DE112018004244T5 - An intelligent decision making process and system for drilling control parameters of hard rock TBM - Google Patents

Abstract

An intelligent decision making process and system for drilling control parameters of hard rock TBM. The process comprises the following parts: setting up an engineering database; Establishment of a model for mutual feedback of the rock machine and a model for decision-making in intelligent control on the basis of an engineering database; display and update of model parameter output; the control method of automatic / manual drilling parameters; the model's self-learning and self-updating algorithm. The system consists of the following components: engineering database unit; Model unit for mutual feedback of the rock machine; Model unit for decision making in intelligent control; Real-time display module for the output of model parameters; Control module for automatic / manual drilling parameters; self-learning and self-updating model unit. According to the operating parameters of the device, the model for mutual feedback of the rock machine predicts the current parameters of the surrounding rock condition and records the TBM drilling environment in real time. The intelligent control decision making model predicts the optimal drilling control parameters according to the current drilling environment, adjusts the current drilling parameters in time by controlling the drilling parameters to adapt to the current drilling environment and ensure a safe, effective and stable drilling process.

Description

Technical part

The invention relates to the technical field of intelligent control of the construction process of tunnel construction systems, in particular an intelligent decision-making method and system for drilling control parameters of hard rock TBM.

Background technology

The hard rock tunnel boring machine (hereinafter referred to as hard rock TBM) is a type of large, high-tech construction machine that is used especially in rock tunnel and underground construction. In the traditional construction mode, the operator evaluates the parameters of the surrounding rock condition through the slow trial drilling process in the main control room and then repeatedly adjusts the drilling parameters until the drilling parameters remain stable. Such an operating mode, on the one hand, causes a large expenditure of time for the construction work due to the continuous slow test drilling and the repeated setting of the drilling parameters. However, if the surrounding rock changes dramatically, it cannot effectively capture the current parameters of the surrounding rock in real time. As a result, the drilling parameters cannot adapt to the current drilling environment, which leads to abnormal tool wear and a decrease in the rock breaking performance of the cutter head system. In serious cases, this leads to damage and failure of key components and affects the lifespan of TBM. It is therefore very important to find an intelligent decision-making process that not only enables the current parameters of the surrounding rock condition to be recorded in real time, but also the optimal drilling parameters suitable for the current drilling environment can be found quickly.

Summary of the invention

• Schritt 1: Erstellen Sie die Engineering-Datenbank einschließlich der Datenbank des umgebenden Gesteinszustands und Bohrparameterdatenbank.
• Schritt 2: Gemäß den Parametern des umgebenden Gesteinszustands wird der kontinuierliche umgebende Gesteinsgrad W erhalten. Unter Verwendung des dreischichtigen neuronalen Netzwerks, der Support-Vektor-Maschine und der Regressionsmethode kleinster Quadrate werden Data Mining und maschinelles Lernen der Parameter des umgebenden Gesteinszustands und die Bohrsteuerungsparameter in der Engineering-Datenbank durchgeführt. Das Modell für gegenseitige Rückmeldung der Gesteinsmaschine und das Modell für die Entscheidungsfindung bei intelligenter Steuerung werden durch mathematische Berechnung erhalten.
• Schritt 3: Das Modell für gegenseitige Rückmeldung der Gesteinsmaschine schätzt die Parameter des umgebenden Gesteinszustands der aktuellen Bohrumgebung gemäß den Bohrparametern, die online und in Echtzeit vom Hostcomputer der TBM erhalten werden. Das Modell für die Entscheidungsfindung bei intelligenter Steuerung sagt die besten Bohrsteuerungsparameter gemäß den Parametern des umgebenden Gesteinszustands voraus und zeigt die kumulativen durchschnittlichen geschätzten Parameter und die aktuellen geschätzten Parameter auf der TBM an.
• Schritt 4: Stellen Sie die aktuellen TBM-Bohrparameter automatisch oder manuell ein, indem Sie die kumulativen durchschnittlichen geschätzten Parameter und die aktuellen geschätzten Parameter verwenden.
• Schritt 5: Übertragen Sie die TBM-Echtzeit-Bohrparameter und andere TBM-Engineering-Datenbanken in die Engineering-Datenbank und geben Sie Schritt 2 ein, um das Selbstlernen und Selbstaktualisierung des Modells für gegenseitige Rückmeldung der Gesteinsmaschine und des Modells für die Entscheidungsfindung bei intelligenter Steuerung durchzuführen.

Considering the existing mode of operation, which can only guess the parameters of the surrounding rock condition based on the human experience of the operator, and the technical problems such as time and labor problems caused by trying different drilling parameters and abnormal wear of TBM Tools, decrease in cutterhead system rock fracture performance, damage and failure of key components, the present invention provides an intelligent decision making method and system for drilling control parameters of hard rock TBM. It uses data mining and machine learning technology to predict drilling parameters. Through self-learning and self-updating, it can adapt to the use of TBM for different surrounding rocks, different diameters, different performance and from the same TBM in different phases of the entire life cycle. To achieve the above purpose, the technical solution of the present invention is implemented as follows: an intelligent decision making method and system for drilling control parameters of hard rock TBM, the steps of which are as follows:

• Step 1: Create the engineering database including the database of the surrounding rock condition and drilling parameter database.
• Step 2: According to the parameters of the surrounding rock state, the continuous surrounding rock grade W is obtained. Using the three-layer neural network, the support vector machine and the least squares regression method, data mining and machine learning of the parameters of the surrounding rock state and the drilling control parameters are carried out in the engineering database. The model for mutual feedback of the rock machine and the model for decision making with intelligent control are obtained by mathematical calculation.
• Step 3: The rock machine mutual feedback model estimates the parameters of the surrounding rock condition of the current drilling environment according to the drilling parameters obtained online and in real time from the TBM host computer. The intelligent control decision making model predicts the best drilling control parameters according to the parameters of the surrounding rock condition and displays the cumulative average estimated parameters and the current estimated parameters on the TBM.
• Step 4: Set the current TBM drilling parameters automatically or manually using the cumulative average estimated parameters and the current estimated parameters.
• Step 5: Transfer the TBM real-time drilling parameters and other TBM engineering databases to the engineering database and enter Step 2 to self-learn and self-update the model for mutual feedback of the rock machine and the model for decision-making with intelligent Control.

The data contained in the database of the surrounding rock condition include: single tool thrust Ft, single tool torque Tn, penetration P, cutter head speed n and propulsion speed V. The data contained in the drilling parameter database include: uniaxial saturated compressive strength Rc of the rock, number of joints Jv per unit volume of the rock mass and surrounding rock grade W.

• Schritt 1: Die Gelenkezahl Jv pro Volumeneinheit der Gesteinsmasse wird zur Interpolation in den Knoten des Integritätsindex Kv umgewandelt. Der Anpassungswert Kv'=e^{(-0.05Jv-0.11)} wird durch Anpassung erhalten;
• Schritt 2: Gemäß dem Klassifizierungsstandard der technischen Gesteinsmasse wird der Grundqualitätsindex BQ der Gesteinsmasse unter Verwendung der einachsigen gesättigten Druckfestigkeit Rc und des Anpassungswerts Kv' berechnet.
• Schritt 3: Transformieren Sie den Basisqualitätsindex BQ der Gesteinsmasse zur Interpolation in den Knoten des umgebenden Gesteinsgrades W. Durch das Anpassen wird der Grad des umgebenden Gesteins erhalten: W=7-0,01*BQ, und der kontinuierliche Grad des umgebenden Gesteins wird erhalten.

According to the parameters of the surrounding rock condition, the method for obtaining a continuous surrounding rock grade W is as follows:

• Step 1: The number of joints Jv per unit volume of the rock mass is converted into the node of the integrity index Kv for interpolation. The adaptation ^value Kv '= e ^{(-0.05Jv-0.11)} is obtained by adaptation;
• Step 2: According to the classification standard of the technical rock mass, the basic quality index BQ of the rock mass is calculated using the uniaxial saturated compressive strength Rc and the adaptation value Kv '.
• Step 3: Transform the base quality index BQ of the rock mass for interpolation into the nodes of the surrounding rock grade W. By fitting the grade of the surrounding rock is obtained: W = 7-0.01 * BQ, and the continuous grade of the surrounding rock is obtained .

