JP6584601B2 - Excavator processing apparatus and work content determination method - Google Patents

Description

  The present invention relates to a shovel processing apparatus and a work content determination method.

  Various sensors for measuring operating variables such as engine speed and hydraulic pressure are attached to work machines such as excavators and lifting magnet type work machines. A technique for diagnosing an abnormality of a work machine based on detection values of these sensors is known. In order to perform an abnormality diagnosis, it may be necessary to specify what kind of work the detection value of the sensor is obtained at.

  Patent Document 1 below discloses a technique for determining the work content of a hydraulic excavator based on a predetermined feature amount indicating the operating state of the hydraulic excavator. Predetermined feature quantities indicating the operating state include boom operation complexity display amount, bucket operation complexity display amount, high-speed turning time, boom reverse operation time, bucket arm stop time, boom operation amount average value, arm operation amount average Value, bucket operation amount average value, turning operation amount average value, right side travel operation amount average value, and left side travel operation amount average value are used.

Japanese Patent Laid-Open No. 10-60948

  With the method that uses the average value of various driving variables as a feature value and discriminates the work content based on this feature value, when there are multiple work contents that show similar tendencies in the average value of the driving variables, It is difficult to make a judgment. An object of the present invention is to provide a shovel processing apparatus and a work content determination method capable of more accurately determining work content.

According to one aspect of the invention,
A shovel processing device that is connected to an input device to which a time change of data obtained by measuring a plurality of data over a certain period is input, and that determines work contents by the shovel,
The processor is
Based on the data divided into a plurality of sections on the time axis, determine the feature amount of the data for each section,
A state for each section is obtained based on the determined feature amount,
There is provided a shovel processing device for calculating a matching probability corresponding to each work content based on the obtained state transition for each section and determining the work content corresponding to the transition of the shovel.

According to another aspect of the invention,
An excavator work content determination method for determining work content by an excavator based on time variation of data obtained by measuring a plurality of data over a certain period,
Based on the data divided into a plurality of sections on the time axis, determine the feature amount of the data for each section,
A state for each section is obtained based on the determined feature amount,
Calculating a matching probability corresponding to each work content on the basis of the transition of the state of each of the obtained interval of the shovel, determine the constant how to determine the work corresponding to the transition is provided.

  By determining the work content based on the state transition for each section obtained based on the feature amount, more accurate determination can be performed as compared with the case of determining based on the feature amount at a certain time. .

FIG. 1 is a block diagram of a work machine management device according to an embodiment and a schematic diagram of a work machine to be managed. FIG. 2 is a diagram illustrating a state transition model stored in the work machine management apparatus according to the embodiment. FIG. 3 is a flowchart of a method for determining a state transition model. FIG. 4 is a graph showing an example of the time change of the operation variable of the work content (simple excavation work) and the reference data determined from the time change of the operation variable. FIG. 5 is a diagram illustrating a plurality of state sequences generated with respect to one reference data, and generation probabilities calculated for each state sequence. FIG. 6 is a flowchart of a method for determining work content corresponding to a time change of an operation variable whose work content is unknown. FIG. 7 is a detailed flowchart of step SB3 in FIG. FIG. 8 is a graph showing an example of a time change of an operation variable whose work content is unknown, and a diagram showing verification data determined from the time change of the operation variable. FIG. 9 is a diagram showing a state series generated corresponding to the verification data, a generation probability calculated for each state series, and a matching probability. 10A and 10B are diagrams showing the contents displayed on the display device of the input / output device.

  FIG. 1 shows a block diagram of a work machine management apparatus 10 according to an embodiment and a schematic diagram of a work machine 18 to be managed. The work machine management device 10 includes an input / output device 11, a processing device 12, and a storage device 13. The input / output device 11 includes a communication device, an image display device, a keyboard, a pointing device, a USB port, a removable memory slot, and the like. The input / output device 11 transmits and receives various data to and from the work machine 18 via the communication line 19. In FIG. 1, an excavator is shown as the work machine 18 to be managed, but other work machines may be set as management objects.

