Systems and methods for ultrasonic characterization of permafrost, frozen soil and saturated soil samples

A method for characterizing a permafrost or frozen soil sample having a plurality of physical parameters, said method comprising:

• (a) during a test of said permafrost or frozen soil sample
• (i) from an ultrasonic transmitter, transmitting an input signal through the permafrost or frozen soil sample;
• (ii) at an ultrasonic receiver positioned oppositely of said ultrasonic transmitter across the permafrost or frozen soil sample, receiving said input signal, and generating a sample response signal in response thereto;
• (b) digitally storing actual test result data that at least comprises response signal data representative of the sample response signal from the ultrasonic receiver during said test;
• (c) through execution by one or more processors of statements and instructions stored in one or more non-transitory computer readable media, perform the following steps:
• (i) access said actual test result data;
• (ii) run a predictive poromechanical model with a plurality of different candidate parameter value combinations as input to thereby derive a plurality of resultant predictive datasets each representing a predicted test result, and apply machine learning-based optimization to identify preferred-candidate parameter value combinations for which the resultant predictive datasets from the predictive poromechanical model better approximate the actual test result data than other candidate parameter value combinations;
• (iii) perform iterative repetitions of step (c)(ii), each time using the preferred-candidate parameter value combinations from a prior iteration of step (c)(ii) as the input to the predictive poromechanical model, until at least one optimal parameter value combination whose resultant predictive dataset optimally matches the actual test result data is found;
• (iv) select said at least one optimal parameter value combination as a resultant measurement dataset for quantifying the physical parameters of the permafrost or frozen soil sample.

A system for characterization of a saturated soil sample of which a plurality of physical parameters is to be measured, said system comprising:

• an ultrasonic transmitter operable to emit an input signal through the saturated soil sample;
• an ultrasonic receiver positioned or positionable oppositely of said ultrasonic transmitter to reside across the saturated soil sample therefrom, and operable to generate a sample response signal in response to receipt of said input signal from the ultrasonic transmitter during a test of said saturated soil sample;
• one or more non-transitory computer readable media having stored thereon:
• a data store for storing actual test result data that at least comprises response signal data representative of the sample response signal from the ultrasonic receiver during said test;
• a predictive poromechanical model useful for calculating predictions of at least some of the actual test result data based on different combinations of potential parameter values for said physical parameters of said saturated soil sample;
• statements and instructions executable by one or more computer processors to perform the following steps:
• (a) access said actual test result data;
• (b) run the predictive poromechanical model with a plurality of different candidate parameter value combinations as input to thereby derive a plurality of resultant predictive datasets each representing a predicted test result, and apply machine learning-based optimization to identify preferred-candidate parameter value combinations for which the resultant predictive datasets from the predictive poromechanical model better approximate the actual test result data than other candidate parameter value combinations;
• (c) perform iterative repetitions of step (b), each time using the preferred-candidate parameter value combinations from a prior iteration of step (b) as the input to the predictive poromechanical model, until at least one optimal candidate parameter value combination whose resultant predictive dataset optimally matches the actual test result data is found;
• and
• (d) select said at least one optimal candidate parameter value combination as a resultant measurement dataset for quantifying the physical parameters of the saturated soil sample.

A method for characterizing a saturated soil sample having a plurality of physical parameters, said method comprising:

• (a) during a test of said saturated soil sample
• (i) from an ultrasonic transmitter, transmitting an input signal through the saturated soil sample;
• (ii) at an ultrasonic receiver positioned oppositely of said ultrasonic transmitter across the saturated soil sample, receiving said input signal, and generating a sample response signal in response thereto;
• (b) digitally storing actual test result data that at least comprises response signal data representative of the sample response signal from the ultrasonic receiver during said test;
• (c) through execution by one or more processors of statements and instructions stored in one or more non-transitory computer readable media, perform the following steps:
• (i) access said actual test result data;
• (ii) run a predictive poromechanical model with a plurality of different candidate parameter value combinations as input to thereby derive a plurality of resultant predictive datasets each representing a predicted test result, and apply machine learning-based optimization to identify preferred-candidate parameter value combinations for which the resultant predictive datasets from the predictive poromechanical model better approximate the actual test result data than other candidate parameter value combinations;
• (iii) perform iterative repetitions of step (c)(ii), each time using the preferred-candidate parameter value combinations from a prior iteration of step (c)(ii) as the input to the predictive poromechanical model, until at least one optimal candidate parameter value combination whose resultant predictive dataset optimally matches the actual test result data is found;
• (iv) select said at least one optimal candidate parameter value combination as resultant measurement dataset for quantifying the physical parameters of the saturated soil sample.