• Schritt 1: Verwenden Sie die Datenbank des umgebenden Gesteinszustands in der Engineering-Datenbank, um die Parametermatrix des umgebenden Gesteinszustands N=[Rc, Jv, W] zu berechnen. Extrahieren Sie die ersten 10% Bohrparameterdaten, die für den Bohrzyklus nützlich sind, aus der Bohrparameterdatenbank, um die Bohrparametermatrix MI=[Ft, Tn,P, n, V] in einen ansteigenden Abschnitt zu bilden;
• Schritt 2: Erstellen Sie ein dreischichtiges neuronales Netzwerk, nehmen Sie die Bohrparametermatrix MI des ansteigenden Abschnitts als Eingabe und die Parametermatrix N des umgebenden Gesteinszustands als Ausgabe und untersuchen Sie das anfängliche neuronale Netzwerk unter Aufsicht. 70% der Daten in der Bohrparameterdatenbank werden für das Training ausgewählt, 30% der Daten werden getestet und ein ausgereiftes neuronales Netzwerk Netl wird erhalten, und das Vorhersageergebnis der Parameter Ynetl des umgebenden Gesteinszustands wird erhalten;
• Schritt 3: Die Support-Vektor-Maschine wird verwendet, um die Bohrparametermatrix MI des ansteigenden Abschnitts als Eingabe, und die Parametermatrix N des umgebenden Gesteinszustands als Ausgabe zu nehmen, um die Daten zu regressieren. 70% der Daten in der Bohrparameterdatenbank werden für das Training ausgewählt, 30% der Daten werden getestet, eine ausgereifte Regressionslernmaschine svml wird erhalten und das Vorhersageergebnis der Parameter des umgebenden Gesteinszustands Ysvml wird erhalten;
• Schritt 4: Verwenden Sie die Regressionsmethode der kleinsten Quadrate, um die Bohrparametermatrix MI des ansteigenden Abschnitts als Eingabe und die Parametermatrix N des umgebenden Gesteinszustands als Ausgabe zu nehmen, um das mathematische Modell wie folgt zu erhalten: $$\begin{array}{l}\text{Rc}=64.6108+0.2061\mathrm{\ast }\frac{\text{Ft}}{1000}-0.4065\mathrm{\ast }\frac{\text{P}}{100}-90.504\mathrm{\ast }\frac{\text{Tn}}{100}+8.831\mathrm{\ast }\frac{{\text{Tn}}^{\text{2}}}{1000}\\ -4.82\mathrm{3}*n;\end{array}$$
  
  $$\begin{array}{l}\text{Jv}=27.344-0.0743\mathrm{\ast }\frac{\text{Ft}}{1000}+1.0493\mathrm{\ast }\frac{\text{P}}{100}-48.8794\mathrm{\ast }\frac{\text{Tn}}{100}+5.344\mathrm{\ast }\frac{{\text{<figure-callout id="2" label="Tn" filenames="DE112018004244T5\_0034.png" state="{{state}}">Tn</figure-callout>}}^{\text{2}}}{1000}+0.165\\ \mathrm{\ast }n;\end{array}$$
  
  $$\text{W}=4.252-8.889\mathrm{\ast }\frac{\text{Ft}}{1000}+9.146\mathrm{\ast }\frac{\text{P}}{100}-1.0664\mathrm{\ast }\frac{\text{Tn}}{100}+8.5778\mathrm{\ast }\frac{{\text{<figure-callout id="2" label="Tn" filenames="DE112018004244T5\_0034.png" state="{{state}}">Tn</figure-callout>}}^{\text{2}}}{1000}+0.253\mathrm{\ast }n$$
  
• Das Vorhersageergebnis Yregl der Parameter des umgebenden Gesteinszustands wird erhalten;
• Schritt 5: Gemäß den Vorhersageergebnissen Ynetl, Ysvml und Yregl der in Schritt 2-4 erhaltenen Parameter des umgebenden Gesteinszustands wird der mathematische Durchschnittswert berechnet: $\text{Y}1=\frac{\text{Ynet}1+\text{Ysvm}1+\text{Yreg1}}{3},$
  
  und das Modell YI für gegenseitige Rückmeldung der Gesteinsmaschine wird erhalten.

The procedure for obtaining the rock machine mutual feedback model is as follows:

• Step 1: Use the surrounding rock state database in the engineering database to calculate the parameter matrix of the surrounding rock state N = [Rc, Jv, W]. Extract the first 10% drilling parameter data useful for the drilling cycle from the drilling parameter database to form the drilling parameter matrix MI = [Ft, Tn, P, n, V] in an ascending section;
• Step 2: Create a three-layer neural network, take the drilling parameter matrix MI of the rising section as input and the parameter matrix N of the surrounding rock state as output and examine the initial neural network under supervision. 70% of the data in the drilling parameter database is selected for training, 30% of the data is tested and a mature neural network Netl is obtained, and the prediction result of the parameters Ynetl of the surrounding rock state is obtained;
• Step 3: The support vector machine is used to take the drilling parameter matrix MI of the rising section as input and the parameter matrix N of the surrounding rock state as output to regress the data. 70% of the data in the drilling parameter database are selected for training, 30% of the data are tested, a mature regression learning machine svml is obtained and the prediction result of the parameters of the surrounding rock state Ysvml is obtained;
• Step 4: Use the least squares regression method to take the drilling parameter matrix MI of the rising section as input and the parameter matrix N of the surrounding rock state as output to get the mathematical model as follows: $$\begin{array}{l}\text{Rc}=64.6108+0.2061\mathrm{\ast }\frac{\text{Ft}}{1000}-0.4065\mathrm{\ast }\frac{\text{P}}{100}-90.504\mathrm{\ast }\frac{\text{Tn}}{100}+\mathrm{8,831}\mathrm{\ast }\frac{{\text{Tn}}^{\text{2nd}}}{1000}\\ -4.82\mathrm{3rd}*n;\end{array}$$
  
  $$\begin{array}{l}\text{Jv}=\mathrm{27,344}-0.0743\mathrm{\ast }\frac{\text{Ft}}{1000}+1.0493\mathrm{\ast }\frac{\text{P}}{100}-48.8794\mathrm{\ast }\frac{\text{Tn}}{100}+\mathrm{5,344}\mathrm{\ast }\frac{{\text{Tn}}^{\text{2nd}}}{1000}+0.165\\ \mathrm{\ast }n;\end{array}$$
  
  $$\text{W}=\mathrm{4,252}-\mathrm{8,889}\mathrm{\ast }\frac{\text{Ft}}{1000}+\mathrm{9,146}\mathrm{\ast }\frac{\text{P}}{100}-1.0664\mathrm{\ast }\frac{\text{Tn}}{100}+8.5778\mathrm{\ast }\frac{{\text{Tn}}^{\text{2nd}}}{1000}+0.253\mathrm{\ast }n$$
  
• The prediction result Yregl of the parameters of the surrounding rock condition is obtained;
• Step 5: According to the prediction results Ynetl, Ysvml and Yregl of the parameters of the surrounding rock state obtained in Step 2-4, the mathematical average value is calculated: $\text{Y}1=\frac{\text{Ynet}1+\text{Ysvm}1+\text{Yreg1}}{\mathrm{3rd}},$
  
  and the YI mutual model of the rock machine is obtained.

• Schritt 1: Verwenden Sie die Datenbank des umgebenden Gesteinszustands in der Engineering-Datenbank, um die Parametermatrix des umgebenden Gesteinszustands N = [Rc, Jv, W] zu berechnen. Bitte extrahieren Sie die letzten 90% der Bohrparameterdaten, die für den Bohrzyklus nützlich sind, in der Bohrparameterdatenbank und berechnen Sie den Durchschnittswert des Bohrparameters des Bohrzyklus, um die Bohrparametermatrix M2 =[Ft, Tn, P, n, V] des stabilen Abschnitts zu bilden.
• Schritt 2: Erstellen Sie ein dreischichtiges neuronales Netzwerk, nehmen Sie die Parametermatrix N des umgebenden Gesteinszustands als Eingabe, die Bohrparametermatrix M2 des stabilen Abschnitts als Ausgabe und untersuchen Sie das anfängliche neuronale Netzwerk unter Aufsicht. Bitte wählen Sie 70% der Daten in der Bohrparameterdatenbank für das Training, und 30% der Daten für Tests aus, um ein ausgereiftes neuronales Netzwerk Net2 zu erhalten und das Vorhersageergebnis Ynet2 der Parameter des umgebenden Gesteinszustands zu erhalten.
• Schritt 3: Die Support-Vektor-Maschine wird verwendet, um die Parametermatrix N des umgebenden Gesteinszustands als Eingabe, und die Bohrparametermatrix M2 des stabilen Abschnitts wird als Ausgabe zu nehmen und eine Datenregression durchzuführen. Bitte wählen Sie 70% der Daten in der Bohrparameterdatenbank für das Training und 30% der Daten für Tests aus, um eine ausgereifte Regressionslernmaschine svm2 zu erhalten und das Vorhersageergebnis Ysvm2 der Parameter des umgebenden Gesteinszustands zu erhalten.
• Schritt 4: Unter Verwendung der Regressionsmethode der kleinsten Quadrate werden die Parametermatrix N des umgebenden Gesteinszustands wird als Eingabe und die Bohrparametermatrix M2 des stabilen Abschnitts als Ausgabe verwendet, um das mathematische Modell wie folgt zu erhalten: $$\text{Ft}=464.853-9.012\mathrm{\ast }\text{Jv}-80.4\mathrm{\ast }\text{W}+0.118\mathrm{\ast }\text{Rc}$$
  
