CN115302507A - Intelligent decision-making method for disassembly process of industrial robot driven by digital twin - Google Patents
Intelligent decision-making method for disassembly process of industrial robot driven by digital twin Download PDFInfo
- Publication number
- CN115302507A CN115302507A CN202210954612.9A CN202210954612A CN115302507A CN 115302507 A CN115302507 A CN 115302507A CN 202210954612 A CN202210954612 A CN 202210954612A CN 115302507 A CN115302507 A CN 115302507A
- Authority
- CN
- China
- Prior art keywords
- industrial robot
- disassembly
- disassembling
- product
- behavior
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 179
- 230000008569 process Effects 0.000 title claims abstract description 121
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 15
- 230000003993 interaction Effects 0.000 claims abstract description 15
- 230000002457 bidirectional effect Effects 0.000 claims abstract description 7
- 238000013507 mapping Methods 0.000 claims abstract description 7
- 230000006399 behavior Effects 0.000 claims description 65
- 239000011159 matrix material Substances 0.000 claims description 17
- 230000009471 action Effects 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 10
- 230000005540 biological transmission Effects 0.000 claims description 5
- 238000012544 monitoring process Methods 0.000 claims description 4
- 238000003491 array Methods 0.000 claims description 3
- 230000002787 reinforcement Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000012800 visualization Methods 0.000 claims description 2
- 239000002699 waste material Substances 0.000 description 15
- 238000005516 engineering process Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 238000004088 simulation Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 102100031102 C-C motif chemokine 4 Human genes 0.000 description 1
- 101100054773 Caenorhabditis elegans act-2 gene Proteins 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1679—Programme controls characterised by the tasks executed
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1602—Programme controls characterised by the control system, structure, architecture
- B25J9/1605—Simulation of manipulator lay-out, design, modelling of manipulator
Landscapes
- Engineering & Computer Science (AREA)
- Robotics (AREA)
- Mechanical Engineering (AREA)
- Automation & Control Theory (AREA)
- Numerical Control (AREA)
Abstract
The invention discloses an intelligent decision method for a disassembly process of an industrial robot driven by a digital twin, which comprises the following steps: considering the uncertainty of the part missing state, accurately describing the part disassembling constraint relation of the product under the part missing state; constructing a digital twin model for describing geometrical, behavior and time sequence characteristics of the physical industrial robot in a disassembling process at high fidelity; the characteristics of digital twin virtual-real connection interaction are fused, and the bidirectional mapping of a physical disassembly process and a digital twin model is realized through a data interaction mode based on 'state-decision'; and (4) considering unpredictability of the uncertain missing state of the part, constructing an intelligent decision model of the industrial robot in the disassembling process through a depth Q network algorithm, and dynamically deciding to generate an optimal disassembling scheme of the product part under the uncertain condition of the missing state of the part. The method can dynamically decide to generate an optimal product disassembly scheme in the industrial robot disassembly process under the uncertain condition of the part missing state.
Description
Technical Field
The invention belongs to the field of cross research combining intelligent manufacturing and artificial intelligence, and particularly relates to an intelligent decision-making method for a disassembly process of an industrial robot driven by a digital twin.
Background
In recent years, the quantity of waste electric and electronic products is huge, and according to the measured data of related mechanisms, the theoretical rejection quantity of the electric and electronic products is about 7.16 hundred million in 2020, the treatment quantity of the waste electric and electronic products is about 8300 ten thousand, and the percentage is about 11.6%. If the waste products are not properly disposed, serious environmental damage can be caused, and meanwhile, the waste products also have huge potential value. Disassembling is a key step in recycling of waste products, the traditional method for disassembling the waste products mainly depends on manual disassembling, the problem of low disassembling efficiency is solved, and the use of an industrial robot to replace manual disassembling is an important means for improving the disassembling efficiency of the waste products. The intelligent decision of the disassembling process is to utilize a disassembling process model to realize the planning of the disassembling process of the industrial robot, and the intelligent decision of the disassembling process is to improve the disassembling efficiency of waste products. Uncertain factors in the process of disassembling the waste products are not negligible. Uncertain factors of the waste product disassembly process are numerous and the formation mechanism is abnormally complex, and part loss is the most common uncertain factor of the waste product disassembly process. However, the disassembling scheme obtained by the existing intelligent decision method for the disassembling process of the industrial robot is difficult to be applied to the disassembling process of the industrial robot under the condition of uncertain part missing states. Therefore, an intelligent decision method for the industrial robot dismantling process under uncertain factors in the product dismantling process is urgently needed to be explored.
Disclosure of Invention
The invention aims to provide an intelligent decision method for a disassembly process of an industrial robot driven by a digital twin, and solves the problem that the disassembly decision method caused by uncertainty of a part missing state and complexity of the disassembly process of the industrial robot is difficult to be applied to the disassembly process of the industrial robot in the part missing state.
