topic Re: Matrix is not positive definite in Robot Structural Analysis Forum

Matrix is not positive definite

Anonymous — Thu, 14 Dec 2017 17:31:37 GMT

I have a frame with lateral and gravity loads. I wish to run a P-delta analysis to account for second-order effects. When I do so I get the error "Matrix is not positive definite". To troubleshoot the issue, I have added a lateral restraint at the top of each frame. This would negate 100% of the sway. Robot gives me the same error. To troubleshoot further, I due a simple static analysis with the lateral restraints and the members are only lightly stressed. I do not understand how to get rid of the "Matrix is not positive definite" notice for P-delta analysis.

I am using RSA 2018 with the latest patch. I've included the model with the lateral restraints at the top of each frame and the P-delta analysis turned on.

Re: Matrix is not positive definite

Artur.Kosakowski — Fri, 15 Dec 2017 09:11:31 GMT

Hi @Anonymous

You cannot intersect truss bars as they have no out of their plane stiffness. Try to replace them with bars with releases.

If you find your post answered press the Accept as Solution button please. This will help other users to find solutions much faster. Thank you.

Re: Matrix is not positive definite

Anonymous — Wed, 03 Jan 2018 21:47:36 GMT

Thank you Artur,

Sorry, but I sent you the incorrect model. The model I intended to send did not have tension only bars. It did have compatible nodes. As it turns out, it will not converge using a P-delta analysis unless the compatible nodes are deleted. I'm sure that I've used compatible nodes with a P-delta analysis in previous versions of Robot. Has this feature changed in the 2018 version?

Best regards,

Carl

Re: Matrix is not positive definite

Rafacascudo — Wed, 03 Jan 2018 23:14:53 GMT

then send the correct "incorrect" model!!

Re: Matrix is not positive definite

Artur.Kosakowski — Thu, 04 Jan 2018 09:33:35 GMT

Hi @Anonymous

I'd say the definition of compatible nodes is such that some part o the model becomes unstable under the level of the applied load. It can be the same story - if you released all rotations in the 'middle' node rather than only these in the bracing plane ...

As @Rafacascudo wrote - we need to see the 'right' model.

Re: Matrix is not positive definite

Anonymous — Thu, 04 Jan 2018 14:35:34 GMT

Good day,

Here is the model using compatible nodes. I've been trying different options to get Robot to use P-Delta including tension only and axial releases on the bars. All of which work in non P-delta analysis, but not using P-delta.

Best regards.

Re: Matrix is not positive definite

Artur.Kosakowski — Mon, 08 Jan 2018 14:36:10 GMT

Hi @Anonymous

If I managed to answer your question(s) press the Accept as Solution button please. This will help other users to find solution(s) much faster. Thank you.

Re: Matrix is not positive definite

Anonymous — Wed, 10 Jan 2018 20:11:44 GMT

As follow-up to this, we have successfully modelled this frame using compatible nodes and tension only members when using a P-delta analysis. We accomplished this by changing the method of solving the system of equations to "Multi-threaded solver" from "Automatic" under Tools -> Job Preferences -> Structure Analysis.