  $$\text{Tn}=61.865-7.012\mathrm{\ast }\text{Jv}-7.4\mathrm{\ast }\text{W}+0.064\mathrm{\ast }\text{Rc}$$
  
  $$\text{P}=0.1105+0.12\mathrm{\ast }\text{Jv}\mathrm{+2}\mathrm{.446\ast }\text{W}+0.058\mathrm{\ast }\text{Rc}$$
  
  $$\text{n}=11.743+0.16\mathrm{\ast }\text{Jv}-1.132\mathrm{\ast }\text{W}\mathrm{-}0.044\mathrm{\ast }\text{Rc}$$
  
  $$\text{V}=\text{p}\mathrm{\ast }\text{n}\mathrm{,}$$
  
  Das Vorhersageergebnis Yreg2 von Bohrzustandsparametern des stabilen Abschnitts wird erhalten;
• Schritt 5: Gemäß den Vorhersageergebnissen Ynet2, Ysvm2 und Yreg2 der in Schritt 2-4 erhaltenen Bohrzustandsparameter des stabilen Abschnitts wird der mathematische Durchschnittswert berechnet: $\text{Y2}=\frac{\text{Ynet2}+\text{Ysvm2}+\text{Yreg2}}{3},$
  
  und das Modell Y2 für die Entscheidungsfindung bei intelligenter Steuerung wird erhalten.

The procedure for obtaining the intelligent control decision making model is as follows:

• Step 1: Use the surrounding rock state database in the engineering database to calculate the parameter matrix of the surrounding rock state N = [Rc, Jv, W]. Please extract the last 90% of the drilling parameter data useful for the drilling cycle in the drilling parameter database and calculate the average value of the drilling parameter of the drilling cycle in order to calculate the drilling parameter matrix M2 = [ Ft , Tn , P , n , V ] of the stable section.
• Step 2: Create a three-layer neural network, take the parameter matrix N of the surrounding rock state as input, the drilling parameter matrix M2 of the stable section as output and examine the initial neural network under supervision. Please select 70% of the data in the drilling parameter database for training and 30% of the data for testing in order to obtain a mature neural network Net2 and to obtain the prediction result Ynet2 of the parameters of the surrounding rock state.
• Step 3: The support vector machine is used to take the parameter matrix N of the surrounding rock condition as input, and the drilling parameter matrix M2 of the stable section is taken as output and perform a data regression. Please select 70% of the data in the drilling parameter database for training and 30% of the data for tests in order to obtain a mature regression learning machine svm2 and to obtain the prediction result Ysvm2 of the parameters of the surrounding rock condition.
• Step 4: Using the least squares regression method, the parameter matrix N of the surrounding rock state is used as input and the drilling parameter matrix M2 of the stable section is used as output to obtain the mathematical model as follows: $$\text{Ft}=464.853-\mathrm{9,012}\mathrm{\ast }\text{Jv}-80.4\mathrm{\ast }\text{W}+0.118\mathrm{\ast }\text{Rc}$$
  
  $$\text{Tn}=\mathrm{61,865}-\mathrm{7,012}\mathrm{\ast }\text{Jv}-7.4\mathrm{\ast }\text{W}+0.064\mathrm{\ast }\text{Rc}$$
  
  $$\text{P}=0.1105+0.12\mathrm{\ast }\text{Jv}\mathrm{+2}\mathrm{.446\; \ast }\text{W}+0.058\mathrm{\ast }\text{Rc}$$
  
  $$\text{n}=\mathrm{11,743}+0.16\mathrm{\ast }\text{Jv}-\mathrm{1,132}\mathrm{\ast }\text{W}\mathrm{-}0.044\mathrm{\ast }\text{Rc}$$
  
  $$\text{V}=\text{p}\mathrm{\ast }\text{n}\mathrm{,}$$
  
  The prediction result Yreg2 of drilling state parameters of the stable section is obtained;
• Step 5: According to the prediction results Ynet2, Ysvm2 and Yreg2 of the drilling state parameters of the stable section obtained in Step 2-4, the mathematical average value is calculated: $\text{Y2}=\frac{\text{Ynet2}+\text{Ysvm2}+\text{Yreg2}}{\mathrm{3rd}},$
  
  and the Y2 model for intelligent control decision making is obtained.

The currently estimated parameters are: the average value of the parameter Nk of the surrounding rock and the stable drilling parameter Mk by the current kth group of drilling parameters is predicted. The cumulative average estimated parameters are as follows: the parameters of the surrounding rock condition are predicted according to the Y1 mutual-machine feedback model, and the parameters of the surrounding rock Ns, which are predicted from Sections K-2 through K-1 of the drilling parameters, will get. If the average deviation between the current estimated parameters and the cumulative average estimated parameters is less than 10%, the surrounding rock is in a stable section and the impact of the current drilling parameters is stable. If the average deviation between the estimated parameters and the cumulative average estimated parameters is greater than 90%, the current drilling parameters are unstable and must be set.

The manual setting in step 4 is that the main operator of the TBM sets the current drilling parameters such as the cutter head speed n and the advance speed V in real time according to the difference between the accumulated average estimated parameters and the current estimated parameters and other drilling parameters other drilling parameters within the stable Area controls and ensures the safe and efficient drilling of TBM. The automatic setting in step 4 consists of: Depending on the difference between the pre-accumulated average estimated parameters and the current estimated parameters, the current drilling parameters such as the cutter head speed n and the advance speed V are set in real time by the PLC control and other drilling parameters are set within the stable one Area controlled to ensure the safe and efficient drilling of the TBM.

If the increase in data in the engineering database is more than 30%, self-learning and self-updating of the model for mutual feedback of the rock machine and the model for decision-making with intelligent control in TBM must be carried out manually or by host computers during the downtime or the casting time, to get a new model for mutual feedback of the rock machine and a new model for decision making with intelligent control. After the update, the current model for mutual feedback of the rock machine and the model for decision-making with intelligent control dominate. If the prediction result of the new rock machine mutual feedback model is better than that of the current rock machine mutual feedback model or that of the intelligent control decision making model, the new rock machine mutual feedback model replaces the current mutual feedback model the rock machine, and if the prediction result of the new intelligent control decision making model is better than that of the current intelligent control decision making model, the new intelligent control decision making model replaces the current intelligent control decision making model.