In order to achieve the purpose, the invention provides an intelligent decision method for a disassembly process of an industrial robot driven by a digital twin, which is used for dynamically deciding to generate an optimal disassembly scheme of a product part in the disassembly process of the industrial robot under the condition of uncertain part missing states and comprises the following steps:
(1) Considering the uncertainty of the part missing state, and accurately describing the product disassembly constraint relation under the part missing state;
(2) Constructing a digital twin model for describing the geometrical, behavior, time sequence and other characteristics of the physical industrial robot in a high-fidelity manner;
(3) The characteristics of digital twin virtual-real connection interaction are fused, and bidirectional mapping of the physical process and the digital process of the industrial robot disassembly is realized through a data interaction mode based on 'state-decision';
(4) And (4) considering unpredictability of the missing state of the part, constructing an intelligent decision model of the industrial robot in the disassembling process through a depth Q network algorithm, and dynamically deciding to generate an optimal disassembling scheme of the product part under the uncertain condition of the missing state of the part.
In the step (1), a product disassembly constraint relation under the condition of uncertain part missing state is constructed by the following method:
(1) in the product disassembly constraint matrix, element c ij Representing the spatially constrained relationship of part i and part j. And if the two have a spatial constraint relation in the disassembly direction, setting the element in the corresponding matrix to be 1, and otherwise, setting the element to be 0. Where "1" indicates that part i is subject to disassembly constraints of part j.
(2) And analyzing the influence action relation of the uncertain state of part missing on the product disassembly constraint matrix based on the constructed product disassembly constraint matrix. When the industrial robot identifies the part missing state, the product disassembly constraint model releases the constraint of the missing part on other parts by deleting corresponding rows and columns in the part disassembly constraint matrix. Since the part missing state is recognized, at this time, the number of detachable parts may be increased.
In the step (2), a digital twin model of the industrial robot dismantling process is constructed by the following method, and the characteristics of geometry, behavior, time sequence and the like of the physical industrial robot dismantling process are described in high fidelity:
(1) geometrical characteristics of the disassembly process: and (3) introducing three-dimensional models of the industrial robot, the disassembled product and the disassembled tool, and adjusting the proportion, the spatial relative position and the distance of the three models to ensure that the geometric dimension of the three models is close to a physical entity, so that the geometric dimension of the digital twin model is consistent with the disassembling scene of the physical industrial robot. The method is characterized in that a rotation direction and an angle limiting component are added to an industrial robot virtual model in the digital twin model, so that accurate description of space motion rules of the physical industrial robot in the disassembling process is realized.
(2) The behavior characteristics of the disassembly process are as follows: behavior 1: the industrial robot moves from a reset point (initial point) to a disassembling tool area, a disassembling tool is replaced, and then the industrial robot returns to the reset point; behavior 2: the industrial robot moves to a product part disassembling area to execute a part disassembling action; behavior 3: after the industrial robot moves to a part disassembling point, recognizing that the part is in a missing state, and immediately returning to a reset point; behavior 4: the industrial robot grabs the disassembled part to move to the part area and places the disassembled part; behavior 5: the industrial robot returns to the reset point.
(3) The time sequence characteristics of the disassembly process: considering the characteristics of the time sequence combination of the disassembling behaviors of the industrial robot, the method mainly comprises the following steps: sequence 1: during initial state, industrial robot does not carry the instrument of disassembling, and the product is disassembled in-process part and is not lost, and its chronogenesis is: behavior 1 → behavior 2 → behavior 4 → behavior 5; and (2) in sequence: during initial state, industrial robot does not carry and solves the instrument, and the product is disassembled in-process part disappearance, and its time sequence is: behavior 1 → behavior 2 → behavior 3; sequence 3: during initial state, the instrument is unanimous with the part is disassembled to the instrument of disassembling that industrial robot carried, and the in-process part is disassembled to the product does not lack, and its chronogenesis is: behavior 2 → behavior 4 → behavior 5; and (4) in sequence 4: during initial state, the instrument is unanimous with the part is disassembled to the instrument of disassembling that industrial robot carried, and the in-process part is disassembled to the product lacks, and its time sequence is: behavior 2 → behavior 3; when the initial disassembling tool of the industrial robot is inconsistent with the part disassembling tool, if the part is in a non-missing state, the disassembling time sequence is the same as the time sequence 1; if the part is in the missing state, the disassembly sequence is the same as sequence 2.
In the step (3), the bidirectional mapping of the physical disassembling process of the industrial robot and the digital twin model thereof is realized in a state-decision data interaction mode:
(1) physical deconstruction process to "state" of the digital twin model: the physical disassembling process of the industrial robot sends 'state' data to the digital twin model, wherein the 'state' data consists of three arrays: the 1 st section is a serial number array of disassembled parts; the 2 nd section is a part serial number array currently disassembled by the industrial robot; section 3 is a part missing state array recognized by the industrial robot; and transmitting the state to the digital twin model to realize the transmission from the physical disassembly process to the state of the digital twin model.