  A plurality of sensors are mounted on the work machine 18. The sensor measures a plurality of operating variables depending on the operating state of the work machine 18. The operating variables include, for example, engine speed, hydraulic pump pressure, operating pressure for controlling excavator forward, backward, turning, hydraulic cylinder operating pressure for controlling booms, and the like. The measured values of these operating variables are transmitted to the work machine management device 10.

  The processing device 12 executes the program stored in the storage device 13 to determine the work content when the operation variable is detected based on the time change of the operation variable received from the work machine 18. The storage device 13 stores a program executed by the processing device 12 and various data used in processing for determining work contents.

  With reference to FIG. 2, a state transition model used in the work content determination process will be described. A state transition model STM is defined for each work content W. The state transition model STM includes a state transition path that defines a path of transition of the state S, a transition probability t between states, and an appearance probability e of an event that appears for each state S. The plurality of states S are arranged in time series, and transition from one state to the next state with the passage of time. There is no return from one state to the previous state. Furthermore, the transition destination state is self or another state and does not branch from one state to a plurality of next states. This state transition model STM is stored in the storage device 13 (FIG. 1).

  Each state S corresponds to one operation of the work machine. An event that appears for each state S is represented by a value taken by the feature amount E in this embodiment. As an example, the events that appear for each state S include an event in which the feature quantity E takes a value “a”, an event that takes a value “b”, and an event that takes a value “c”. The feature amount E is determined based on the value of the driving variable. Note that the number of events is not limited to three. It is also possible to define the state transition model STM so that four or more events appear for each state S.

  The work content W includes, for example, simple excavation work (work content W1), turning ground leveling work (work content W2), loading work (work content W3), and the like. The state transition model STM1 of simple excavation work (work content W1) includes six states from state S11 to state S16. For example, states S11 to S16 correspond to a boom raising operation, an arm extending operation, a bucket releasing operation, a boom lowering operation, an arm bending operation, and a bucket excavating operation, respectively.

Transitions between states consist of self-transitions and transitions to the next state. The self-transition probability of the state Si is represented by t (ii), and the transition probability from the state Si to the next state Sj is represented by t (ij). If a random variable at a certain time n is Xn, t (ii) is expressed by a conditional probability that when the random variable Xn is in the state Si, Xn + 1 is also in the state Si. That is, t (ii) is
t (ii) = P (Xn + 1 = Si | Xn = Si)
It is expressed. Further, the self-transition probability of the state S11 is expressed as t (11 11), and the transition probability from the state S11 to the next state S12 is expressed as t (11 12).

  In the state Si, the appearance probability of an event in which the feature quantity E has the value “k” is expressed as ei (k). For example, in the state S11, the event appearance probabilities that the feature quantity E has the values “a”, “b”, and “c” are expressed as e11 (a), e11 (b), and e11 (c), respectively. .

  FIG. 3 shows a flowchart of a method for determining the state transition model STM. By executing the process shown in FIG. 3 for each work content W, the state transition model STM of each work content W is determined. Each state S of the state transition model STM and a route through which the state transitions are defined in advance according to the work content W. When the processing device 12 (FIG. 1) executes the process shown in FIG. 3, the transition probability t and the appearance probability e of the state transition model are determined.

  In step SA1, the time change of the operation variable V whose work content W is known is acquired. The time change of the operation variable V is obtained by measuring a plurality of operation variables V of a work machine that is performing work corresponding to a certain work content W over a certain period. A plurality of time variations of the operation variable V may be acquired from different aircrafts of the same model, and a plurality of operation variables V are measured by measuring the operation variable V at a plurality of times with different operation dates and times on the same aircraft. You may acquire the time change of.

  In step SA2, the time change of the plurality of operating variables V is divided into a plurality of sections on the time axis.