• Die Engineering-Datenbankeinheit, die regelmäßig die Daten des umgebenden Gesteinszustands und die Bohrparameterdaten vom Hostcomputer der TBM abruft;
• Die Modelleinheit für gegenseitige Rückmeldung der Gesteinsmaschine, die die Bohrparameter der Engineering-Datenbankeinheit verwendet, um die Parameter des umgebenden Gesteinszustands der aktuellen Bohrumgebung vorherzusagen.
• Die Modelleinheit für die Entscheidungsfindung bei intelligenter Steuerung, die gemäß den Parametern des umgebenden Gesteinszustands die besten Bohrsteuerungsparameter schätzt;
• Das Echtzeit-Anzeigemodul zur Ausgabe von Modellparametern kommuniziert mit dem Hostcomputer über die E/A-Schnittstelle und zeigt die Ausgabeparameter auf der Bedienoberfläche des Hauptführers an.
• Das Steuermodul für automatische / manuelle Bohrparameter befindet sich an der Bedienoberfläche des Hostcomputers. Der Hauptführer oder das System steuert die Bohrparameter der TBM über die SPS-Steuerung gemäß den geschätzten Bohrparametern.
• Die selbstlernende und selbstaktualisierende Modelleinheit befindet sich auf der Bedienoberfläche des Hostcomputers. Der Hauptführer legt die Modellaktualisierungszeit fest oder das Hintergrundprogramm wählt den geeigneten Zeitpunkt für die automatische Aktualisierung des Modells aus. Der geeignete Zeitpunkt umfasst, ohne darauf beschränkt zu sein, die Ausfallzeit und die Vergusszeit.
• Die Parameter der Modelleinheit für gegenseitige Rückmeldung der Gesteinsmaschine und der Modelleinheit für die Entscheidungsfindung bei intelligenter Steuerung können gelesen und geschrieben werden. Wenn die selbstlernende und selbstaktualisierende Modelleinheit gestartet wird, können die Parameter der Modelleinheit für gegenseitige Rückmeldung der Gesteinsmaschine und der Modelleinheit für die Entscheidungsfindung bei intelligenter Steuerung geschrieben werden, und andere Prozesse können nur gelesen und nicht geschrieben werden.
• Die Daten, die von der Engineering-Datenbankeinheit vom Hostcomputer der TBM erhalten werden, umfassen die Echtzeit-TBM-Bohrparameter und andere Engineering-Datenbanken der TBM. Die Engineering-Datenbank anderer TBM umfasst, ohne darauf beschränkt zu sein, die Engineering-Datenbank von TBM desselben Strukturtyps oder einer ähnlichen geologischen Ausgrabung.
• Der vorteilhafte Effekt der vorliegenden Erfindung besteht darin, dass die Modelleinheit für gegenseitige Rückmeldung der Gesteinsmaschine die aktuellen Parameter des umgebenden Gesteinszustands gemäß den Betriebsparametern des Geräts vorhersagt und die TBM-Bohrumgebung in Echtzeit erfasst. Entsprechend der aktuellen Bohrumgebung sagt die Modelleinheit für die Entscheidungsfindung bei intelligenter Steuerung die optimalen Bohrsteuerungsparameter voraus. Über die Modellparameter-Ausgabeeinheit werden die Parameter des umgebenden Gesteinszustands und die optimalen Bohrsteuerungsparameter in Echtzeit auf der Bedienoberfläche des Hauptführers angezeigt. Das automatische / manuelle Bohrparameter-Steuermodul wird verwendet, um den Bohrparameter-Steuermodus auszuwählen, um die aktuellen Bohrparameter rechtzeitig gemäß den empfohlenen optimalen Bohrsteuerungsparametern einzustellen, um sich an die aktuelle Bohrumgebung anzupassen und einen sicheren, effektiven und stabilen Bohrprozess zu gewährleisten. Gleichzeitig verfügt es über selbstlernende und selbstaktualisierende Modelleinheit, um sich an die Verwendung von TBM für unterschiedliche umgebende Gesteine, unterschiedlichen Durchmesser, unterschiedliche Leistung und von derselben TBM in verschiedenen Phasen des gesamten Lebenszyklus anzupassen.

An intelligent decision making process and system for drilling control parameters of hard rock TBM includes:

• The engineering database unit that periodically retrieves the surrounding rock condition data and drilling parameter data from the TBM host computer;
• The rock machine mutual feedback model unit that uses the drilling parameters of the engineering database unit to predict the parameters of the surrounding rock condition of the current drilling environment.
• The model unit for intelligent control decision making that estimates the best drilling control parameters according to the parameters of the surrounding rock condition;
• The real-time display module for the output of model parameters communicates with the host computer via the I / O interface and displays the output parameters on the main operator's user interface.
• The control module for automatic / manual drilling parameters is located on the user interface of the host computer. The main operator or the system controls the drilling parameters of the TBM via the PLC control according to the estimated drilling parameters.
• The self-learning and self-updating model unit is located on the user interface of the host computer. The main guide sets the model update time or the background program selects the appropriate time for the automatic update of the model. The appropriate time includes, but is not limited to, the downtime and the casting time.
• The Parameters of the model unit for mutual feedback of the rock machine and the model unit for decision-making in intelligent control can be read and written. When the self-learning and self-updating model unit is started, the parameters of the model unit for mutual feedback of the rock machine and the model unit for decision making with intelligent control can be written, and other processes can only be read and not written.
• The data obtained from the engineering database unit from the TBM host computer includes the real-time TBM drilling parameters and other TBM engineering databases. The engineering database of other TBMs includes, but is not limited to, the engineering database of TBMs of the same structure type or a similar geological excavation.
• The advantageous effect of the present invention is that the model unit for mutual feedback of the rock machine predicts the current parameters of the surrounding rock state according to the operating parameters of the device and detects the TBM drilling environment in real time. According to the current drilling environment, the model unit for intelligent control decision making predicts the optimal drilling control parameters. The parameters of the surrounding rock condition and the optimal drilling control parameters are displayed in real time on the operator's user interface via the model parameter output unit. The automatic / manual drilling parameter control module is used to select the drilling parameter control mode to set the current drilling parameters in time according to the recommended optimal drilling control parameters, to adapt to the current drilling environment and to ensure a safe, effective and stable drilling process. At the same time, it has a self-learning and self-updating model unit to adapt to the use of TBM for different surrounding rocks, different diameters, different performance and from the same TBM in different phases of the entire life cycle.

Figure list

• 1 ist das Flussdiagramm des intelligenten Entscheidungsfindungsverfahrens für TBM-Bohrsteuerungsparameter gemäß der vorliegenden Erfindung.
• 2 ist das Strukturdiagramm des intelligenten Entscheidungsfindungssystems der TBM-Bohrsteuerungsparameter gemäß der vorliegenden Erfindung.
• 3 ist das Flussdiagramm zum Erhalten des Modells für gegenseitige Rückmeldung der Gesteinsmaschine gemäß der vorliegenden Erfindung.
• 4 ist ein Flussdiagramm zum Erhalten des Modells für die Entscheidungsfindung bei intelligenter Steuerung gemäß der vorliegenden Erfindung.
• 5 ist ein Flussdiagramm der Berechnung und Ausgabe der Modellparameter gemäß der vorliegenden Erfindung.
• 6 ist ein Flussdiagramm des Selbstlernens und der Selbstaktualisierung des Modells gemäß der vorliegenden Erfindung.

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, a brief introduction is given below to the figures which are required in the embodiments or in the description of the prior art. Obviously, the figures in the following description are only a few embodiments of the present invention. For the ordinary technicians in the field, other figures according to these figures can be obtained without creative work.

• 1 Figure 11 is the flowchart of the intelligent decision making process for TBM drilling control parameters in accordance with the present invention.
• 2nd Fig. 4 is the structural diagram of the intelligent decision making system of the TBM drilling control parameters according to the present invention.
• 3rd Fig. 4 is the flowchart for obtaining the rock machine mutual feedback model according to the present invention.
• 4th FIG. 14 is a flowchart for obtaining the intelligent control decision making model according to the present invention.
• 5 is a flowchart of the calculation and output of the model parameters according to the present invention.
• 6 10 is a flowchart of self-learning and self-updating of the model according to the present invention.

Embodiments

The technical solution of the present utility model is clearly and completely described with the aid of the attached figures. Obviously, the described embodiment is only part of the embodiments of the utility model, not all of them. Based on the embodiment of the utility model, all other embodiments obtained by ordinary technicians in the field without creative work belong to the scope of the present utility model.

• Schritt 1: Erstellen Sie die Engineering-Datenbank einschließlich der Datenbank des umgebenden Gesteinszustands und Bohrparameterdatenbank.

As in 1 , an intelligent decision making process for drilling control parameters of hard rock TBM is provided. The steps are as follows:

• Step 1: Create the engineering database including the database of the surrounding rock condition and drilling parameter database.

The engineering database contains the parameter database of the surrounding rock condition and the drilling control parameter database for highly efficient drilling, which are used as examples for later use. The data contained in the database of the surrounding rock condition must include at least the following: single tool thrust Ft, single tool torque Tn, penetration P, cutter head speed n and advance speed V. The data contained in the drilling parameter database must include at least the following: uniaxial saturated compressive strength Rc of the rock, number of joints Jv pro Volume unit of the rock mass and basic quality grade W of the rock mass. Please take core samples of the surrounding rock in the area where the parameter database of the surrounding rock state is located and draw the joint diagram of the surrounding rock. Perform an indoor experiment according to the rock core obtained to obtain the uniaxial saturated compressive strength Rc. Count the value of the number of joints Jv per unit volume of the rock mass according to the joint diagram of the surrounding rock near the rock core sampling stack number and the classification standard for technical rock masses.

Step 2: According to the parameters of the surrounding rock state, the continuous surrounding rock grade W is obtained. Using the three-layer neural network, the support vector machine and the least squares regression method, data mining and machine learning of the parameters of the surrounding rock state and the drilling control parameters are carried out in the engineering database. The model for mutual feedback of the rock machine and the model for decision making with intelligent control are obtained by mathematical calculation.

The rock machine mutual feedback model is obtained based on information from the parameter database of the surrounding rock condition and the drilling control parameter database of the rising section. The intelligent control decision making model is obtained based on information from the parameter database of the surrounding rock condition and the drilling control parameter database of the stable section. The breakdown of the rising and stable portion of the drilling control parameters is the 10%. To take the data point of the drilling cycle as the segmentation critical point. The segmentation critical point includes, but is not limited to, the 10%. Data point of the drilling cycle and can also be other location points of neighboring proportions. The methods of mathematical calculation include, but are not limited to, mathematical averages, geometric averages, or other weighted averages. It uses the rock machine mutual feedback model to predict the parameters of the surrounding rock condition in the current drilling environment. It uses the decision making model with intelligent control of the rock machine to predict the best drilling parameters.