(2) Digital twinning model to physical disassembly process "decision": the deep reinforcement learning algorithm is combined with the current uncertain missing state of the product part, the optimal disassembly behavior in the current state is generated through dynamic decision, the disassembly behavior is regarded as the next product part to be disassembled, the next product part is sent to the physical disassembly process of the industrial robot in a character string mode, the physical disassembly process is enabled, and the decision from a digital twin model to the physical disassembly process is realized.
In the step (4), the construction of an intelligent decision model of the industrial robot disassembling process is realized by the following method:
(1) environmental model of deep Q network algorithm: the method mainly comprises a product part disassembly time data set (comprising disassembly time of parts in good state and disassembly time of parts in missing state and the like), a product disassembly information model (comprising product disassembly restriction, parts in missing state, parts to be disassembled, disassembled parts and the like) and a reward model (a difference value between a constant and disassembly time) which are combined with digital twin simulation.
(2) State-cost function approximation: will s t As the input of the neural network, constructing a state-value function approximation model in a mode of cascading an input layer, a full connection layer and an output layer; based on the model, the current part disassembly state s is combined t A corresponding Q value is generated as an input to the Q learning algorithm.
(3) Training of deep Q networksRefining: the total reward value of the product part disassembly scheme is composed of all actions a t Corresponding prize r t Performing summation calculation by using<s t ,a t ,r t ,s t+1 >Training the depth Q network by using a quadruple experience pool, and continuously updating parameters of the depth Q network by using a gradient descent algorithm in combination with an experience playback method strategy based on uniform sampling;
(4) testing and verifying a decision model: assuming that the product parts are in a complete state, generating an optimal disassembly scheme of the product parts in the complete state through a depth Q network, and taking the optimal disassembly scheme as a pre-planning scheme; based on the method, the industrial robot disassembles any waste product, when a part of the product is identified to be in a missing state in the disassembling process, the state is updated and input to the deep Q network, and the decision model dynamically selects the optimal disassembled part in the current state so as to generate an optimal scheme after dynamic decision; and comparing and analyzing the difference of the preplanning scheme and the optimal scheme after dynamic decision in the product disassembly time, and verifying the effectiveness of the proposed decision model.
Compared with the prior art, the invention has the following advantages and beneficial effects:
aiming at the problem that a disassembly decision method caused by uncertainty of a part missing state and complexity of the industrial robot disassembly process is difficult to adapt to the industrial robot disassembly process in the part missing state, the invention provides an intelligent decision method of the industrial robot disassembly process by fully utilizing a digital twin theory and technology, so that the intelligent decision method is flexibly adapted to the industrial robot disassembly process with uncertain part missing state, and an optimal disassembly sequence suitable for a product part under the current condition is dynamically decided and generated.
Drawings
FIG. 1 is a flow chart of an intelligent decision method for a disassembly process of an industrial robot driven by a digital twin;
fig. 2 is a flow chart of the behaviour of an industrial robot dismantling process;
fig. 3 is a timing flowchart of a disassembling process of the industrial robot;
FIG. 4 is a flow chart of an intelligent decision-making method for a dismantling process of an industrial robot based on a deep Q network;
FIG. 5 is a graph of loss function during deep Q-network algorithm training;
fig. 6 is a reward convergence curve during the training process of the deep Q network algorithm.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
Aiming at the problem that a disassembly decision method caused by uncertainty of a part missing state and complexity of a disassembly process of an industrial robot is difficult to be suitable for the disassembly process of the industrial robot under the part missing state, the invention provides an intelligent decision method of the disassembly process of the industrial robot by fully utilizing a digital twin theory and technology, so that the intelligent decision method is flexibly suitable for the disassembly process of the industrial robot with uncertain part missing state, and an optimal disassembly sequence suitable for a product part under the current condition is dynamically decided and generated.
