Solved

PM Condition based, due date calculation

  • 2 November 2021
  • 14 replies
  • 434 views
indmlk
Rank
Userlevel 2
Badge +3

PM Maintenance Plan Due Date calculation is NOT CLEAR.

PM Maintenance plan, next planned value is calculated based on , either

(a) The Last Measurement at Maintenance

(b) The Predicted Value at Maintenance

when performed Value Base is YES. The calculation for Planned value is clear using 2 fomulars, but it is not clear how the Due date is calculated.

It says, LENEAR REGRESSION is used to calculate the Due Dates of the maintenance plan based on the planned value.

we did some testing using some measurements created but linear graph we drawn using Ms.Excel is not matching with the maintenance plan due dates.

Does anyone of you know how the Due date is calculated using LENEAR REGRESSION.

If you know any equation, please help me.

 

icon

Best answer by Dilani 3 November 2021, 12:10

View original
This topic has been closed for comments

14 replies

Nipun Gunaratne
Rank
Userlevel 5
Badge +9

Hi,

Predicted value at maintenance will be derived either using Interpolation method or Running Average. 

 

Best Regards,

Nipun

indmlk
Rank
Userlevel 2
Badge +3

Hi Nipuna, Next planned value is calculated using some formulas. That is correct. My Question is how the Due date is calculated using planned values.

Dilani
Rank
Userlevel 3
Badge +4

It is calculated by plotting a graph using last five valid measurements and their registered dates using a linear regression formula. 
Graph axis represents time variation and measurement variance. Due date is forecasted using that plotted graph by feeding planned value. Graph varies when new measurements added for a performed value based true plan. Hope this helps.

T
Rank
Badge +1

Hi Dilani,

We have been trying to recalculate Due Dates using Linear regression but we still cannot get exact dates provided by IFS. I have used your hints from message above but my dates still do not match with IFS calculations.

The equipment used has Planned Value based = NO. Are you able to review attached file and let me know if we are missing something in the manual calculation? Are there other parameters included in the calculation apart from measurements and date variations?

 

Thank you,

Francis

1 Attachment
Dilani
Rank
Userlevel 3
Badge +4

Hi Francis,

It would take lot of investigation time for this. I am sorry I don’t have enough time to go through the calculation and get back. I think it would be great if you could create a question to GSD, and they would be able to assist you on this.

Thanks & Best Regards,

Dilani

T
Rank
Badge +1

Hi Dilani

Thank you, I am discussing with consulting team to get details on calculations.


Regards,

Francis

indmlk
Rank
Userlevel 2
Badge +3

It is calculated by plotting a graph using last five valid measurements and their registered dates using a linear regression formula. 
Graph axis represents time variation and measurement variance. Due date is forecasted using that plotted graph by feeding planned value. Graph varies when new measurements added for a performed value based true plan. Hope this helps.

Hi Dilani,

This seems doesn't work. Do you have the exact formula to calculate the Due Date for condition Based PM maintenance lines. 

Thank you.

/Indik

T
Rank
Badge +1

Hi Indik,

I struggled to understand the calculations, with help from an IFS employee,  I found that:

  1. The calculation doesn’t use Linear regression
  2. All measurements are included in the calculation, unless a Calculation Start date is provided on the Test point.

The formula to obtain Next PM Due Date = Last Measurement date + Duration for next plan (*)

(*) Duration for next plan = Interval from Condition criteria / Average running hours per day (**)

(**) Average running hours per day = (Last reading - First reading) / (Last measurement date - First measurement date)

Hope this helps.

Regards,

Francis

Dilani
Rank
Userlevel 3
Badge +4

Hi Fransis,

This statement is untrue. We are using different calculation methods like interpolation and linear regression depending on conditions, but we never use all the measurements to forecast the plan (there is a max limit of 5). Indika had specifically asked on how linear regression calculation works which I explained.

If the calculation you have mentioned is constructed based on our IFS Documentation (Section on Condition-based generation in the about page for Generate WO inside PM Planning) , then the running average is calculated using linear regression for that scenario.

@Indika 

The formula is the standard linear regression. Please have a look in the code and I suggest raising a question to GSD if needed further assistant on this.

Best Regards,

Dilani

indmlk
Rank
Userlevel 2
Badge +3

Hi Dilani,

Thank you for the reply. we observed this issue with migrated measurements. We have migrated app8 measurements in to app10 and then observed this issue even for the new measurements.

We are still investigating that and I will update if we could find the reason behind this behavior.

Thank you and best regards,

Indika

T
Rank
Badge +1

Hi @Dilani 

May be you are right but I have done couple of scenario using Linear regression with tips provided but the Due date was never near IFS calculation. I sent you some scenario for you to confirm that your statement could be verified, I understand you said it would take time to check.

Using the above formula, I was able to get the dates right so what to conclude? What to do if I am getting two different statements from IFS? I am running IFS Apps 10 Upd 11, may be what you are saying is in other IFS versions. 

Regards,

Francis

Dilani
Rank
Userlevel 3
Badge +4

Hi Fransis,

What I said untrue was two facts. Fact that stated as if calculation never considers Linear regression and the fact that stated calculation will consider all measurements after calculation start date.

There is a scenario that Linear Regression will come to play when Predicted Value at Maintenance flag is selected and when it does it will consider 5 latest measurements after calculation start date. 

But as you mentioned, may be there is some difference when comes to the app10 updates.

 

I now expanded the calculation given by you. It would look like below when expanded.

Next PM Due Date = Last Measurement date + Interval* (Last measurement date - First measurement date)/(Last reading - First reading)

So it kind of look like a scenario of interpolation method. Yet, sorry about confusing your calculation’s Running Average with IFS documentation’s Running Average. There are so many things to consider while looking at these type of question and I had no intention to say formula you provided is wrong. It’s just hard to give generic answer when there are many things to consider, that’s why I kept suggesting to raise question as then someone can spend time to go deeper and provide a better explanation.

 

Thanks & Best Regards,

Dilani

T
Rank
Badge +1

Hi Dilani

Thank you for the feedback and clarification. I agree with you on the fact that Predicted Value at Maintenance can use Linear regression considering only a given number of measurements. 
My point was on the Due Date, referring to this topic. The system calculates a Due Date for next PM as well as the Planned Value. I don’t think we should consider that the same calculation method is used for Planned Value and Due Date.
 

Regards 

Francis 

 

 


 

indmlk
Rank
Userlevel 2
Badge +3

Hi Fransis and Dilani,

We did some technical investigations and found out the reason for causing wrong Due date in the maintenance plan for the measurements which are migrated from app8 to app10.

in our customer environmnet, we found it works fine for the new objects created and newly entered measurements which are showing correct duedates in the maintenance plan for condition based PMs according to the linear regression.

After investigating we realize. there was a pacth ID will solve this issue, patch ID is 160263.

From this patch, Order by is changed from VALUE_SEQ to REG_DATE. this is available from upd12 onwards.

we are investigating this further and testing ongoing with this fix and I will update if we found some more about this issue or calculation.

Thank you very much for both of your support on sharing your thoughts on this issue.

Best Regards,

Indika

Terms & ConditionsCookie settings