JP2869245B2 - Control and operation support method for shield machine - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a control and operation support method applied to the direction control of a shield machine and the like.

[0002]

2. Description of the Related Art Jack operation for determining the direction of excavation of a shield excavator is usually performed based on operational knowledge cultivated by experienced operators at various construction sites. However, the construction environment (ground conditions such as soil and water pressure) and the characteristics of the shield machine used (differences in operability depending on the position of the center of gravity of the aircraft) differ from site to site. Otherwise, construction that satisfies construction accuracy and safety cannot be performed. For this reason, the operator memorizes the experience of the shield machine currently in operation as a case, and if it is not possible to respond only with the operation knowledge cultivated in the past, properly uses the newly stored case and the past operation knowledge, Or, it is considered that they are operated in combination.

[0003] In recent years, a number of control systems using fuzzy inference methods have been proposed to automate the operation of a shield machine ("Application of Fuzzy Theory to Shield Tunnel Machines" Measurement and Control, Vol. 27, NO. 11 etc.). These proposals have been evaluated as effective techniques for incorporating the above-mentioned expert operator's operational knowledge into a control system.

Further, as a method of reflecting the difference in the construction environment and the characteristics of the shield machine on each site, a learning function by auto-tuning (automatic adjustment) of a membership function is provided to a control system using a fuzzy inference method. Has already been proposed (Japanese Unexamined Patent Publication No. Hei 18-18609).
7).

[0005]

FIG. 6 shows an example of a direction control system of a shield machine based on jack pattern control using a fuzzy inference method. In the figure, reference numeral 11 denotes a state evaluation unit which inputs measurement data representing the control state of the shield machine 17 and evaluates the degree of conformity (grade) of an input parameter to a fuzzy set by a membership function shown in a block 12. The fuzzy rule indicated by the block 14 is a rule base in which the operation knowledge of a skilled operator is made into a rule. The fuzzy inference unit 13 applies the fuzzy rule to the input parameter to perform fuzzy inference, and performs a control operation amount (output amount) as an output parameter. One-sided pressing change, etc.)
Ask for. The development rule indicated by the block 16 is a rule base in which the knowledge of the skilled operator regarding the jack pattern development is ruled. The pattern development unit 15 performs the pattern development of the shield jack using the inference value according to the development rule. And outputs a jack on / off command.

[0006] Many of the conventionally proposed direction control systems for shield excavators perform jack pattern control based on inference results based on a rule base.
However, the general fuzzy inference has a problem that the membership functions of the fuzzy rules and the input / output parameters are fixed, so that it is impossible to cope with changes in the construction environment and the characteristics of the shield machine. To solve this problem, a learning method by auto-tuning the membership function has been proposed. In this method, for example, when the ground is hard and it is difficult to cut a curve, the tendency is determined from the control result, and a measure is taken by changing the membership function of the output parameter of the fuzzy rule. However, when the membership function is tuned, if the basis for performing the tuning is not clear, an incorrect inference result may be caused. In addition, when a control result that is not expected is obtained due to factors other than the control parameters that define the membership function, the change of the membership function cannot handle the problem.
For example, if the curve cannot be cut because the excavation speed is too fast, all of the above methods deal with tuning processing of the membership function in spite of the necessity of a control operation to reduce the excavation speed before controlling the jack pattern. Therefore, improvement of the control state cannot be expected.

An object of the present invention is to control a shield machine based on an inference result of a rule-based inference incorporating an operation knowledge of a skilled operator, and to cope with a situation in which control by a rule-based inference alone is not sufficient. It is an object of the present invention to provide a shield excavator control and operation support method capable of accurately determining the cause from past cases and performing corresponding control or operation support of the shield excavator.

[0008]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention provides a method for controlling a shield machine based on a rule-based inference result in which a control rule is registered in a shield construction. If the expected control result is not obtained by rule-based inference, a case base is constructed by editing and registering the control results and other excavator operating data that affect the control results, and registering them as cases. It is characterized by retrieving past cases with similar inference application conditions from the base, determining the cause, and determining the corresponding control or operation support policy of the shield machine.