• Schritt 1: Die Gelenkezahl Jv pro Volumeneinheit der Gesteinsmasse wird zur Interpolation in den Knoten des Integritätsindex Kv umgewandelt, und der Anpassungswert Kv'=e^{(-00.5Jv-0.11)} wird durch Anpassung erhalten.

When converting the Jv value of the number of joints per unit volume of the rock mass into the Kv value of the integrity index, discrete Kv values are obtained through the transformation method specified by the technical classification standard for the rock mass, which is not conducive to the realization of fine control . Therefore, the present invention improves the process and uses the value at the node to perform adjustment and interpolation. According to the parameters of the surrounding rock condition, the method for obtaining the continuous surrounding rock grade W is as follows:

• Step 1: The number of joints Jv per unit volume of the rock mass is converted into the node of the integrity index Kv for interpolation, and the adaptation ^value Kv '= e ^{(-00.5Jv-0.11)} is obtained by adaptation.

According to the calculation, the mean square error of the adjustment value Kv 'and the real value Kv MSE = 0.016, the relative error is less than 6% and the prediction accuracy is greater than 94%, which meets the technical requirements.

Step 2: According to the classification standard of the technical rock mass, the basic quality index BQ of the rock mass is calculated using the uniaxial saturated compressive strength Rc and the adaptation value Kv '.

Step 3: When converting the basic quality index BQ value of the rock mass into the surrounding rock grade W, the classification standard of the technical rock mass is only divided linearly into 5 classes, which is not conducive to the implementation of fine control. Therefore, the present invention improves the process and uses the value at the node to perform adjustment and interpolation. The basic quality index BQ of the rock mass is used for interpolation in the nodes of the surrounding Rock grade W converted. The surrounding rock grade is obtained by the adjustment: W = 7-0.01 * BQ, and the surrounding rock grade W is a continuous output.

• Schritt 1: Verwenden Sie die Datenbank des umgebenden Gesteinszustands in der Engineering-Datenbank, um die Parametermatrix des umgebenden Gesteinszustands N=[Rc, Jv, W] zu berechnen. Extrahieren Sie die ersten 10% Bohrparameterdaten, die für den Bohrzyklus nützlich sind, aus der Bohrparameterdatenbank, um die Bohrparametermatrix MI=[Ft, Tn,P, n, V] in einen ansteigenden Abschnitt zu bilden.
• Schritt 2: Erstellen Sie ein dreischichtiges neuronales Netzwerk, nehmen Sie die Bohrparametermatrix MI des ansteigenden Abschnitts als Eingabe und die Parametermatrix N des umgebenden Gesteinszustands als Ausgabe und untersuchen Sie das anfängliche neuronale Netzwerk unter Aufsicht. 70% der Daten in der Bohrparameterdatenbank werden für das Training ausgewählt, 30% der Daten werden getestet und ein ausgereiftes neuronales Netzwerk Netl wird erhalten, und das Vorhersageergebnis der Parameter Ynetl des umgebenden Gesteinszustands wird erhalten.

As in 3rd shown, the process for obtaining the rock machine mutual feedback model is as follows:

• Step 1: Use the surrounding rock state database in the engineering database to calculate the parameter matrix of the surrounding rock state N = [Rc, Jv, W]. Extract the first 10% drilling parameter data useful for the drilling cycle from the drilling parameter database to form the drilling parameter matrix MI = [Ft, Tn, P, n, V] in an ascending section.
• Step 2: Create a three-layer neural network, take the drilling parameter matrix MI of the rising section as input and the parameter matrix N of the surrounding rock state as output and examine the initial neural network under supervision. 70% of the data in the drilling parameter database are selected for training, 30% of the data are tested and a mature neural network Netl is obtained, and the prediction result of the parameters Ynetl of the surrounding rock state is obtained.

For a three-layer neural network, the number of nodes in each layer is 20, 10 and 10. After testing, the prediction accuracy of the neural network Netl for the uniaxial saturated compressive strength Rc of the rock, the number of joints Jv per unit volume of the rock mass and the basic quality grade W the rock mass 92.7%, 85.4% and 95.9%, which meets the technical requirements.

Step 3: The support vector machine is used to take the drilling parameter matrix MI of the rising section as input and the parameter matrix N of the surrounding rock state as output to regress the data. 70% of the data in the drilling parameter database are selected for training, 30% of the data are tested, a mature regression learning machine svml is obtained and the prediction result of the parameters of the surrounding rock state Ysvml is obtained.

In order to improve the prediction accuracy, the support vector machine SVM is used for the data regression. After testing, the prediction accuracy of the regression learning machine svml for the uniaxial saturated compressive strength Rc of the rock, the number of joints Jv per unit volume of the rock mass and the basic quality level W of the rock mass is 93.2%, 84.2% and 94.7%, respectively, which is the technical Requirements fulfilled.

$$\begin{array}{l}\text{Jv}=27.344-0.0743\mathrm{\ast }\frac{\text{Ft}}{1000}+1.0493\mathrm{\ast }\frac{\text{P}}{100}-48.8794\mathrm{\ast }\frac{\text{Tn}}{100}+5.344\mathrm{\ast }\frac{{\text{Tn}}^{\text{2}}}{1000}+0.165\\ \mathrm{\ast }n;\end{array}$$

$$\text{W}=4.252-8.889\mathrm{\ast }\frac{\text{Ft}}{1000}+9.146\mathrm{\ast }\frac{\text{P}}{100}-1.0664\mathrm{\ast }\frac{\text{Tn}}{100}+8.5778\mathrm{\ast }\frac{{\text{Tn}}^{\text{2}}}{1000}+0.253\mathrm{\ast }n;$$

Das Vorhersageergebnis Yregl der Parameter des umgebenden Gesteinszustands wird erhalten.Step 4: Use the least squares regression method to take the drilling parameter matrix MI of the rising section as input and the parameter matrix N of the surrounding rock state as output to get the mathematical model as follows: $$\begin{array}{l}\text{Rc}=64.6108+0.2061\mathrm{\ast }\frac{\text{Ft}}{1000}-0.4065\mathrm{\ast }\frac{\text{P}}{100}-90.504\mathrm{\ast }\frac{\text{Tn}}{100}+\mathrm{8,831}\mathrm{\ast }\frac{{\text{Tn}}^{\text{2nd}}}{1000}\\ -\mathrm{4,823}\mathrm{\ast }n;\end{array}$$

$$\begin{array}{l}\text{Jv}=\mathrm{27,344}-0.0743\mathrm{\ast }\frac{\text{Ft}}{1000}+1.0493\mathrm{\ast }\frac{\text{P}}{100}-48.8794\mathrm{\ast }\frac{\text{Tn}}{100}+\mathrm{5,344}\mathrm{\ast }\frac{{\text{Tn}}^{\text{2nd}}}{1000}+0.165\\ \mathrm{\ast }n;\end{array}$$

$$\text{W}=\mathrm{4,252}-\mathrm{8,889}\mathrm{\ast }\frac{\text{Ft}}{1000}+\mathrm{9,146}\mathrm{\ast }\frac{\text{P}}{100}-1.0664\mathrm{\ast }\frac{\text{Tn}}{100}+8.5778\mathrm{\ast }\frac{{\text{Tn}}^{\text{2nd}}}{1000}+0.253\mathrm{\ast }n;$$

The prediction result Yregl of the parameters of the surrounding rock condition is obtained.

Neural network or support vector machine (SVM) are used to solve the implicit function model so that the physical interpretation is not significant. Therefore, the present invention regresses data according to the least square regression method. After testing, the relative accuracy of the above three models of uniaxial saturated prediction results Compressive strength Rc of the rock, number of joints Jv per unit volume of the rock mass and surrounding rock grade W 92.5%, 82.4% and 90.3%, which meets the technical requirements.

und das Modell YI für gegenseitige Rückmeldung der Gesteinsmaschine wird erhalten.Step 5: According to the prediction results Ynetl, Ysvml and Yregl of the parameters of the surrounding rock state obtained in Step 2-4, the mathematical average value is calculated: $\text{Y}1=\frac{\text{Ynet}1+\text{Ysvm}1+\text{Yreg1}}{\mathrm{3rd}},$

and the YI mutual model of the rock machine is obtained.