The digital twinning refers to the theory and technology of describing and modeling the geometry, behavior, time sequence and the like of a physical entity object by using a digital technology. The invention fully utilizes the characteristics of digital twinning technology such as multidimensional high-fidelity simulation, virtual-real interaction and the like, provides an intelligent decision method for the disassembly process of the industrial robot driven by digital twinning, and is applicable to the disassembly process of the industrial robot under the condition of uncertain part missing state: the method is characterized by integrating the characteristics of geometry, behavior, time sequence and the like of the industrial robot in the process of disassembling the waste products, and constructing a digital twin model of the industrial robot in a high-fidelity manner in the process of disassembling the waste products; and dynamically deciding to generate an optimal waste product disassembling sequence under an industrial robot disassembling mode by utilizing the characteristic of digital twin virtual-real interaction and through a state-decision data interaction mode. As shown in fig. 1, the intelligent decision-making method for the industrial robot dismantling process mainly comprises the following steps:
(1) Considering the uncertainty of the part missing state, accurately describing the part disassembling constraint relation of the product under the part missing state;
(2) Constructing a digital twin model for describing geometrical, behavior and time sequence characteristics of the physical industrial robot in a disassembling process with high fidelity;
(3) The characteristics of digital twin virtual-real connection interaction are fused, and the bidirectional mapping of a physical dismantling process and a digital twin model is realized through a data interaction mode based on 'state-decision';
(4) And (4) considering unpredictability of the uncertain missing state of the part, constructing an intelligent decision model of the industrial robot in the disassembling process through a depth Q network algorithm, and dynamically deciding to generate an optimal disassembling scheme of the product part under the uncertain condition of the missing state of the part.
The steps are from product constraint modeling to dynamic decision making in a disassembly process in the embodiment of the invention.
In the step (1), the specific steps for constructing the product disassembly constraint relationship are as follows:
(1) constructing a disassembly constraint matrix: analyzing the disassembly constraint relationship between different parts, setting the matrix elements corresponding to the two parts with the disassembly constraint relationship to be 1, otherwise, setting the value to be 0;
(2) generating feasible disassembled parts: summing the disassembly constraint matrix line by line to form a column vector, wherein the part corresponding to the value of 0 is a part which can be disassembled; by the method, a part set which can be disassembled in a given state can be generated, and an action a of a deep Q network model can be selected from the part set t ；
(3) Constraint updating: when the industrial robot identifies that the part is in a missing state or a certain part is disassembled, the disassembling constraint of the missing part or the disassembled part on other parts is released by deleting corresponding rows and columns in the part disassembling constraint matrix.
In step (2), as shown in fig. 2 and 3, considering the different points of the initial states of the first part and the non-first part of the industrial robot to be disassembled, the digital twin model of the industrial robot to be disassembled is described by the following method:
(1) the geometric model of the industrial robot disassembling process is divided into four regions: 1. a product disassembling area 2, an industrial robot area 3, a tool disassembling area 4 and a part disassembling placement area;
(2) four regions based on industrial robot disassembly process geometric model consider dividing the industrial robot disassembly process into five behaviors: behavior 1 (change tools); act 2 (disassemble); action 3 (part missing); action 4 (place part); action 5 (reset). Five actions are performed in the above four regions. For example, when the part seq needs to be disassembled n The different dismantling behaviour is explained as follows: behavior 1 (change tool): the tail end of the industrial robot moves from a reset point to a disassembling tool area and then moves from the disassembling tool area to an adjusting point; behavior 2 (disassemble): moving the tail end of the industrial robot from the adjusting point to the point where the disassembled parts are given in sequence in the disassembled product area; behavior 3 (part missing): the tail end of the industrial robot moves from the point where the disassembled parts are given in the sequence to a reset point; action 4 (place part): the tail end of the industrial robot moves from the point where the disassembled parts are given in the sequence to the disassembled part placing area; behavior 5 (reset): the industrial robot end moves from the disassembled part placing area to the reset point. Because industrial robot has the singularity in the motion process, set up the reset point for industrial robot joint coordinate is the coordinate point of all zero, make the terminal motion of industrial robot to the reset point, help the automatic generation to disassemble part simulation time data set, need not the manual regulation industrial robot gesture.
(3) The time sequence model of the disassembling process of the industrial robot is used for describing the disassembling process of a single part based on the five behaviors, and is used for acquiring a simulated data set of disassembling time of all parts of a product, and the disassembling sequence is assumed to be Seq = [ Seq ] 1 ,seq 2 ,seq 3 ,...,seq n ]Considering the execution time of the complete disassembly sequence, i.e. for the industrial robot for a single part Seq i (i∈[1,n]) The total disassembly time for a given disassembly sequence Seq, in combination with the disassembly process, can be calculated by the following equation:
in the formula, T seq Total dismantling time, T, representing dismantling sequence Seq seqj,Bm Showing disassembled part Seq j Time spent in behavior m, tool seqj Showing disassembled part Seq j When part Seq j-1 Tool for disassembling Tool seqj-1 And part Seq j Tool for disassembling Tool seqj When it is the same dismantling tool, x j-1,j 0, the part Seq is disassembled j Does not include the time T of action 1 (change tool) seqj,B1 On the contrary, x j-1,j 1, part Seq is disassembled j Includes the time T of action 1 (change tool) seqj,B1 。
When part seq is found j In the absence, the part seq is considered to be j Has been disassembled, action 4 (place part) is replaced with action 3 (part missing).