[0009]

According to the present invention, the control of the shield machine is performed based on the rule-based inference result in which the control rules are registered. If the expected control result cannot be obtained by the rule-based inference, the control is performed in the past. Of inference application conditions, control results, and excavator operation data that affect control results in a simulated control system, search for cases with similar inference application conditions from a case base registered as cases, investigate the cause, and shield correspondingly The decision of the control of the excavator or the operation support is made, and learning based on the case is performed, so if the situation cannot be dealt with only by the rule-based reasoning, the control by the learning or the operation support should be used. Can be.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention applied to direction control of a shield machine will be described below.

FIG. 1 is a block diagram showing one embodiment of the present invention, and FIG. 2 is a block diagram showing another embodiment of the present invention.
Functionally equivalent parts of the two are denoted by the same reference numerals.
In FIGS. 1 and 2, reference numeral 1 denotes a state evaluation unit for rule-based reasoning, which is a part for performing a state evaluation from measurement data representing the control state of the shield machine 4 and inputting by a membership function in jack pattern control by fuzzy reasoning. Evaluate the grade of the parameter. Reference numeral 2 denotes a rule-based inference unit which performs inference processing by applying the control rules registered in the rule base 3 to graded input parameters. In jack pattern control by fuzzy inference, the fuzzy inference unit 13 shown in FIG. And pattern development unit 15
Is equivalent to The inference result here is given to the shield machine 4 as a control command. Reference numeral 5 denotes a data editing unit for case registration, which edits case data to be registered in the case base 6. Reference numeral 7 denotes a problem comprehension unit for case-based reasoning. Here, when an expected control result cannot be obtained by control based on rule-based reasoning, a problem to be subjected to case-based reasoning is grasped and presented as inference parameters. Reference numerals 8 and 8 'denote search / inference units for case-based reasoning, which search for similar cases from the case base 6 based on the presented inference parameters, investigate the cause of the problem, and determine measures. In the example of (1), the operation of the operator is supported based on the inference result of the search / inference unit 8, and in the example of FIG. 2, the inference result of the search / inference unit 8 'is given to the shield machine 4 as a direct control command.

The jack pattern control for controlling the excavation direction of the shield excavator is divided into determining an initial jack pattern at the start of one-ring excavation and correcting the pattern during excavation.

Japanese Patent Application No. 3-659 filed earlier by the present applicant
As described in the specification of Japanese Patent No. 49, the initial jack pattern is determined, for example, from the target azimuth angle θh, the target pitching angle θv and the measured current shield azimuth angle θh ′ and pitching angle θv ′ in one ring excavation. The calculated target control angle in each of the horizontal and vertical directions Δθh = θh−θh ′,
Δθv = θv−θv ′, and what percentage of the target half angle of the shield required for calculation in curve construction is
Using the horizontal and vertical angle halfway achievement ratios ηh and ηv indicating whether or not the center is broken as input parameters, the horizontal and vertical necessary to achieve the target control angle using a fuzzy rule and a membership function determined in advance. Degree of one-sided pushing in each vertical direction (the degree of one-sided pushing = the sum of the distance of the used jack from the shield center line / the number of used jacks, where the longest distance from the shield center line of the jack is 1) Eh, Ev The used jack pattern that best satisfies the required degree of single pressing Eh, Ev is obtained by previously calculating the degree of single pressing in each of the horizontal and vertical directions for all jack patterns and searching the registered pattern database.

In this case, a fuzzy rule for obtaining the required degree of partial pressing is expressed in the following format.

(Horizontal direction) The target control angle θh is a,
And if the half-fold achievement rate ηh is b, the degree of single-sided pressing Eh is x.

(Vertical direction) If the target control angle .theta.v is a 'and the center folding achievement rate .eta.v is b', the degree of one-side pushing Ev is x '.

Here, a, a ', b, b' represent the degree of the input parameter, and x, x 'represent the degree of the output parameter.

FIGS. 4A and 4B are explanatory diagrams of control targets in the horizontal direction control. FIG. 4A shows the state of the shield machine 4 at the start of the excavation, and FIG. 4B shows the shield at the end of the excavation for one ring of the segment 9. This shows the state that the excavator 4 should be. In other words, the azimuth angle θh
Pointing the shield machine from 'to θh is the horizontal control target for this ring.