• Schritt 1: Verwenden Sie die Datenbank des umgebenden Gesteinszustands in der Engineering-Datenbank, um die Parametermatrix des umgebenden Gesteinszustands N = [Rc, Jv, W] zu berechnen. Bitte extrahieren Sie die letzten 90% der Bohrparameterdaten, die für den Bohrzyklus nützlich sind, in der Bohrparameterdatenbank und berechnen Sie den Durchschnittswert des Bohrparameters des Bohrzyklus, um die Bohrparametermatrix M2 =[Ft, Tn, P, n, V]des stabilen Abschnitts zu bilden.
• Schritt 2: Erstellen Sie ein dreischichtiges neuronales Netzwerk, nehmen Sie die Parametermatrix N des umgebenden Gesteinszustands als Eingabe, die Bohrparametermatrix M2 des stabilen Abschnitts als Ausgabe und untersuchen Sie das anfängliche neuronale Netzwerk unter Aufsicht. Bitte wählen Sie 70% der Daten in der Bohrparameterdatenbank für das Training, und 30% der Daten für Tests aus, um ein ausgereiftes neuronales Netzwerk Net2 zu erhalten und das Vorhersageergebnis Ynet2 der Parameter des umgebenden Gesteinszustands zu erhalten.

According to the prediction results of the model of the neural network model, the model of the support vector machine and the model of the least squares regression, the average mathematical value is calculated, which not only avoids the calculation error caused by a single mathematical method in data prediction , but also the benefits of each of the three models are adequately presented and the accuracy of the prediction is improved. As in 4th shown, the procedure for obtaining the model for intelligent control decision making is as follows:

• Step 1: Use the surrounding rock state database in the engineering database to calculate the parameter matrix of the surrounding rock state N = [Rc, Jv, W]. Please extract the last 90% of the drilling parameter data useful for the drilling cycle in the drilling parameter database and calculate the average value of the drilling parameter of the drilling cycle in order to calculate the drilling parameter matrix M2 = [ Ft , Tn , P , n , V ] of the stable section.
• Step 2: Create a three-layer neural network, take the parameter matrix N of the surrounding rock state as input, the drilling parameter matrix M2 of the stable section as output and examine the initial neural network under supervision. Please select 70% of the data in the drilling parameter database for training and 30% of the data for testing in order to obtain a mature neural network Net2 and to obtain the prediction result Ynet2 of the parameters of the surrounding rock state.

For a three-layer neural network, the number of nodes in each layer is 20, 10 and 10. After testing, the prediction accuracy of the neural network is Net2 for Ft , Tn , P , n , v 93.4%, 82.4%, 91.9%, 95.4% and 88.9%, which meets the technical requirements.

Step 3: The support vector machine is used to take the parameter matrix N of the surrounding rock condition as input, and the drilling parameter matrix M2 of the stable section is taken as output and perform a data regression. Please select 70% of the data in the drilling parameter database for training and 30% of the data for tests in order to obtain a mature regression learning machine svm2 and to obtain the prediction result Ysvm2 of the parameters of the surrounding rock condition.

To improve the prediction accuracy, support vector machine (SVM) is used for the data regression. After testing, the prediction accuracy of the regression learning machine is svm2 for Ft , Tn , P , n , v 92.6%, 84.7%, 90.3%, 94.8% and 90.2%, which meets the technical requirements.

$$\text{Tn}=61.865-7.012\mathrm{\ast }\text{Jv}-7.4\mathrm{\ast }\text{W}+0.064\mathrm{\ast }\text{Rc}$$

$$\text{P}=0.1105+0.12\mathrm{\ast }\text{Jv}\mathrm{+2}\mathrm{.446\ast }\text{W}+0.058\mathrm{\ast }\text{Rc}$$

$$\text{n}=11.743+0.16\mathrm{\ast }\text{Jv}-1.132\mathrm{\ast }\text{W}\mathrm{-}0.044\mathrm{\ast }\text{Rc}$$

$$\text{V}=\text{p}\mathrm{\ast }\text{n};$$

Step 4: Using the least squares regression method, the parameter matrix N of the surrounding rock state is used as input and the drilling parameter matrix M2 of the stable section is used as output to obtain the mathematical model as follows: $$\text{Ft}=464.853-\mathrm{9,012}\mathrm{\ast }\text{Jv}-80.4\mathrm{\ast }\text{W}+0.118\mathrm{\ast }\text{Rc}$$

$$\text{Tn}=\mathrm{61,865}-\mathrm{7,012}\mathrm{\ast }\text{Jv}-7.4\mathrm{\ast }\text{W}+0.064\mathrm{\ast }\text{Rc}$$

$$\text{P}=0.1105+0.12\mathrm{\ast }\text{Jv}\mathrm{+2}\mathrm{.446\; \ast }\text{W}+0.058\mathrm{\ast }\text{Rc}$$

$$\text{n}=\mathrm{11,743}+0.16\mathrm{\ast }\text{Jv}-\mathrm{1,132}\mathrm{\ast }\text{W}\mathrm{-}0.044\mathrm{\ast }\text{Rc}$$

$$\text{V}=\text{p}\mathrm{\ast }\text{n};$$

The prediction result Yreg2 of drilling state parameters of the stable section is obtained.

Neural networks or support vector machine (SVM) are used to solve the implicit function model so that the physical interpretation is not significant. Therefore, the data is trained and predicted using the least squares regression method. The relative accuracy of the above five mathematical models is 91.7%, 83.6%, 90.3%, 92.8% and 87.6%, which meets the technical requirements.

und das Modell Y2 für die Entscheidungsfindung bei intelligenter Steuerung wird erhalten.Step 5: According to the prediction results Ynet2, Ysvm2 and Yreg2 of the drilling state parameters of the stable section obtained in Step 2-4, the mathematical average value is calculated: $\text{Y2}=\frac{\text{Ynet2}+\text{Ysvm2}+\text{Yreg2}}{\mathrm{3rd}},$

and the Y2 model for intelligent control decision making is obtained.

When using the surrounding rock parameters to predict the drilling state parameters of the stable section, the forecast results Ynet2, Ysvm2, Yreg2 are obtained simultaneously by calculating the mathematical average value according to the three models. The mathematical average of the prediction results of the three models is calculated, not only avoiding the calculation error caused by a single mathematical method in data prediction, but also appropriately presenting the advantages of each of the three models and improving the accuracy of the prediction. Step 3: The rock machine mutual feedback model estimates the parameters of the surrounding rock condition of the current drilling environment according to the drilling parameters obtained online and in real time from the TBM host computer. The intelligent control decision making model predicts the best drilling control parameters according to the parameters of the surrounding rock condition and displays the cumulative average estimated parameters and the current estimated parameters on the TBM.

As in 5 shown, the drilling parameters are saved once a second and uploaded to the host computer when the tunnel boring machine (TBM) enters a new drilling cycle. The drilling parameters are read by the host computer every 10 seconds and saved as m1 m2, ... mk, provided that the current drilling parameter is in the k. Section is located. The interval time can be set. In general, the interval time can be shortened if the condition of the surrounding rock changes dramatically. If the condition of the surrounding rock is relatively stable, the interval time can be extended accordingly. According to model Y1 for mutual feedback of the rock machine, the parameters of the surrounding rock condition are estimated and the parameters of the surrounding rock condition are obtained. According to model Y2 for intelligent decision making of the drilling parameters, the drilling parameters of the stable section are estimated. The currently estimated parameters are: the average value of the parameter Nk of the surrounding rock and the stable drilling parameter Mk, which is predicted by the current kth group of drilling parameters. The cumulative average estimated parameters are as follows: the parameters of the surrounding rock condition are predicted according to the Y1 mutual-machine feedback model, and the parameters of the surrounding rock Ns, which are predicted from Sections K-2 through K-1 of the drilling parameters, will get. The current estimated parameters are displayed in the second column of the user interface and the cumulative average estimated parameters are displayed in the first column of the user interface.

If the average deviation between the current estimated parameters and the cumulative average estimated parameters is less than 10%, the surrounding rock is in a stable section and the impact of the current drilling parameters is stable. If the average deviation between the estimated parameters and the cumulative average estimated parameters is greater than 90%, the current drilling parameters are unstable and must be set.

Step 4: Set the current TBM drilling parameters automatically or manually using the cumulative average estimated parameters and the current estimated parameters.

Manual mode is selected for the method of controlling the drilling parameters, i.e. the manual setting is that the main leader of the TBM has the current drilling parameters such as the cutter head speed n and the advance speed V in real time according to the difference between the accumulated average estimated parameters and the current one estimated parameters and other drilling parameters controls other drilling parameters within the stable range and ensures the safe and efficient drilling of TBM. The automatic mode is selected for the method for controlling the drilling parameters, ie the automatic setting consists of: According to the difference between the pre-accumulated average estimated parameters and the current estimated parameters, the current drilling parameters such as the cutter head speed n and the advance speed V are Control set in real time and other drilling parameters are controlled within the stable range to ensure safe and efficient drilling of the TBM.