In the step (3), the bidirectional mapping of the physical disassembling process of the industrial robot and the digital twin model is realized in a state-decision data interaction mode:
"State s t "data consists of three arrays: paragraph 1 is an array of 1 × n, describing the parts that have been disassembled, wherein each element has two states of 0 and 1, "0" indicates that the part of the corresponding serial number has not been disassembled, "1" indicates that the part of the corresponding serial number has been disassembled; paragraph 2 is an array of 1 × n describing the part being disassembled, wherein each element has two states, 0 and 1, "0" indicating that the industrial robot is not currently disassembling the part and "1" indicating that the industrial robot is currently disassembling the part; paragraph 3 is an array of 1 × n, describing the missing state of the part, wherein each element has two states, 0 and 1, "0" indicates that the part is in a good state, and "1" indicates that the part is in a missing state; and transmitting the state data to the digital twin model in a Socket communication mode, so that the transmission from the physical disassembly process to the state data of the digital twin model is realized.
The digital twin model sends 'decision' data to the physical disassembly process: the deep Q network obtains the optimal disassembly behavior in the current state (i.e., the optimal part number to be disassembled next step) according to the "state" data of the physical disassembly process, and converts it into a character string form (e.g., if the part 8 is to be disassembled, a character string "disassemblable _8" is generated). Based on this, the data is transmitted to the physical dismantling process in a Socket communication mode, the physical industrial robot control cabinet analyzes the received character string and executes a corresponding dismantling function (for example, if the character string received by the industrial robot is "disassemblable _8", the control cabinet executes the function "Dis _8 ()", "), so that the physical process of the part 8 can be dismantled, and the data transmission from the digital twin model to the physical dismantling process" decision "is realized.
In the step (4), an intelligent decision method for the disassembly process of the industrial robot under the condition of uncertain part missing state is constructed through a deep reinforcement learning algorithm based on a deep Q network, as shown in FIG. 4, the method mainly comprises the following steps:
(1) combining the product part disassembly constraint matrix in the step (1) and the product part disassembly time data set constructed in the step (2) to construct an environment model of a depth Q network algorithm;
(2) combined with the disassembly state s of the current product t A 1 is to t Inputting the data into a deep Q network, wherein the deep Q network can output a disassembly behavior a in the current state t Then a is added t Input to the environment model to obtain the next disassembly state s of the product t+1 Generating a product disassembling scheme in a loop iteration mode;
(3) and generating a product disassembly scheme by using the deep Q network as a generation, accumulating the disassembly state, the disassembly behavior and the reward value of the product, and storing the disassembly state, the disassembly behavior and the reward value as experience in an experience pool. And when the size of the experience pool meets the experience playback condition, randomly extracting samples from the experience pool to update the network model parameters of the deep Q network. Based on the method, a product disassembly scheme is generated by using the updated deep Q network, whether the product disassembly scheme is the optimal product disassembly scheme or not is judged, if yes, the deep Q network at the moment is stored, and if not, the deep Q network is trained continuously until the optimal deep Q network is stored;
(4) monitoring the reward value and loss value curves of each epoch by using a Tensioboard visualization tool while training the deep Q network, and monitoring the training process of the deep Q network according to the curve convergence degree as shown in figures 5 and 6;
(5) and testing the trained deep Q network, comparing and analyzing the difference of the optimal disassembly scheme generated under the condition that all parts of the product are intact with the optimal disassembly scheme generated under the condition that a certain part of the product is lost in disassembly time, and verifying the effectiveness of the provided industrial robot disassembly process decision method.
In summary, the invention provides an intelligent decision method for a disassembly process of an industrial robot driven by a digital twin, which is used for dynamically deciding and generating an optimal disassembly scheme of a product in the disassembly process of the industrial robot under the condition of uncertain part missing states, and the method mainly comprises the following steps: considering the uncertainty of the part missing state, accurately describing the part disassembling constraint relation of the product under the part missing state; constructing a digital twin model for describing geometrical, behavior and time sequence characteristics of the physical industrial robot in a disassembling process at high fidelity; the characteristics of digital twin virtual-real connection interaction are fused, and the bidirectional mapping of a physical process and a digital process is realized through a data interaction mode based on 'state-decision'; and (4) considering unpredictability of the missing state of the part, constructing an intelligent decision model of the industrial robot in the disassembling process through a depth Q network algorithm, and dynamically deciding to generate an optimal disassembling scheme of the product part under the uncertain condition of the missing state of the part. Compared with the prior art, the intelligent decision method for the industrial robot disassembly process is provided by fully utilizing the digital twin theory and technology aiming at the problem that the disassembly decision method caused by uncertainty of the part missing state and complexity of the industrial robot disassembly process is difficult to be suitable for the industrial robot disassembly process in the part missing state, so that the intelligent decision method for the industrial robot disassembly process is flexibly suitable for the industrial robot disassembly process with uncertain part missing state, and the optimal disassembly sequence suitable for the product part under the current condition is dynamically decided and generated.
