p-matrix Data Visualizer Software

 

p-matrix is a penalty matrix based data visualization software for analysing in-process data. The bubble diagram and penalty matrix technique enables you to implement ISO 9001:2015's Risk Based Thinking in your organisation.

 

Features of p-matrix Data Visualizer Software - For Windows PC

Free Version

Paid Version

p-matrix Input Data Sheet

Responses

1

10

Factors

20

200

Observations

30,000

30,000

LB, HB, NB, Manual Response Types

Included

Included

Lower the Better, Higher the Better and
Nominal the Best

Discrete and Continuous Factors

Included

Included

Sampling rate of factors and responses

Equal

Equal, Lower, Higher

Analysis

Finding Desired & Undesired response

Manual

Automatic

thresholds (Major step in analysis )

Analysing Correlated responses

Not included

Included

Discover Main Effects of Factors on

Included

Included

Response(s)

Discover Effect of Interactions between

Not Included

Included

multiple factors over Response(s)

p-matrix Report Sheets

Summary Sheet

Not Included

Included

Penalty Matrix - Main Effects

Included

Included

sheet

Penalty Matrix - Interactions sheet

Not Included

Included

Main Effects sheet

Not Included

Included

Interactions sheet

Not Included

Included

No Effects sheet

Not Included

Included

Training & Technical Support

CD with Recorded Presentations

Not Included

Included

Online Technical support

Not Included

Included

Join a 3.5 hour online training

Essential

Essential

foundation course

Join our next one day advanced training

Recommended

Recommended

course

A one day onsite advanced training

Available upon request

Available upon request

course for your organisation

 

More about Factor-Response Correlations found using p-matrix Data Visualizer Software

  • Develops a scheme to retrieve your data and undertake an initial analysis
  • It analyses hundreds and thousands of penalty matrices among factors
  • Classifies factor settings as Optimal, Avoid or No Effect
  • p-matrix discovers correlations based on data collected
  • Correlation does not mean Causation. Domain knowledge is necessary to interpret results
  • Ranks correlations in order of importance 1 = good to know, 10 = exceptionally strong
  • Patterns are statistically significant for strengths above 3.5 - 4
  • At end of analysis, a process engineer would get a list of top 15-20 matrices he/she needs to look at
  • Its findings are pointers for discussion to make a decision
  • Corrective actions suggest changes to process settings within current specification in your data set
  • You do not need to implement all of them
  • Hidden causes may be found which require further investigation E.g. If furnace 2-6 and ladle 1-3 are optimal , you would like to further investigate pouring temperature and pouring time parameters.
  • If no correlation is found include data on additional factors and rerun analysis to discover new correlations
  • It cannot and does not recommend extrapolating the results outside current specifications of your data set

 

In-Process Data Input

  • The data on ‘process response’ values is normally recorded per batch, per heat or per component. The date and time of manufacture is also normally recorded. Examples of process response include
    • % rejection values per batch,
    • porosity score or rework time per component,
    • defects like inclusions, cracks, and shrinkage, porosity, gas holes, cold shuts, etc
    • material properties such as UTS, YS, %EL, machinability index etc
  • The data for every response value (e.g. ‘% rejections due to inclusions’ are recorded per batch) can either be associated with
    • a corresponding value for a factor or process parameter (equal sampling rate: e.g. a single value for ‘compactibility of sand’ is recorded per batch) OR
    • more than one values for a factor (higher sampling rate: e.g. multiple values say 5-6 values of ‘compactibility of sand’ are recorded per batch).
  • Combining date, time and location of measurement for a response value can generate a unique identifier that can be used to link equal or higher sampling rate factor values with the corresponding response value. This connection will help to establish a meaningful traceability in the data.
  • Factors / process parameters may be associated with a number of sub-processes e.g. mould/shell making process, core making process, melting process, casting process, heat treatment and machining process.
  • Factors can have continuous values (e.g. temperature and humidity values, % carbon values etc) or discrete values (e.g. operator name, furnace type, oven, shift etc.)
  • It is very common to have missing data values for one or more factors and process responses.
  • The p-matrix software has an ability to visualise in-process data with equal and higher sampling rates including missing data for up to 200 factors and 30,000 observations.
  • An observation can be for a heat or a batch or any unique identifier that holds traceability for the response and corresponding process parameter setting
  • A typical p-matrix project has
  • One to two ‘process responses' (defect / material property)
  • 20 to 30 ‘factors’ (continuous/discrete, with higher/equal/lower sampling rates) and
  • as low as 30-50 observations (Note: Observations should include both good data i.e. when desired response was achieved as well as bad data i.e. when desired response was NOT achieved collected over a period of time.)

 

Detailed p-matrix Report

p-matrix software generates a Report on its findings in Excel spreadsheet format displaying the following information.

  • p-matrix Report consists of 6 sheets of Excel spreadsheet.
  • The first 5 sheets namely - Summary, Penalty Matrix Main Effects, Penalty Matrix Interactions, Main Effects and Interactions sheets show optimal and avoid classification of factor settings for one or more process responses.
  • Optimal classification improves the chances of realizing desired response behaviour.
  • Avoid classification increases the risk of achieving unacceptable response values.
  • Report is colour coded for ease of readability. Blue is used for optimal settings and Red for avoid.
  • Summary Sheet - For each response, it displays ranked optimal and avoid Main Effect of factor settings and interactions with other factor settings.
  • Penalty matrix Main Effects Sheet - For each response, it displays corresponding factor-response penalty matrices. Use this sheet -
    • to cross validate the findings shown on the Main Effects sheet,
    • to visualize penalty value variation for a given response across settings of multiple process parameters
    • to obtain actionable information from your in-process data to develop a confirmation trial plan to verify results of analysis
  • Penalty matrix Interactions Sheet - For each response, it displays corresponding interactions between two factor settings. Use this sheet -
    • to cross validate the findings shown on the Interactions sheet,
    • to visualize penalty value variation for a given response across two process parameter settings.
    • to obtain actionable information from your in-process data to develop a confirmation trial plan to verify results of analysis
  • Main Effects Sheet - For each response, it displays
    • Main Effect of a factor setting on one or more process responses along with strength of Main Effect.
    • It also highlights number of interactions and highest interaction strength with other factors settings.
    • You may also sort this sheet to view effect of
      • multiple process parameters on each defect OR
      • Each process parameter across multiple defects
    • You may use Excel filters to hide factor settings with low strength values
  • Interactions Sheet - For each response, it displays detailed interactions between multiple factor settings along with strength of interaction.
    • You may also sort this sheet to view effect of
      • multiple process parameters on each defect OR
      • Each process parameter across multiple defects
    • You may use Excel filters to hide factor settings with low strength values
  • Ranked Findings - All Findings are ranked. Ranking done from 1 to 10 in order of importance.
    • Findings with 1 or 2 strength value may have useful interactions. Findings with strength value 3.5 and above are significant.
  • No Effects Sheet - This sheet highlights process parameter settings that have no effect on the response. These settings 'Do not change the risk level' as 'No evidence' of correlation is discovered.