Cognitive Automated Test System

Four Hound Solutions, LLC. is commercializing a novel Cognitive Automated Test System and Tool Set (CATS2) using a proprietary analysis algorithm. The system would provide a self-healing capability to any automated test system allowing manufacturing and diagnostic testing to continue even with the loss of critical test system assets.

CATS2 will consist of advanced software tools that automates the setup, execution, data collection and results analysis for use on automated test systems. CATS2 is planned to be the most advanced commercially available test software package in the industry today. CATS2 combined with Synthetic Instrumentation provides test engineers and technicians a turnkey, automated hardware/software solution for test development.

Test Software: Automated test programs and systems are typically written as a series of functional end-to-end tests with measurements made at the output pins or operator observations in order to assure that the product under test is operating correctly and ready for issue. They can be written in a variety of languages not limited to: LabVIEW®, ATEasy®, C/C++, Basic, Pascal, FORTRAN, and ATLAS. The test system sequencing is handled in one of two ways. The first method is for the program to execute all functional end-to-end tests and then call a diagnostic routine if a failure is detected. The second method is for the program to execute the functional end-to-end tests until a failure is detected then call a diagnostic routine associated to the failing test. Essentially each test is followed by diagnostic tests to isolate the fault to the level required by the specification.

ATS fault isolation is a refinement of the fault detection and characterization process. The goal is to locate the system failure, determine a reconfiguration alternative and automatically implement to restore the system to operation. This may be aided by historical data and procedural activities to narrow the field of possible alternatives. A solution to this problem is to implement system diagnostics in the maintenance test environment using a model-based diagnostic approach. According to AI-ESTATE, it is necessary to uncouple the diagnostic logic from the test logic[1],[2],[3]. This will allow design changes to be implemented into the test process without software coding changes.

Normally, when automating a test process, the engineer would create a test step for every adjustment required to each instrument used in the test process just like someone manually setting up the instrumentation. This can become a redundant process if a unit under test requires multiple tests using a similar test procedure. The engineer is also required to monitor what instrumentation exists on the system and not in use at run time of the specific test being created. After a test process is created and deployed for use, it relies on the specific instruments that are called in the test process to be present on the system to run. Because of this, if a required instrument is out for calibration, being repaired, or removed from the system for any other reason, the system is inoperable for any test that requires the missing instrument.

Functional and Diagnostic Test Flowchart

Functional and Diagnostic Test Flowchart

System Description: Cognitive Automated Test System and Tool Set (CATS2), this system will dynamically load and provide a scalable ATS test software algorithm. Through the use of the CATS2 the engineer can insert common tests into an automated test process instead of individual steps. The advantage of this method is engineering efficiency or reduction in development time. There are other derivative advantages: common test style among multiple engineers; developers with limited experience with a particular test can begin developing with minimal training; and test system downtime is reduced and in some instances eliminated.

A CATS2 system will be capable of dynamically loading instrumentation specific measurements depending on hardware available. This will allow for an instrument to be removed from the test station and an already deployed process that uses this instrument to still run as long as another method to accomplish each test exists. The system would provide a self-healing capability to any automated test system allowing manufacturing and diagnostic testing to continue even with the loss of critical test system assets.

When it comes to maintaining the operational status of a test system, one size does not fit all. Every system and environment has unique requirements associated with its operating environment, user interfaces, data acquisition, control flow and intercommunications. The proposed software architecture of CATS2 will provide flexibility.

The CATS reasoning algorithm is planned to be structured as a library of functions that provide reasoning services to an application. This structure enables the CATS2 to fit into unique operating environments, as opposed to regulating how interfaces, functions, and controls are executed. The result will be a client-server design and open architecture interface to the overall system. This should allow for great flexibility in various implementation scenarios, where different systems have different operating environments, requirements, and characteristics that the diagnostics or health monitoring system must fit into.

CATS2 will provide many benefits through its dynamic analyses, service-oriented implementation, capability for development and maintenance. In summary:

  • The model must be deterministic, and based on first principles of design. The outcomes must be precisely determined through known relationships among equipment states and test results.
  • Provide effective and accurate use of available information. This will eliminate errors that manually generated ad hoc procedures often exhibit.
  • Provide alternatives to operators and maintainers even when some measurements or observations results cannot be made available.
  • Accept data in any order and amount and provide results that are as accurate as the information allows. Thus, it is event driven for asynchronously collected data.
  • Provides best next test information that satisfies order requirements provided by the model developer and will minimize the cost of locating the fault.
  • Provides best next test information that takes into account groupings of measurements related to system state to more efficiently minimize the cost of locating a fault.
  • Implemented as a server to an application.
  • Implemented on multiple operating systems and for many applications. It could be ported to other environments as necessary.

[1] IEEE 1232-2002: “IEEE Standard for Artificial Intelligence Exchange and Service Tie to All Test Environments (AI-ESTATE)”, Piscataway, NJ: IEEE Standards Press.

[2] Sheppard, J. W., “Inducing Diagnostics Inference Model from Case Data”, Research Perspectives and Case Studies in System Test and Diagnosis. Boston: Kluwer Academic Publishers, 1998. P69-102

[3] Sheppard, J. W. and Simpson W. R., “Prototyping a Diagnostic Interface” AUTOTESTCON 98, Proceedings, 1998, pp 276-283