It should be noted that, according to the implementation requirement, each step/component described in the present application can be divided into more steps/components, and two or more steps/components or partial operations of the steps/components can be combined into new steps/components to achieve the purpose of the present invention.
It will be understood by those skilled in the art that the foregoing is merely a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included within the scope of the present invention.
Claims (7)
1. A digital twin driven industrial robot disassembling process intelligent decision method is characterized by being used for dynamically deciding and generating an optimal product disassembling scheme in the industrial robot disassembling process under the condition that a part missing state is uncertain, and the method comprises the following steps:
(1) Considering the uncertainty of the part missing state, and accurately describing the part disassembling constraint relation of the product under the part missing state;
(2) Constructing a digital twin model for describing geometrical, behavior and time sequence characteristics of the physical industrial robot in a disassembling process at high fidelity;
(3) The characteristics of digital twin virtual-real connection interaction are fused, and the bidirectional mapping of a physical disassembly process and a digital twin model is realized through a data interaction mode based on 'state-decision';
(4) And (4) considering unpredictability of the uncertain missing state of the part, constructing an intelligent decision model of the industrial robot in the disassembling process through a depth Q network algorithm, and dynamically deciding to generate an optimal disassembling scheme of the product part under the uncertain condition of the uncertain missing state of the part.
2. A digital twin driven industrial robot dismantling process intelligent decision making method according to claim 1 and wherein step (1) includes:
(1) constructing and disassembling a constraint matrix: analyzing the disassembly constraint relationship between different parts, and setting matrix elements corresponding to two parts with the disassembly constraint relationship to be 1, or else, setting the matrix elements to be 0; the constraint matrix is disassembled as follows:
element c ij Representing the spatial constraint relationship between the part i and the part j, wherein '1' represents that the part i is subject to the disassembly constraint of the part j;
(2) generating feasible disassembled parts: summing the disassembly constraint matrix line by line to form a column vector, wherein the part corresponding to the value of 0 is a part which can be disassembled, and thus generating a part set which can be disassembled in a given state;
(3) constraint updating: when the industrial robot identifies that the part is in a missing state or a certain part is disassembled, the disassembling constraint of the missing part or the disassembled part on other parts is released by deleting corresponding rows and columns in the part disassembling constraint matrix.
3. The intelligent decision method for the disassembly process of the digital twin driven industrial robot as claimed in claim 2, wherein the step (2) comprises:
(1) geometrical characteristics of the disassembly process: introducing three-dimensional models of an industrial robot, a disassembled product and a disassembling tool, adjusting the proportion, the space relative position and the distance of the three models, and ensuring that the geometric dimension of the digital twin model is consistent with the disassembling scene of the physical industrial robot; therefore, the geometric model of the industrial robot disassembling process is divided into four areas: 1. a product disassembling area 2, an industrial robot area 3, a tool disassembling area 4 and a part disassembling placement area;
(2) the behavior characteristics of the disassembly process are as follows: dividing the industrial robot dismantling process into five behaviors based on four areas of a geometric model of the industrial robot dismantling process;
behavior 1: the industrial robot moves from the reset point to a disassembling tool area, a disassembling tool is replaced, and then the industrial robot returns to the reset point; behavior 2: the industrial robot moves to a product part disassembling area to execute a part disassembling action; behavior 3: after the industrial robot moves to a part disassembling point, recognizing that the part is in a missing state, and immediately returning to a reset point; behavior 4: the industrial robot grabs the disassembled part to move to the part area and places the disassembled part; behavior 5: the industrial robot returns to the reset point;
(3) the time sequence characteristics of the disassembly process: consider characteristics of industrial robot disassembling action time sequence combination, including:
sequence 1: during initial state, industrial robot does not carry the instrument of disassembling, and the product is disassembled in-process part and is not lost, and its chronogenesis is: behavior 1 → behavior 2 → behavior 4 → behavior 5; and (2) time sequence: during initial state, industrial robot does not carry and solves the instrument, and the in-process part is lacked in the product is disassembled, and its time sequence is: behavior 1 → behavior 2 → behavior 3; sequence 3: during initial state, the instrument is disassembled with the part to the instrument unanimously of disassembling that industrial robot carried, and the product is disassembled in-process part and is not lost, and its chronogenesis is: behavior 2 → behavior 4 → behavior 5; and (4) in sequence 4: during initial state, the instrument is unanimous with the part is disassembled to the instrument of disassembling that industrial robot carried, and the in-process part is disassembled to the product lacks, and its time sequence is: behavior 2 → behavior 3; when the initial disassembling tool of the industrial robot is inconsistent with the part disassembling tool, if the part is in a non-missing state, the disassembling time sequence is the same as the time sequence 1; if the part is in a missing state, the disassembly time sequence is the same as the time sequence 2;
and acquiring a disassembly time data set of all parts of the product based on the disassembly process of a single part.