  FIG. 4 shows an example of changes over time in the operation variables V1, V2, V3,... Of the work content W1 (simple excavation work). The time change of the operation variables V1, V2, V3,... Is divided into a plurality of sections D1 to D10 on the time axis. The length of time of each of the sections D1 to D10 is, for example, 1 second. What is necessary is just to set suitably the length of each time of the divisions D1-D10. If the length of one section is shortened, the determination accuracy of the work content can be improved, but the processing time for determining the work content becomes long. The length of one section may be set according to the required determination accuracy and the processing capability of the processing device 12 (FIG. 1). When the length of one section changes, the number of sections obtained by partitioning also changes. In the embodiment, a case where the section is divided into ten sections D1 to D10 will be described.

  In step SA3 (FIG. 3), the feature amount E is determined for each of the sections D1 to D10 based on the values of the operation variables V1, V2, V3,. The reference data R is generated by arranging the determined feature value E in time series. In the example shown in FIG. 4, the values of the feature amount E are “a”, “b”, “c”, “a”, “a”, “c”, “b” in the sections D1 to D10, respectively. , “B”, “a”, “a”. In this case, the generated reference data R is “abcaacbbaa”.

  Since a plurality of time changes are prepared for each of the operation variables V1, V2, V3,..., A plurality of reference data R is generated. The length of time of a plurality of time changes is not necessarily the same. When the length of the time change is different, the number of sections obtained by dividing the time change is different. For this reason, the number of feature quantities E constituting the reference data R is not necessarily the same for all reference data R.

  In step SA4 (FIG. 3), provisional values of the transition probability t and the appearance probability e of the state transition model STM are set. Appropriate values may be set as provisional values for the transition probability t and the appearance probability e.

  In step SA5 (FIG. 3), a plurality of state series K is generated by applying each state S of the state transition model STM to each of the plurality of feature amounts E constituting the reference data R.

  FIG. 5 shows an example of a plurality of state series K (1) to K (N) generated for one reference data R. The number of feature amounts E constituting the reference data R is ten, and the number of states S11 to S16 constituting the state transition model STM is six. For this reason, the same state S is assigned to a plurality of feature amounts E. In the state series K (i) shown in FIG. 5, the same state S11 is assigned to the feature quantity E that takes the first value “a” and the feature quantity E that takes the second value “b”. Yes.

  FIG. 5 shows the state series K (1) to K (N) applied to one reference data R. Actually, however, the state series K is similarly generated for all the reference data R. The

  In step SA6 (FIG. 3), the generation probability PK (i) of the state sequence K (i) is calculated for each state sequence K (i) using the provisional values of the transition probability t and the appearance probability e. For example, in the state S11 assigned to the first feature quantity E that takes the value “a” of the state series K (1) shown in FIG. And since the transition destination state is S11 (self-transition), this probability is e11 (a) × t (1111). The appearance probability e11 (a) and the transition probability t (1111) are as defined in FIG. Similar probabilities can be obtained for the ten states S constituting the state sequence K (1). By multiplying all the probabilities obtained for these ten states S, the generation probability PK (1) of the state series K (1) is calculated. The generation probability PK (i) of the state series K (i) shown in FIG. 5 is expressed by the following equation.

PK (i) = e11 (a) × t (11 11)
× e11 (b) × t (11 12)
× e12 (c) × t (12 12)
× e12 (a) × t (12 13)
× e13 (a) × t (13 13)
× e13 (c) × t (13 14)
× e14 (b) × t (14 14)
× e14 (b) × t (14 15)
× e15 (a) × t (15 16)
× e16 (a) × t (16 end)

  In step SA7 (FIG. 3), the transition probability t and the appearance probability e are calculated based on the generation probability PK of the state series K, and the provisional values are replaced with newly calculated values. Specifically, it appears in all state sequences K (1) to K (N) shown in FIG. 5 and all state sequences K applied to other reference data R not shown in FIG. For all the states S11, the number of appearances of the event in which the feature amount E has the values “a”, “b”, and “c” is weighted by the generation probability PK and totaled. Based on the total number of times, the appearance probabilities e11 (a), e11 (b), and e11 (c) can be calculated. Similarly, all appearance probabilities e can be calculated.