Step 5: Transfer the TBM real-time drilling parameters and other TBM engineering databases to the engineering database and enter Step 2 to self-learn and self-update the model for mutual feedback of the rock machine and the model for decision-making with intelligent Control.

The model's self-learning and self-updating algorithm is used to update the rock machine mutual feedback model and the intelligent control decision making model in real time and quickly. The model's self-learning and self-updating algorithm is derived from the continuous enrichment of technical data during the drilling process. The drilling parameters can be read directly from the host computer of this device from the engineering database or called from the engineering database of other TBMs. Other TBMs include, but are not limited to, TBMs of the same structure type or a similar geological excavation. The new engineering database is used to update the rock machine mutual feedback model and the intelligent control decision making model in real time and quickly to reflect the use of TBM for different surrounding rocks, different diameters, different performance and adapt from the same TBM in different phases of the entire life cycle. As the engineering database sample database grows, you can manually choose when to update the model, or the background program can choose the appropriate time to automatically update the model. The appropriate time includes, but is not limited to, downtime, casting time, etc.

As in 6 if the increase in the data in the engineering database is more than 30%, self-learning and self-updating of the model for mutual feedback of the rock machine and the model for decision-making with intelligent control in TBM are manual or by host computers during the downtime or the casting time to obtain a new model for mutual feedback of the rock machine and a new model for decision-making with intelligent control. After the update, the current model for mutual feedback of the rock machine and the model for decision-making with intelligent control dominate. If the prediction result of the new rock machine mutual feedback model is better than that of the current rock machine mutual feedback model or that of the intelligent control decision making model, the new rock machine mutual feedback model replaces the current mutual feedback model of the rock machine, and if the prediction result of the new intelligent control decision making model is better than that of the current intelligent control decision making model, the new intelligent control decision making model replaces the current intelligent control decision making model.

• Die Engineering-Datenbankeinheit, die regelmäßig die Daten des umgebenden Gesteinszustands und die Bohrparameterdaten vom Hostcomputer der TBM abruft. Die Daten, die von der Engineering-Datenbankeinheit vom Hostcomputer der TBM erhalten werden, umfassen die Echtzeit-TBM-Bohrparameter und andere Engineering-Datenbanken der TBM. Die Engineering-Datenbank anderer TBM umfasst, ohne darauf beschränkt zu sein, die Engineering-Datenbank von TBM desselben Strukturtyps oder einer ähnlichen geologischen Ausgrabung.

As in 2nd , an intelligent decision making system for drilling control parameters of hard rock TBM includes:

• The engineering database unit, which regularly retrieves the data of the surrounding rock condition and the drilling parameter data from the host computer of the TBM. The data obtained from the engineering database unit from the TBM host computer includes the real-time TBM drilling parameters and other TBM engineering databases. The engineering database of other TBMs includes, but is not limited to, the engineering database of TBMs of the same structure type or a similar geological excavation.

The rock machine mutual feedback model unit that uses the drilling parameters of the engineering database unit to predict the parameters of the surrounding rock condition of the current drilling environment. The model unit for intelligent control decision making that estimates the best drilling control parameters according to the parameters of the surrounding rock condition.

The parameters of the model unit for mutual feedback of the rock machine and the model unit for decision-making in intelligent control can be read and written. When the self-learning and self-updating model unit is started, the parameters of the model unit for mutual feedback of the rock machine and the model unit for decision making with intelligent control can be written, and other processes can only be read and not written.

The real-time display module for the output of model parameters communicates with the host computer via the I / O interface and displays the output parameters on the main operator's user interface. The output parameters include the cumulative average estimated parameters and current estimated parameters used to control TBM drilling.

The control module for automatic / manual drilling parameters is located on the user interface of the host computer. The main operator or the system controls the drilling parameters of the TBM via the PLC control according to the estimated drilling parameters.

The self-learning and self-updating model unit is located on the user interface of the host computer. The main guide sets the model update time or the background program selects the appropriate time for the automatic update of the model. The appropriate time includes, but is not limited to, the downtime and the casting time.

The engineering database unit of the present invention can continuously enrich the large database of technical parameters. Using data mining and machine learning technology, the rock machine mutual feedback model and intelligent control decision making model are created using the neural network, support vector machine, and least squares regression method, and drilling parameters are estimated. The model's parameter prediction results are displayed on the host computer's user interface via the real-time display unit module of the model parameter output, and the drilling parameter control mode can be selected via the automatic / manual drilling parameter control module. The self-learning and self-updating model unit can carry out the self-learning and self-updating, the model for mutual feedback of the rock machine and the model for decision-making with intelligent control in accordance with the continuous updating of the large database of technical parameters. It is used to adapt to the use of TBM for different surrounding rocks, different diameters, different performance and from the same TBM in different phases of the entire life cycle.

An intelligent decision making method and system for drilling control parameters of TBM of the present invention is described in detail above. However, the description of the above development steps is only for a better understanding of the method and the core idea of the present invention and is not to be understood as a limitation of the invention. Any change or replacement that will occur to those skilled in the art in the technical field of the present invention is within the scope of the present invention.

Claims (11)

An intelligent decision making process for drilling control parameters of hard rock TBM, characterized in that it comprises the following steps: Step 1: Create the engineering database including the database of the surrounding rock condition and drilling parameter database. Step 2: According to the parameters of the surrounding rock state, the continuous surrounding rock grade W is obtained. Using the three-layer neural network, the support vector machine and the least squares regression method, data mining and machine learning of the parameters of the surrounding rock state and the drilling control parameters are carried out in the engineering database. The model for mutual feedback of the rock machine and the model for decision making with intelligent control are obtained by mathematical calculation. Step 3: The rock machine mutual feedback model estimates the parameters of the surrounding rock condition of the current drilling environment according to the drilling parameters obtained online and in real time from the TBM host computer. The intelligent control decision making model says the best drilling control parameters according to the parameters of the surrounding rock condition ahead and displays the cumulative average estimated parameters and the current estimated parameters on the TBM. Step 4: Set the current TBM drilling parameters automatically or manually using the cumulative average estimated parameters and the current estimated parameters. Step 5: Transfer the TBM real-time drilling parameters and other TBM engineering databases to the engineering database and enter Step 2 to self-learn and self-update the model for mutual feedback of the rock machine and the model for decision-making with intelligent Control. An intelligent decision making process for drilling control parameters from hard rock TBM Claim 1 , characterized in that the data contained in the database of the surrounding rock state include the single tool thrust Ft, single tool torque Tn, penetration P, cutter head speed n and propulsion speed V and the data contained in the drilling parameter database the uniaxial saturated compressive strength Rc of the rock, number of joints Jv per unit volume of the Include rock mass and surrounding rock grade W. An intelligent decision making process for drilling control parameters from hard rock TBM Claim 1 or 2nd , characterized in that the method for obtaining a continuous surrounding rock grade W according to the parameters of the surrounding rock condition is as follows: Step 1: The number of joints Jv per unit volume of the rock mass is converted to the node of the integrity index Kv for interpolation. The adaptation ^value Kv '= e ^{(-0.05Jv-0.11)} is obtained by adaptation. Step 2: According to the classification standard of the technical rock mass, the basic quality index BQ of the rock mass is calculated using the uniaxial saturated compressive strength Rc and the adaptation value Kv '. Step 3: Transform the base quality index BQ of the rock mass for interpolation into the nodes of the surrounding rock grade W. By fitting the grade of the surrounding rock is obtained: W = 7-0.01 * BQ, and the continuous grade of the surrounding rock is obtained . An intelligent decision making process for drilling control parameters from hard rock TBM Claim 3 , characterized in that the procedure for obtaining the model for mutual feedback of the rock machine is as follows: Step 1: Use the database of the surrounding rock state in the engineering database to obtain the parameter matrix of the surrounding rock state N = [Rc, Jv, W ] to calculate. Extract the first 10% drilling parameter data useful for the drilling cycle from the drilling parameter database to form the drilling parameter matrix MI = [Ft, Tn, P, n, V] in an ascending section. Step 2: Create a three-layer neural network, take the drilling parameter matrix MI of the rising section as input and the parameter matrix N of the surrounding rock state as output and examine the initial neural network under supervision. 70% of the data in the drilling parameter database is selected for training, 30% of the data is tested and a mature neural network Netl is obtained, and the prediction result of the parameters Ynetl of the surrounding rock state is obtained. Step 3: The support vector machine is used to take the drilling parameter matrix MI of the ascending section as input and the parameter matrix N of the surrounding rock state as output to regress the data. 70% of the data in the drilling parameter database are selected for training, 30% of the data are tested, a sophisticated regression learning machine svml is obtained and the prediction result Ysvml of the parameters of the surrounding rock state is obtained. Step 4: Use the least squares regression method to take the drilling parameter matrix MI of the rising section as input and the parameter matrix N of the surrounding rock state as output to get the mathematical model as follows: $$\begin{array}{l}\text{Rc}=64.6108+0.2061\mathrm{\ast }\frac{\text{Ft}}{1000}-0.4065\mathrm{\ast }\frac{\text{P}}{100}-90.504\mathrm{\ast }\frac{\text{Tn}}{100}+\mathrm{8,831}\mathrm{\ast }\frac{{\text{Tn}}^{\text{2nd}}}{1000}\\ -\mathrm{4,823}*n;\end{array}$$