The pattern correction in the course of excavation is performed by controlling the excavation rate α, 1 ring, which indicates the percentage of excavation per one ring excavation stroke, based on control state measurement data taken in at specified intervals (every check stroke). Control amounts Δθhi and Δθvi indicating how many times the shield machine has been controlled in each of the horizontal and vertical directions since the start of the excavation. If the current jack pattern is maintained until the end of one ring excavation,
Calculate the predicted control amounts θhi and θvi indicating how many control amounts can be obtained in each vertical direction, and use these values as input parameters in advance in the horizontal and vertical directions using fuzzy rules and membership functions determined in advance. Of one-sided pressing (the amount of change in one-sided pressing) ΔEh, ΔEv to be corrected, and when it is determined that the pattern correction is necessary when at least one of the horizontal and vertical corrected one-sided pressing degrees is equal to or more than a specified value, horizontal and vertical The jack that best satisfies the degree of correction in each direction is calculated by calculating the amount of change in the degree of pressing when the jack is added or subtracted for each jack in advance, and is selected by searching the registered jack database. This is done by adding a modified jack to the current pattern (on control) or deleting it from the current pattern (off control).

In this case, the fuzzy rule for obtaining the corrected partial pressing degree is expressed in the following format.

(Horizontal direction) If the control amount Δθhi is c, the predicted control amount θhi is d, and the excavation rate α is e, the corrected partial depression degree ΔEh is y.

(Vertical direction) The control amount Δθvi is c ′,
And the predicted control amount θvi is d ′, and the excavation rate α is e ′
, The corrected partial depression degree ΔEv is y ′.

Here, c, c ', d, d', e, e 'represent the degree of the input parameter, and y, y' represent the degree of the output parameter.

FIG. 5 is a diagram showing a control locus for jack pattern correction. The control prediction amount is a control amount obtained as a result of performing linear or non-linear prediction based on the current control state. FIG. 5 shows the control amount at the time of the i-th check stroke and the control amount at the time of determining the initial jack pattern. The example which performed linear prediction from the state of the shield machine is shown.

In the jack pattern control based on the above-described fuzzy inference, problems and cases to be compared and compared in case-based inference are as shown in Tables 1 and 2 below, for example.

[0026]

[Table 1]

[0027]

[Table 2]

In this example, parameters used for control and a grade value obtained by evaluating the degree of certainty with a membership function are registered as application conditions of fuzzy inference in the case data. In addition, excavator operation data (total thrust, frontal earth pressure,
Excavation speed, cutter torque, extra excavation, etc.) are also registered as reference data, and finally a control change amount (change in control amount per unit stroke length) representing a control result is registered.

FIG. 3 shows a system flow of the present invention incorporating case-based reasoning. After performing rule-based reasoning in the course of excavating one ring, and after executing jack pattern correction control based on the reasoning result, FIG. The control result is checked in accordance with a predetermined criterion, and if the expected control result is not obtained, the control and inference application conditions performed this time by the search / inference units 8 and 8 'of FIGS. Is retrieved from the case base 6 and compared with the given problem. The case in which the inference application conditions are similar specifically refers to the same problem as the problem given the degree of the input parameter used for the control.

The above Table 1 summarizes data in the case where an expected control result is not obtained as a problem. Now, it is assumed that a case 21 shown in Table 2 is obtained as a case similar to the problem-113 shown in Table 1 and the inference application conditions. In case-21, a large control change amount is obtained as compared to problem-113.
In step 13, the cause of the failure to obtain the expected control result is determined. As a procedure, data is collated sequentially according to check items when an expected control result is not obtained. Check items are incorporated in the program as meta-knowledge in advance. Here, the items shown in Table 3 below are checked.

[0031]

[Table 3]

From the above collation results, the reason why the expected control result could not be obtained and the coping method are inferred by using the hierarchical knowledge shown in the following Table 4 incorporated in the program in advance.