  Furthermore, all the state sequences K (1) to K (N) shown in FIG. 5 and all the state sequences K applied to other reference data R not shown in FIG. Regarding the state S11, the number of times of self-transition and the number of times of transition to the next state S12 are respectively weighted by the generation probability PK and totaled. Based on the total number of times, the transition probabilities t (11 11) and t (11 12) can be calculated. Similarly, all transition probabilities t can be calculated.

  In step SA8 (FIG. 3), it is determined whether or not the generation probabilities PK of the state series K have converged. Specifically, when the difference between the value of the generation probability PK before calculating the generation probability PK in step SA6 and the value of the generation probability PK after calculation is smaller than the convergence determination value, the generation probability PK is converged. It is determined that If it is determined that the generation probability PK has not converged, the process returns to step SA6, and the generation probability PK is calculated using the new provisional value.

  If it is determined that the generation probability PK has converged, the nearest provisional values of the transition probability t and the appearance probability e are set as definite values in step SA9. This determined value is stored in the storage device 13 (FIG. 1) as a part of parameters defining the state transition model.

  FIG. 6 shows a flowchart of a method for determining the work contents of the time change of the operation variable V whose work contents W are unknown. The processing of the flowchart is executed by the processing device 12 (FIG. 1).

  In step SB1, the time change of the operation variable V whose work content W is unknown is input to the processing device 12 (FIG. 1) via the input / output device 11 (FIG. 1). FIG. 8 shows an example of temporal changes in the operation variables V1, V2, and V3 whose work contents W are unknown. The operation variables V1, V2, and V3 are the same as the operation variables V1, V2, and V3 (FIG. 4) used when defining the state transition model STM.

  In step SB2, verification data T including a time series of the feature amount E is generated based on the time change of the input operation variable V. Hereinafter, a method for generating the verification data T will be described.

  In step SB21, the time change of the operating variable V is divided into a plurality of sections on the time axis. In the example shown in FIG. 8, the time change of the operation variables V1, V2, and V3 is divided into eight sections D1 to D8. The time length of each of the sections D1 to D8 is the same as the time length of the sections divided in step SA2 (FIG. 3) when the state transition model STM is defined.

  In step SB22 (FIG. 6), the feature amount E is determined for each of the sections D1 to D8 based on the values of the operation variables V1, V2, and V3. Verification data T in which a plurality of feature amounts E are arranged in time series is generated. In the example illustrated in FIG. 8, the verification data T is “abbacbac”.

  In step SB3 (FIG. 6), a matching probability Pc between the verification data T and the state transition model STM is calculated.

  FIG. 7 shows a detailed flowchart of step SB3. In step SB3, for all work contents W1, W2, W3,..., The processes of steps SB31 to SB34 are repeatedly executed for each work content.

  In step SB31, a plurality of state series L are generated from the verification data T. The process of step SB31 is the same as the process of generating the state series K in step SA5 (FIG. 3) of the process of defining the state transition model STM. FIG. 9 shows the generated state sequences L (1) to L (M). Since the feature quantity E constituting the verification data T is eight, each of the state series L includes eight states S arranged in time series.

  In step SB32 (FIG. 7), the generation probabilities PL (1) to PL (M) (FIG. 9) of the state series L (1) to L (M) are calculated. The method for calculating the generation probabilities PL (1) to PL (M) is the same as the method for calculating the generation probability PK in step SA6 (FIG. 3) of the process for defining the state transition model STM. Note that, when the generation probability PL is calculated, the definite values are used as the transition probability t and the appearance probability e.

