Curve Filter

Description Include average, median, mean, min, max, mode, variance, RMS, symmetric nearest neighbour.
Module(s) Quantitative Interpretation, Explorationist *
Requirements
Related
Works with

* This process is made available in the Explorationist module temporarily.

The Curve Filter takes input well curves, applies a filter and outputs the result.

  1. In the Control Panel, go to View > New Single Well View.
  2. In the Single Well window, open the Process tab.
  3. At the tab header, click the blue "+" icon.
  4. Click on Curve Filter.
  5. Type a name for the process and click OK.

Configure the curve filter

Any number of input curves can be specified. The number of output curves matches the number of input curves. Each curve is filtered independently - the values of one curve do not affect another.

  1. In the Curves section, click the blue "+" icon to add a curve.
  2. Click the curve selector and select the intended curve from the Well Curve Chooser window.
  3. Click OK.
  4. Missing / Invalid Values: Choose the action when input variables are unavailable, null or invalid prior to filtering. This rule is applied when calculating each sample:
    • Value is 0: Replace NaN values with 0
    • Use nearest: Nearest Non-NaN Sample Interpolation
    • Interpolate: Interpolate missing values using the class interpolation setting
    • Interpolate (Linear): Interpolate missing values using linear method

Sliding window filter

These algorithms analyse a window of values. They can operate in any domain (TWT, TVD, MD). Input curve values are used directly in the filter. Runtime domain conversion requires checkshot velocity.

  1. Methods:
    • Average
    • Average Harmonic
    • Median
    • Trimmed Mean
    • Min
    • Max
    • Mode
    • Variance
    • RMS
    • Symmetric Nearest Neighbour: This algorithm smoothes data while preserving edges / large changes.
  2. Length: The filter window specified in output domain units.
    • For an output sample location, the filter window will be centered at the output sample.
      • Start of filter window: output sample location minus Length
      • End of filter window: output sample location plus Length
      • For the first output sample location, the filter window will be:
        • Start of filter window: first output sample location
        • End of filter location: first output sample location + Length
      • For the last output sample location:
        • Start of filter window: last output sample location - Length
        • End of filter location: last output sample location
    • Default Length:
      • TWT output domain: 5 ms.
      • TVD and MDKB output domain: 5 m.
    • If output and input domain are different, the filter window start, centre and end locations are converted to input domain to determine the location of the input curve for filtering. The filtered value will then output to the output location.
  3. Weighting: Applicable for method Average, Average Harmonic, Median, Trimmed Mean, Variance, RMS and SNN
    • Equal: All values in the window are weighted equally.
    • Linear: Values are scaled linearly from zero (at window start) to a maximum (at centre sample), then down to zero (at window end).
    • Gaussian: Values are scaled according to a normal (gaussian) curve.
  4. Trim %: The percent of extreme samples to remove from the window before calculating the mean. Applicable for method Trimmed mean.
    • Example: Trim=25%
      • Sort the values in the window.
      • Average the middle 75% of the values, i.e.
      • Do not include the upper 12.5% and lower 12.5% of the sorted values.

Transform filter

Transforms are applied to the full log curve in one pass.

  1. Methods:
    • Frequency filters (bandpass, high-pass, low-pass).
    • Discrete Wavelet Transform (DWT).
  2. Frequency Filters:
    • Only valid in TWT domain.
    • For TVD or MDKB domain, the curve will be converted to TWT domain before filtering. Time depth pairs are used for conversion.
    • Options for frequency filters:
      • Low pass
      • High pass
      • Bandpass
    • Parameters are similar to Band-Pass Filter process.
  1. Discrete Wavelet Transform: The entire curves are filtered in their native domain, then converted to the output domain.
    • Curves are filtered to "blocks" according to a scaled Daubechies wavelet i.e. each block contains the features of the curve corresponding to the wavelet at that scale.
    • Enter a mute value to reduce or remove the features for the corresponding block.
    • See https://en.wikipedia.org/wiki/Discrete_wavelet_transform for general information on wavelet transforms.

Output settings

  • Default output curve name: Curve name-Curve Class-Filter values
  • Default output curve class: Input curve class.
  • For all filter type and methods except for frequency filter,
    • Extent: Output can be in MDKB, TWT or TVDSS. By default, the process uses the domain and sequence of the first curve in the Input Curves section.
    • Enter the start, stop and step size.
    • Reset: automatically set the start and stop value to match the input data.
  • For frequency filter type
    • Extent: Output in TWT only.
    • Enter the start, stop. By default uses the start and stop of the first curve in the input curve section. The increment is determined by input curve.
    • Reset: automatically set the start and stop value to match the input data.