayst commented Dec 10, 2020

kks32 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kks32 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kks32 commented Dec 10, 2020

ezrayst commented Dec 16, 2020

Choose a reason for hiding this comment

kks32 commented Jul 15, 2021

ezrayst commented Nov 27, 2020 •

edited

Loading

codecov bot commented Nov 27, 2020 •

edited

Loading

jgiven100 commented Nov 27, 2020 •

edited

Loading