Toda Construction Co., Ltd asked TEAC and IF Computer to develop a Real Time Expert System to monitor the civil enginering process of tunnel excavation using the Pressurized Slurry Shield Tunnelling method.

The system was build in order to increase the safety of drilling operations for very large diameter tunnels. One target site is the Tokyo Bay Tunnel project linking Kawasaki and Chiba. This site is 14 meters in diameter, one of the largest under construction today.

During shield-tunneling with the pressurized slurry shield tunneling method in large cities, which is heavily used in metropolitan areas, safety control is a high-priority concern. Under the conditions of high depth, high water pressure, wide shield diameter (more than 10m), only highly reliable monitoring and data management at the construction site can assure safety.

Usually very experienced operators continuously monitor sensor data to recognize trends that may require preventive action or corrective adjustments to the drilling parameters.

Toda Construction recognized the potential of automatic monitoring to relief operator stress, to increase reliability, and to document both operator expertise and drilling processes.

The expert system receives a large amount of sensor data, such as slurry pressure, pump speeds, material flows: about 200 analog and digital data points every 2 seconds. The system uses this raw information together with built-in expert knowledge to analyse the ongoing process, predict possible trouble, infer reasons of arising problems and suggests the most suitable action or countermeasure. The normal turnaround time for a full inference cycle is 10 about seconds.

Inference method employed is a combination of forward chaining and expectation driven verification, i.e. local mixed inference. As IF/Prolog is used for all program modules except for the sensor data acquisition, which employs the C langauge.

Business Partner
TEAC Corporation (Sensor Systems)
Hardware
Hewlett-Packard
Software
IF/Prolog, C, UNIX


Partial view of raw data and evaluation result