[0033]

[Table 4]

As a result, in the case of Problem-113 in contrast to Case-21 in which a relatively good control result is obtained, the cause of the failure in obtaining the expected control result and a countermeasure thereof are derived as follows.

[0035] Based on the inference result, control or operation support of the corresponding shield machine is executed.

That is, in the example of FIG. 1, the search / inference unit 8 gives the following advice to the operator through display or the like.

"The excavation speed is too fast for the construction curvature. It is considered desirable to reduce the speed to about 3 cm / min. If there is no problem in face management, raise the earth pressure and reduce the front reaction force. It is easier to cut the curve if it is generated. ”Also, in the case where the control of the shield digging machine is performed automatically, for example, as shown in FIG. And excavation speed is 3cm
The operating speed of the shield jack can be controlled so as to be about / min.

In FIG. 3, when the expected control result is obtained by the control based on the rule-based inference, the program is terminated without performing the case search, and when the similar case is not obtained by the case search, Is registered as a case, and the program is terminated.

In the above embodiment, the direction control of the shield excavator has been described. However, the present invention is not limited to only the direction control.
It can also be applied to face control and the like. Further, the control rules registered in the rule base are not limited to fuzzy rules. Depending on the type of control, a rule base can be constructed using a so-called production rule.

[0040]

According to the present invention, a shield machine is controlled based on a control rule incorporating the operation knowledge of a skilled operator. By learning the cause, the cause can be determined, and the control or operation support of the shield machine can be performed according to the cause, so that it is possible to more appropriately cope with the change in the construction environment and the characteristics of the shield machine.

[Brief description of the drawings]

FIG. 1 is a block diagram showing one embodiment of the method of the present invention.

FIG. 2 is a block diagram showing another embodiment of the method of the present invention.

FIG. 3 is a system flowchart of the method of the present invention.

FIG. 4 is an explanatory diagram of a control target in horizontal direction control;
(A) is a state diagram of the shield machine at the start of excavation, (b)
FIG. 4 is a view showing a state where the shield machine should be after one ring excavation is completed.

FIG. 5 is a diagram showing a control trajectory of jack pattern correction during excavation of one ring.

FIG. 6 is an explanatory diagram of a conventional technique of jack pattern control by fuzzy inference.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... State evaluation part, 2 ... Rule base reasoning part, 3 ... Rule base, 4 ... Shield machine, 5 ... Data editing part, 6 ...
Case base, 7: Problem grasping unit, 8, 8 '... Search / inference unit.

──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Hitoshi Takahashi 650, Kandamachi, Tsuchiura-shi, Ibaraki Hitachi Construction Machinery Co., Ltd. Tsuchiura Plant (56) References JP-A-2-115492 (JP, A) JP-A-3 -228996 (JP, A) JP-A-2-18697 (JP, A) JP-A-5-5396 (JP, A) (58) Fields investigated (Int. Cl. ⁶ , DB name) E21D 9/06 301 E21D 9/06 302

Claims (3)

(57) [Claims] At the time of shield construction, a shield excavator is controlled based on an inference result based on a rule base in which control rules are registered, and the inference application conditions, control results, and other excavator operations affecting the control results are provided. By editing the data and registering it as a case, a case base is constructed.If the expected control result is not obtained by rule-based inference, past cases with similar inference application conditions are searched from the case base. A method for controlling and operating a shield machine, wherein the cause is determined and a policy for controlling or operating the shield machine is correspondingly determined. 2. A shield construction method for controlling a direction of a shield machine based on a rule-based inference result in which a control rule is registered, and a condition for applying the inference, a control result, and another excavation which affects the control result. By editing machine operation data and registering it as a case, a case base is constructed. If the expected control result is not obtained by rule-based inference, past cases with similar inference application conditions are searched from the case base. And determining the cause and determining the policy of control or operation support of the shield machine corresponding to the cause. 3. The method according to claim 1, wherein the rule base is constructed by fuzzy rules, the fuzzy parameters are also used as inference application conditions registered in the case base, and the certainty is evaluated by a membership function used in fuzzy inference. Item 3. The method for controlling and operating a shield machine according to item 1 or 2.