$$\begin{array}{l}\text{Jv}=\mathrm{27,344}-0.0743\mathrm{\ast }\frac{\text{Ft}}{1000}+1.0493\mathrm{\ast }\frac{\text{P}}{100}-48.8794\mathrm{\ast }\frac{\text{Tn}}{100}+\mathrm{5,344}\mathrm{\ast }\frac{{\text{Tn}}^{\text{2nd}}}{1000}+0.165\\ \mathrm{\ast }n;\end{array}$$

$$\text{W}=\mathrm{4,252}-\mathrm{8,889}\mathrm{\ast }\frac{\text{Ft}}{1000}+\mathrm{9,146}\mathrm{\ast }\frac{\text{P}}{100}-1.0664\mathrm{\ast }\frac{\text{Tn}}{100}+8.5778\mathrm{\ast }\frac{{\text{Tn}}^{\text{2nd}}}{1000}+0.253\mathrm{\ast }n;$$

The prediction result Yregl of the parameters of the surrounding rock condition is obtained; Step 5: According to the prediction results Ynetl, Ysvml and Yregl of the parameters of the surrounding rock state obtained in Step 2-4, the mathematical average value is calculated: $\text{Y}1=\frac{\text{Ynet}1+\text{Ysvm}1+\text{Yreg1}}{\mathrm{3rd}},$

and the YI mutual model of the rock machine is obtained. An intelligent decision making process for drilling control parameters from hard rock TBM Claim 3 , characterized in that the method for obtaining the model for decision making in intelligent control is as follows: Step 1: Use the database of the surrounding rock state in the engineering database to obtain the parameter matrix of the surrounding rock state N = [Rc, Jv, W] to be calculated. Please extract the last 90% of the drilling parameter data useful for the drilling cycle in the drilling parameter database and calculate the average value of the drilling parameter of the drilling cycle in order to calculate the drilling parameter matrix M2 = [ Ft , Tn , P , n , V ] of the stable section. Step 2: Create a three-layer neural network, take the parameter matrix N of the surrounding rock state as input, the drilling parameter matrix M2 of the stable section as output and examine the initial neural network under supervision. Please select 70% of the data in the drilling parameter database for training and 30% of the data for testing in order to obtain a mature neural network Net2 and to obtain the prediction result Ynet2 of the parameters of the surrounding rock state. Step 3: The support vector machine is used to take the parameter matrix N of the surrounding rock condition as input, and the drilling parameter matrix M2 of the stable section is taken as output and perform a data regression. Please select 70% of the data in the drilling parameter database for training and 30% of the data for tests in order to obtain a mature regression learning machine svm2 and to obtain the prediction result Ysvm2 of the parameters of the surrounding rock condition. Step 4: Using the least squares regression method, the parameter matrix N of the surrounding rock state is used as input and the drilling parameter matrix M2 of the stable section is used as output to obtain the mathematical model as follows: $$\text{Ft}=464.853-\mathrm{9,012}\mathrm{\ast }\text{Jv}-80.4\mathrm{\ast }\text{W}+0.118\mathrm{\ast }\text{Rc}$$

$$\text{Tn}=\mathrm{61,865}-\mathrm{7,012}\mathrm{\ast }\text{Jv}-7.4\mathrm{\ast }\text{W}+0.064\mathrm{\ast }\text{Rc}$$

$$\text{P}=0.1105+0.12\mathrm{\ast }\text{Jv}\mathrm{+2}\mathrm{.446\; \ast }\text{W}+0.058\mathrm{\ast }\text{Rc}$$

$$\text{n}=\mathrm{11,743}+0.16\mathrm{\ast }\text{Jv}-\mathrm{1,132}\mathrm{\ast }\text{W}\mathrm{-}0.044\mathrm{\ast }\text{Rc}$$

$$\text{V}=\text{p}\mathrm{\ast }\text{n}$$

The prediction result Yreg2 of drilling state parameters of the stable section is obtained; Step 5: According to the prediction results Ynet2, Ysvm2 and Yreg2 of the drilling state parameters of the stable section obtained in Step 2-4, the mathematical average is calculated: $\text{Y2}=\frac{\text{Ynet2}+\text{Ysvm2}+\text{Yreg2}}{\mathrm{3rd}},$

and the Y2 model for intelligent control decision making is obtained. An intelligent decision making process for drilling control parameters from hard rock TBM Claim 1 , characterized in that the currently estimated parameters are as follows: the average value of the parameter Nk of the surrounding rock and the stable drilling parameter Mk, which is predicted by the current kth group of drilling parameters. The cumulative average estimated parameters are as follows: the parameters of the surrounding rock condition are predicted according to the model Y1 for mutual feedback of the rock machine, and the parameters Ns of the surrounding one Rocks predicted from sections K-2 through K-1 of the drilling parameters are obtained. If the average deviation between the current estimated parameters and the cumulative average estimated parameters is less than 10%, the surrounding rock is in a stable section and the impact of the current drilling parameters is stable. If the average deviation between the estimated parameters and the cumulative average estimated parameters is greater than 90%, the current drilling parameters are unstable and must be set. An intelligent decision making process for drilling control parameters from hard rock TBM Claim 1 or 6 , characterized in that the manual setting in step 4 is that the main leader of the TBM sets the current drilling parameters such as the cutter head speed n and the advance speed V in real time according to the difference between the cumulative average estimated parameters and the current estimated parameters and other drilling parameters controls other drilling parameters within the stable range and ensures the safe and efficient drilling of TBM. The automatic setting in step 4 consists of: According to the difference between the pre-accumulated average estimated parameters and the current estimated parameters, the current drilling parameters such as the cutter head speed n and the advance speed V are set in real time by the PLC control and other drilling parameters are set within the stable one Area controlled to ensure the safe and efficient drilling of the TBM. An intelligent decision making process for drilling control parameters from hard rock TBM Claim 1 , characterized in that the self-learning and self-updating of the model for mutual feedback of the rock machine and the model for decision-making with intelligent control in TBM are to be carried out manually or by host computers during the downtime or the casting time if the increase in data in the engineering Database is more than 30% to get a new model for mutual feedback of the rock machine and a new model for decision making with intelligent control. After the update, the current model for mutual feedback of the rock machine and the model for decision-making with intelligent control dominate. If the prediction result of the new rock machine mutual feedback model is better than that of the current rock machine mutual feedback model or that of the intelligent control decision making model, the new rock machine mutual feedback model replaces the current mutual feedback model of the rock machine, and if the prediction result of the new intelligent control decision making model is better than that of the current intelligent control decision making model, the new intelligent control decision making model replaces the current intelligent control decision making model. An intelligent hard rock TBM drilling control parameter decision system, characterized in that it includes: The engineering database unit that periodically retrieves the surrounding rock condition data and drilling parameter data from the TBM host computer. The rock machine mutual feedback model unit that uses the drilling parameters of the engineering database unit to predict the parameters of the surrounding rock condition of the current drilling environment. The model unit for intelligent control decision making that estimates the best drilling control parameters according to the parameters of the surrounding rock condition. The real-time display module for the output of model parameters communicates with the host computer via the I / O interface and displays the output parameters on the main operator's user interface. The control module for automatic / manual drilling parameters is located on the user interface of the host computer. The main operator or the system controls the drilling parameters of the TBM via the PLC control according to the estimated drilling parameters. The self-learning and self-updating model unit is located on the user interface of the host computer. The main guide sets the model update time or the background program selects the appropriate time for the automatic update of the model. The appropriate time includes, but is not limited to, the downtime and the casting time. An intelligent decision making system for drilling control parameters from hard rock TBM Claim 9 , characterized in that the parameters of the model unit for mutual feedback of the rock machine and the model unit for decision making with intelligent control can be read and written. When the self-learning and self-updating model unit is started, the parameters of the model unit for mutual feedback of the rock machine and the model unit for decision making with intelligent control can be written, and other processes can only be read and not written. An intelligent decision making system for drilling control parameters from hard rock TBM Claim 9 , characterized in that the data obtained from the engineering database unit from the TBM host computer includes the real-time TBM drilling parameters and other TBM engineering databases. The engineering database of other TBMs includes, but is not limited to, the engineering database of TBMs of the same structure type or a similar geological excavation.

Images