4. The intelligent decision method for the disassembly process of the digital twin driven industrial robot as claimed in claim 3, wherein the reset point is a coordinate point where the coordinates of the joint of the industrial robot are all zero.
5. A digital twin driven industrial robot dismantling process intelligent decision making method according to claim 3 and wherein step (3) includes:
(1) physical deconstruction process to "state" of digital twinning model: sending 'state' data s to digital twin model in physical disassembling process of industrial robot t "status" data s t The device consists of three arrays: paragraph 1 is the part number that has been disassembledThe array is an array of 1 × n and describes the disassembled parts, wherein each element has two states of 0 and 1, 0 represents that the parts with the corresponding serial numbers are not disassembled, and 1 represents that the parts with the corresponding serial numbers are disassembled; paragraph 2 is the serial number array of the part currently being disassembled by the industrial robot, which is an array of 1 × n, describing the part being disassembled, wherein each element has two states of 0 and 1, "0" indicates that the industrial robot is not currently disassembling the part, and "1" indicates that the industrial robot is currently disassembling the part; paragraph 3 is a part missing state array recognized by the industrial robot, which is an array of 1 × n and describes a part missing state, wherein each element has two states of 0 and 1, "0" indicates that the part is in a good state, and "1" indicates that the part is in a missing state;
will be in the above "state s t The transmission to the digital twin model is realized, and the transmission from the physical dismantling process to the state of the digital twin model is realized;
(2) digital twinning model to physical disassembly process "decision": the deep reinforcement learning algorithm is combined with the current uncertain missing state of the product part, the optimal disassembly behavior in the current state is generated through dynamic decision, the disassembly behavior is regarded as the next product part to be disassembled, the next product part is sent to the physical disassembly process of the industrial robot in a character string mode, the physical disassembly process is enabled, and the decision from a digital twin model to the physical disassembly process is realized.
6. The intelligent decision method for the disassembly process of the digital twin driven industrial robot as claimed in claim 5, wherein the step (4) comprises:
(1) combining the product part disassembly constraint matrix in the step (1) and the product part disassembly time data set constructed in the step (2) to construct an environment model of a deep Q network algorithm;
(2) combined with the disassembly state s of the current product t A 1, a t Inputting the data into a deep Q network, and outputting the disassembly behavior a in the current state by the deep Q network t Then a is added t Input to the environment model to obtain the next disassembly state s of the product t+1 By means of loop iterationsGenerating a product disassembly scheme;
(3) a product disassembly scheme generated by the deep Q network is taken as a generation, and all disassembly states, disassembly behaviors and reward values of the product are accumulated and stored in an experience pool as experience; and when the size of the experience pool meets the experience playback condition, randomly extracting samples from the experience pool to update the network model parameters of the deep Q network, and generating a product disassembly scheme by using the updated deep Q network based on the network model parameters.
7. A digital twin driven industrial robot dismantling process intelligent decision making method according to claim 6 and wherein step (4) further includes:
judging whether the product disassembly scheme is the optimal product disassembly scheme, if so, saving the depth Q network at the moment, and if not, continuing training the depth Q network until the optimal depth Q network is saved;
(4) monitoring the reward value and loss value curve of each epoch by using a visualization tool while training the deep Q network, and monitoring the training process of the deep Q network according to the curve convergence degree;
(5) testing the trained deep Q network, comparing and analyzing the difference of the optimal disassembly scheme generated under the condition that all parts of the product are intact and the optimal disassembly scheme generated under the condition that a certain part of the product is lost in disassembly time, and verifying the effectiveness of the provided industrial robot disassembly process decision method; and if the training is effective, the trained deep Q network is an intelligent decision model for the industrial robot disassembling process.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210954612.9A CN115302507A (en) | 2022-08-10 | 2022-08-10 | Intelligent decision-making method for disassembly process of industrial robot driven by digital twin |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210954612.9A CN115302507A (en) | 2022-08-10 | 2022-08-10 | Intelligent decision-making method for disassembly process of industrial robot driven by digital twin |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115302507A true CN115302507A (en) | 2022-11-08 |