  In step SB33 (FIG. 7), the generation probabilities PL (1) to PL (M) are added to the generation probabilities PL (1) to PL (M) for all the state sequences L (1) to L (M). The total value of is calculated. In step SB34, the total value is adopted as the matching probability Pc between the verification data T and the state transition model STM. By executing steps SB31 to SB34 for each work content W1, W2, W3,..., The match probabilities Pc1, Pc2, Pc3,. 9) is calculated.

  In step SB4 (FIG. 6), each of the match probabilities Pc1, Pc2, Pc3,... Is compared with a threshold value PcT. In step SB5, the work content is determined based on the comparison result in step SB4. Specifically, the work content with the matching probability Pc equal to or higher than the threshold value PcT is extracted as a work content candidate when the verification data T is collected.

  In step SB6 (FIG. 6), the determination result of the work content is output to the input / output device 11 (FIG. 1). For example, the determination result of the work content is displayed on the image display device of the input / output device 11.

  FIG. 10A shows an example of the contents displayed on the display device of the input / output device 11. When there is one candidate extracted in step SB5, the work content extracted as a candidate is output to the input / output device 11 together with the verification data and the state transition path.

  FIG. 10B shows a display example when a plurality of candidates are extracted in step SB5. A plurality of work contents extracted as candidates are output to the input / output device 11 together with a match probability and a path of state transition. The operator can estimate the work content corresponding to the verification data T based on the displayed matching probability and the state transition path.

  In the above-described embodiment, the work content is determined based on a change over time (that is, a time series of feature values) instead of an average value of various operation variables acquired from the work machine over a certain period. For this reason, the determination accuracy can be increased.

  Although the present invention has been described with reference to the embodiments, the present invention is not limited thereto. It will be apparent to those skilled in the art that various modifications, improvements, combinations, and the like can be made.

DESCRIPTION OF SYMBOLS 10 Work machine management device 11 Input / output device 12 Processing device 13 Storage device 18 Work machine 19 Communication line E Feature quantity K, L State series S State STM State transition model Pc Match probability PcT Threshold t Transition probability e Appearance probability PK, PL Generation probability R Reference data T Verification data W Work contents V Operating variable

Claims (8)

A shovel processing device that is connected to an input device to which a time change of data obtained by measuring a plurality of data over a certain period is input, and that determines work contents by the shovel,
The processor is
Based on the data divided into a plurality of sections on the time axis, determine the feature amount of the data for each section,
A state for each section is obtained based on the determined feature amount,
A shovel processing device that calculates a matching probability corresponding to each work content based on the obtained state transition for each section, and determines the work content corresponding to the transition of the shovel. A shovel processing device that is connected to an input device to which a time change of data obtained by measuring a plurality of data over a certain period is input, and that determines the operation of the shovel,
The processor is
Based on the data divided into a plurality of sections on the time axis, determine the feature amount of the data for each section,
A shovel processing device that determines a probability of a state corresponding to a motion based on the determined feature amount and determines a shovel motion.   The shovel processing device according to claim 1, wherein the state includes at least a boom raising operation and an arm extending operation.   The excavator processing device according to claim 1, wherein the work content includes at least excavation work and loading work. An excavator work content determination method for determining work content by an excavator based on time variation of data obtained by measuring a plurality of data over a certain period,
Based on the data divided into a plurality of sections on the time axis, determine the feature amount of the data for each section,
A state for each section is obtained based on the determined feature amount,
On the basis of the transition of the obtained in each section condition calculates the matching probability corresponding to each work and the shovel, determine the constant how to determine the work corresponding to the transition. A method for determining the operation of a shovel that determines the operation of the shovel based on a time change of data obtained by measuring a plurality of data over a period of time,
Based on the data divided into a plurality of sections on the time axis, determine the feature amount of the data for each section,
A determination method for determining a shovel operation by obtaining a probability for a state corresponding to an operation based on the determined feature amount. The state, determine the constant A method according to claim 5 or 6 comprising at least boom raising operation and arm stretching operation. The determination method according to claim 5 , wherein the work content includes at least excavation work and loading work.

Images