Family
ID=83860274
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210954612.9A Pending CN115302507A (en) | 2022-08-10 | 2022-08-10 | Intelligent decision-making method for disassembly process of industrial robot driven by digital twin |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115302507A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150273685A1 (en) * | 2014-04-01 | 2015-10-01 | Bot & Dolly, Llc | Software Interface for Authoring Robotic Manufacturing Process |
| CN105975701A (en) * | 2016-05-10 | 2016-09-28 | 浙江大学 | Parallel scheduling disassembly path forming method based on mixing fuzzy model |
| CN112231966A (en) * | 2020-09-08 | 2021-01-15 | 合肥学院 | Cooperative robot assemblability prediction system and method based on digital twin |
| CN112965374A (en) * | 2021-02-02 | 2021-06-15 | 郑州轻工业大学 | Method for disassembling and scheduling in consideration of random demand and operation time under resource constraint |
| US20220195714A1 (en) * | 2020-12-17 | 2022-06-23 | Zoobox Canada Inc. | Method, assembly and system for assembling and disassembling a shelter |
| CN114815688A (en) * | 2022-04-07 | 2022-07-29 | 铭之城科技(苏州)有限公司 | Planning method for disassembly sequence of robot with uncertain interference |
| CN114800510A (en) * | 2022-05-06 | 2022-07-29 | 武汉理工大学 | Industrial robot disassembly sequence planning method considering product part failure characteristics |
-
2022
- 2022-08-10 CN CN202210954612.9A patent/CN115302507A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150273685A1 (en) * | 2014-04-01 | 2015-10-01 | Bot & Dolly, Llc | Software Interface for Authoring Robotic Manufacturing Process |
| CN105975701A (en) * | 2016-05-10 | 2016-09-28 | 浙江大学 | Parallel scheduling disassembly path forming method based on mixing fuzzy model |
| CN112231966A (en) * | 2020-09-08 | 2021-01-15 | 合肥学院 | Cooperative robot assemblability prediction system and method based on digital twin |
| US20220195714A1 (en) * | 2020-12-17 | 2022-06-23 | Zoobox Canada Inc. | Method, assembly and system for assembling and disassembling a shelter |
| CN112965374A (en) * | 2021-02-02 | 2021-06-15 | 郑州轻工业大学 | Method for disassembling and scheduling in consideration of random demand and operation time under resource constraint |
| CN114815688A (en) * | 2022-04-07 | 2022-07-29 | 铭之城科技(苏州)有限公司 | Planning method for disassembly sequence of robot with uncertain interference |
| CN114800510A (en) * | 2022-05-06 | 2022-07-29 | 武汉理工大学 | Industrial robot disassembly sequence planning method considering product part failure characteristics |
Non-Patent Citations (2)
| Title |
|---|
| 刘志峰;胡迪;高洋;张敬东;: "基于贪婪算法的产品拆卸序列规划", 中国机械工程, no. 18, 25 September 2011 (2011-09-25) * |
| 王攀;程培源;王威;张爽;: "基于拆卸Petri网和混沌粒子群的拆卸序列规划", 机械设计与制造, no. 01, 8 January 2015 (2015-01-08) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Killian et al. | Robust and efficient transfer learning with hidden parameter markov decision processes | |
| Chan et al. | ACTRCE: Augmenting Experience via Teacher's Advice For Multi-Goal Reinforcement Learning | |
| EP3114540A1 (en) | Neural network and method of neural network training | |
| CN109344969B (en) | Neural network system, training method thereof, and computer-readable medium | |
| CN109766995A (en) | The compression method and device of deep neural network | |
| Dhahri et al. | Hierarchical multi-dimensional differential evolution for the design of beta basis function neural network | |
| CN110177122A (en) | A kind of method for establishing model and device identifying network security risk | |
| CN112327745B (en) | PLC program design method based on testable digital twin body | |
| US20070179917A1 (en) | Intelligent design optimization method and system | |
| CN114290339B (en) | Robot realistic migration method based on reinforcement learning and residual modeling | |
| US20070233441A1 (en) | Method, system, and apparatus for virtual modeling of biological tissue with adaptive emergent functionality | |
| CN111461284A (en) | Data discretization method, device, equipment and medium | |
| Lu et al. | ModLaNets: learning generalisable dynamics via modularity and physical inductive bias | |
| CN104915714A (en) | Predication method and device based on echo state network (ESN) | |
| CN115302507A (en) | Intelligent decision-making method for disassembly process of industrial robot driven by digital twin | |
| Ping et al. | A mixed isomorphism approach for kinematic structure enumeration graphs based on intelligent design and manufacturing | |
| Patarnello et al. | Learning capabilities of boolean networks | |
| CN115222773A (en) | Single-point motion learning method and device | |
| CN110866607A (en) | Machine learning-based penetration behavior prediction algorithm | |
| KR20190120729A (en) | System Modeling Method by Machine Learning using Big data | |
| CN117236137B (en) | Winter continuous construction control system for deep tunnel in high and cold area | |
| CN117993280A (en) | Dynamic decision method for industrial robot dismantling process under uncertain undetachable state | |
| CN116985150B (en) | Method and device for planning soft and soft collection track of fruit clusters | |
| Kordík et al. | Modified GMDH method and models quality evaluation by visualization | |
| CN117648585B (en) | Intelligent decision model generalization method and device based on task similarity |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |