Winch, winch system, and state estimation system

Abstract

A hoist having a motor and controlling the motor to raise and lower a hoist load comprises: an operation information acquisition unit for acquiring operation information of the motor; and a state estimation unit that estimates the state of the hoist on the basis of the operation information acquired by the operation information acquisition unit and a state estimation formula learned so as to be able to estimate the state of the hoist. The state estimation formula is composed of a common estimation part irrelevant to the type of the winch and a type inherent estimation part different for each type of the winch.

Description

Technical Field

[0001] The invention relates to a winch, a winch system and a state estimation system. Background Art

[0002] In recent years, with the aging of skilled operators of winches and the shortage of manpower caused by the increase in the number of winches, the number of operators with little experience (unskilled operators) is increasing. Therefore, a winch that can be used by unskilled operators to carry out safe and efficient transportation is required. When using a winch, not only does it assist the operator in handling operations by notifying the operator of various operating conditions such as the quality of the load, but there are also various requirements such as preventing overload, predicting life, diagnosing deterioration conditions, suppressing the swing of the load, and detecting the lift-off of the load from the ground.

[0003] In particular, regarding lifting off the ground, since the lifting point deviates from the center of gravity of the load, there is a possibility of a dangerous state such as a large swing of the load or a break of the lifting rope at the moment of lifting off the ground, so it cannot be ignored in order to ensure the safety of the operator. Therefore, a technology for safely lifting off the ground for unskilled operators is proposed. In this regard, a technology is disclosed for obtaining the current value and the rotation speed of the motor constituting the hoist and using the result of comparing them to detect the imminent lifting off the ground and ensure safety (refer to patent document 1).

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application Publication No. 2012-66893 Summary of the invention

[0007] Problems to be solved by the invention

[0008] According to Patent Document 1, it is possible to detect liftoff using output information from the motor. However, the current value of the motor is affected by multiple elements that constitute the winch, such as the deterioration of the transmission system such as bearings and gears, and changes in the temperature of the motor. Therefore, in order to improve the accuracy of liftoff detection, it is necessary to accurately generate a physical model by considering conditional branches. In addition, in addition to liftoff, when estimating various states of the winch (referred to as "state estimation") such as the mass estimation of the hanging weight and the temperature estimation of the motor are added, it is also necessary to construct a physical model for estimating the object items of each state estimation (referred to as "estimation items") and set parameters. Therefore, in the state estimation using the physical model, the factors of the physical items to be modeled increase, which is common to both improving the estimation accuracy and increasing the estimation items, so the setting of parameters is not easy.

[0009] The present invention aims to estimate the state of a device such as a winch without constructing a physical model.

[0010] Technical solutions to solve problems

[0011] As an example of the present invention, a winch having an electric motor and controlling the electric motor to lift and lower a load comprises:

[0012] an operation information acquisition unit for acquiring operation information of the motor; and

[0013] a state estimation unit that estimates the state of the hoist based on the operation information acquired by the operation information acquisition unit and a state estimation formula learned to be able to estimate the state of the hoist,

[0014] The state estimation formula includes a common estimation part irrespective of the model of the hoisting machine and a model-specific estimation part which is different for each model of the hoisting machine.

[0015] Effects of the Invention

[0016] According to the present invention, the state of a device such as a hoist can be estimated. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a diagram showing an example of the structure of the hoisting machine system of the present invention.

[0018] Figure 2 It is a figure which shows an example of the structure of the hoisting machine of this invention.

[0019] Figure 3 It is a figure which shows an example of the structure of the hoisting machine of this invention.

[0020] Figure 4 It is a figure which shows an example of the structure of the hoisting machine of this invention.

[0021] Figure 5 This is a diagram showing an example of the functional configuration of the hoisting machine system according to the first embodiment.

[0022] Figure 6 This is a diagram showing an example of a state estimation formula of the first embodiment.

[0023] Figure 7 This is a diagram showing an example of the learning / adjustment flow of the state estimation formula of the first embodiment.

[0024] Figure 8 This is a diagram showing an example of a mechanism configuration in learning / adjusting the state estimation formula of the first embodiment.

[0025] Fig. 9 This is a diagram showing an example of the functional configuration of the hoisting machine system when learning / adjustment is performed according to the first embodiment.

[0026] Fig.10 This is a diagram showing an example of the correlation between the state estimation formula output in the second embodiment and the measured value of the actual state.

[0027] Fig.11 This is a diagram showing an example of the configuration of a state estimation unit according to the second embodiment.

[0028] Fig.12 This is a diagram showing an example of the structure of the state estimation unit of Example 3. DETAILED DESCRIPTION

[0029] Hereinafter, examples of the present invention will be described in detail, but the present invention is not limited to the examples.

[0030] Example 1

[0031] (Basic form, model difference correction)

[0032] In this embodiment, a winch having an electric motor and controlling the electric motor to lift and lower a load is used as an example for explanation. For example, the winch has an operation information acquisition unit that acquires operation information of the electric motor and a state estimation unit that estimates the state of the winch from the operation information acquired by the operation information acquisition unit. The state of the winch includes the state of the electric motor constituting the winch, the state of the sling on which the load is mounted, the state of the load, and the like. For example, the state estimation unit estimates one or more states of the winch based on a state estimation formula obtained by learning data that corresponds the operation information to information indicating the state of the winch.

[0033] According to the above structure, by using the state estimation formula obtained based on appropriate learning, various states of the winch can be predicted based on the operation information of the motor even without building a physical model. In addition, according to the above structure, the state of the winch in a complex system in which multiple factors affect each other can be estimated. In addition, according to the above structure, the estimated value of the estimation item having a common physical background and the estimated value of the related estimation item can be calculated at the same time.

[0034] Next, embodiments of the present invention are described based on the accompanying drawings. The following description and the accompanying drawings are examples for illustrating the present invention, and are appropriately omitted and simplified to make the description clear. The present invention can also be implemented in various other ways. Unless otherwise specified, each component can be single or multiple.

[0035] In the following description, the same reference numerals are assigned to the same elements in the drawings, and the description is omitted as appropriate. In addition, when describing elements of the same type without distinguishing them, the common parts (parts other than the sub-numbers) in the reference symbols including the sub-numbers are used, and when describing elements of the same type with distinction, the reference symbols including the sub-numbers are used. For example, in this specification, when describing a winch without distinguishing them in particular, it is recorded as "winch 101", and when describing each winch with distinction, it is recorded as "winch 101-1" and "winch 101-2".

[0036] In addition, the expressions "first", "second", "third", etc. in this specification are added to identify the components and do not necessarily limit the number or order. In addition, the numbers used to identify the components are used for each context, and the numbers used in one context do not necessarily represent the same structure in other contexts. In addition, the components identified by a certain number do not prevent the components identified by other numbers from having the same functions.

[0037] For Example 1, use Figures 1 to 8 Provide explanation. Figure 1 In the figure, 100 represents the hoisting machine system of this embodiment as a whole.

[0038] Figure 1 1 is a diagram showing an example of the configuration of the hoisting machine system 100. The hoisting machine system 100 includes a crane 110 on which a hoisting machine 101 is mounted and an operation terminal 120.

[0039] The crane 110 is composed of a track 111 provided along the two side walls of a house (not shown), a bridge 112 movable on the track 111, and a trolley 113 movable along the bridge 112. Transport motors for driving the bridge 112 and the trolley 113 are connected. A hoist 101 is provided on the trolley 113, and a rope suspended from the hoist 101 is hoisted to a hoisting weight 130.

[0040] In addition, the term "cable" mentioned here is used as a general term for the hoisting equipment used to suspend the weight 130. That is, "cable" includes not only cables but also chains, belts, wires, cables, wires, ropes, etc. In addition, below, the machinery involved in the transportation of the weight 130 is sometimes referred to as "product". The product includes at least the winch 101. In addition, the product may be the winch 110, but depending on the product, it does not include the track 111, the bridge 112, the trolley 113, etc.

[0041] Figure 2 to Figure 41 is a diagram showing an example of the structure of the hoist 101. The hoist 101 is configured to include a drum 210 for winding a rope, a motor 220 as a power source for rotating the drum 210, and a bearing 230 for smoothly transmitting power from the motor 220 to the drum 210. In this embodiment, the motor 220 is assumed to be a three-phase motor, but the effect of this embodiment is not related to the type of the motor 220, so various motors 220 such as a DC motor can be applied. In addition, the power from the motor 220 can also be transmitted via gears, transmission belts, etc.

[0042] The transport motors provided in the bridge 112 and the trolley 113, respectively, and the motor 220 provided in the hoist 101 rotate according to the operation command from the operation terminal 120. Thus, the hoist 101 performs a lifting and lowering action, moves the bridge 112 and the trolley 113 in a specified direction, and performs a series of transport operations such as lifting the hanging weight 130, transporting it to a specified position, and lowering it. In addition, it should be noted that the hanging weight 130 is an object transported by the crane 110, and is not a component of the crane 110.

[0043] Here, a typical state of state estimation is described regarding the state of the hoisting machine 101. When the hoisting machine system 100 is off the ground, the mass of the hanging weight 130 (referred to as "hanging weight mass"), the probability of the cable being tightened or loosened (referred to as "tension state"), and the temperature of the motor 220 (referred to as "motor temperature") are estimated, and the weight swing amount of the hanging weight 130 (referred to as "hanging weight swing amount") is estimated immediately after leaving the ground.

[0044] In addition, since the mass of the hanging weight is strongly related to the load applied to the motor 220, the hoisting machine system 100 can also estimate the state (force, acceleration, etc.) expressed by the relational expression of the second law of the motion equation in addition to or instead of the mass of the hanging weight. That is, the hoisting machine system 100 can also estimate the acceleration of the hanging weight 130, the tension applied to the cable, and the rotational load (load torque) of the motor 220 including the transmission system.

[0045] In this embodiment, the reason for assuming that the state is off the ground is that the motor 220 of the winch 101 is mainly driving the state, so it is not easily affected by other factors when the state estimation is performed. However, the state estimation is not limited to the state of off the ground, and can also be performed continuously during transportation and descent. In addition, off the ground refers to the state of Figure 2 The state where the hanging weight 130 is placed on the ground 240 is shown. Figure 3 The hanging weight 130 is shown to be off the ground 240 .

[0046] Here, use Figure 4, indicating that the positional relationship between the hanging weight 130 and the trolley 113 when off the ground has deviated. When off the ground, it is preferred that the trolley 113 is arranged directly above the center of gravity 410 of the hanging weight 130, but if Figure 4 As shown, there is a case where the trolley 113 is arranged at a position deviated from the center of gravity 410 of the hanging weight 130. In this case, the amount of swing of the hanging weight after leaving the ground may increase.

[0047] Figure 5 1 is a diagram showing an example of a configuration of a control signal calculation processing system in the hoisting machine system 100 (functional configuration of the hoisting machine system 100 ).

[0048] The operation terminal 120 has a status display unit 521 and an operation input unit 522. In addition, the operation terminal 120 has hardware such as a CPU or other computing device 523, a main storage device 524 such as a semiconductor memory, an auxiliary storage device 525 such as a hard disk, a communication device 526, and an input / output device 527 composed of a display and a touch panel. The well-known operations of each hardware are described with appropriate omissions. In addition, the operation terminal 120 can be a teaching pendant, or a terminal such as a smart phone, a tablet terminal, a notebook personal computer, a desktop personal computer, etc. The operation terminal 120 can be provided as a dedicated terminal, or the functions of the operation terminal 120 can be provided as an application and used in the operator's respective terminal.

[0049] The functions of the operation terminal 120 (status display unit 521, operation input unit 522, etc.) can be realized, for example, by the computing device 523 reading the program stored in the auxiliary storage device 525 to the main storage device 524 and executing (software), or can be realized by hardware such as a dedicated circuit, or can be realized by combining software and hardware. In addition, one function of the operation terminal 120 can be divided into multiple functions, and multiple functions can also be combined into one function. In addition, part of the functions of the operation terminal 120 can be set as a separate function, or can be included in other functions. In addition, part of the functions of the operation terminal 120 can be realized by other computers that can communicate with the operation terminal 120.

[0050] The state display unit 521 displays, for example, the estimated value of the hanging mass, the estimated value of the motor temperature, and the estimated value of the swing amount as numerical values ​​on the input-output device 527. In addition, the state display unit 521 displays, for example, the estimated value of the tension state as either "tension" or "relaxation" on the input-output device 527. However, the display of the state display unit 521 may be omitted as needed.

[0051] The operation input unit 522 is an interface for the operator to input operation instructions to the winch 101 by operating the input / output device 527 such as buttons, levers, handles, joysticks, keyboards, and mice. The operation input unit 522 can be an interface (hardware) composed of a display and a touch panel, or it can be an interface (software) implemented on a device such as a smart phone. The operation input unit 522 can input the movement amount, the coordinates of the destination, the name of the place, etc. according to the operation performed by the operator of the input / output device 527. The operation input unit 522 can also be connected to other systems such as the operation management system and the production management system that manage the winch 101 installed in the factory, and input operation instructions based on information from other systems. Furthermore, the operation input unit 522 can have an automatic operation function that automatically inputs pre-arranged operation instructions.

[0052] In addition, in this embodiment, the state display unit 521 and the operation input unit 522 are implemented in the operation terminal 120, but they can also be implemented in the hoisting machine 101. For example, the state display unit 521 can also be provided on a circuit board mounted in the hoisting machine 101.

[0053] Next, the hoisting machine 101 will be described. The hoisting machine 101 includes a motor 220, a motor control unit 511, an operation information acquisition unit 512, and a state estimation unit 513. In addition, the hoisting machine 101 includes hardware such as a CPU or other computing device 514, a main storage device 515 such as a semiconductor memory, an auxiliary storage device 516 such as a hard disk, and a communication device 517.

[0054] The functions of the hoist 101 (motor control unit 511, operation information acquisition unit 512, state estimation unit 513, etc.) can be realized, for example, by the computing device 514 reading the program stored in the auxiliary storage device 516 to the main storage device 515 and executing (software), or can be realized by hardware such as dedicated circuits, or can be realized by combining software and hardware. In addition, one function of the hoist 101 can be divided into multiple functions, and multiple functions can also be combined into one function. In addition, part of the functions of the hoist 101 can be set as a separate function, or can be included in other functions. In addition, part of the functions of the hoist 101 can be realized by other computers that can communicate with the hoist 101.

[0055] In addition, in this embodiment, data is assumed to be transmitted by wireless communication, but the effect is not changed even if communication is performed using a wired cable. For example, an operation command inputted by the operation terminal 120 is transmitted to the hoisting machine 101 via the communication device 526 and the communication device 517 .

[0056] The motor control unit 511 receives the operation command inputted by the operation input unit 522 of the operation terminal 120 as input, generates a motor control command for operating the motor 220 according to the operation command, and outputs the command to the motor 220. In addition, the motor control unit 511 sends part or all of the motor control command outputted to the motor 220 to the operation information acquisition unit 512. However, the motor control unit 511 may send the motor control command to the operation information acquisition unit 512 without passing through the motor 220.

[0057] The operation information acquisition unit 512 acquires the operation information sent to the state estimation unit 513. The operation information is input and output information about the operation of the motor 220 used in the state estimation in the state estimation unit 513, and is information indicating a part or all of the motor control command. For example, the operation information is information such as the voltage value applied to the motor 220, the command frequency of the control unit (inverter) of the motor 220, the motor speed calculated based on the sensor value of the encoder provided for the motor 220, the drive current value input to the motor 220, the excitation current value and / or the torque current value in the vector control of the motor 220, the input torque value to the motor 220, the command speed input to the motor 220, the slip value of the speed of the motor 220 corresponding to the command speed input to the motor 220, the motor temperature, the deterioration degree of the motor, the deterioration degree of the components connected to the transmission system of the motor, and the like.

[0058] The state estimation unit 513 performs state estimation based on the operation information acquired by the operation information acquisition unit 512 using the state estimation formula stored in the auxiliary storage device 516 .

[0059] In this embodiment, in the pre-shipment process of the hoist 101, data consisting of operation information indicating the prescribed operation of the motor 220 and state information indicating the state of the hoist 101 during the prescribed operation (information indicating the mass of the hoisting weight, the tensioning state, the motor temperature, and the amount of the hoisting weight swing) is collected, and a high-precision state estimation formula is learned that takes into account individual differences in motor characteristics. Hereinafter, the data set of the collected operation information and state information is referred to as "learning data". In addition, in the operation process, the hoist 101 of this embodiment uses the operation information and the learned state estimation formula to perform state estimation.

[0060] In addition, the state information of the learning data is a numerical value or classification set for learning the state estimation formula, which can be set manually or automatically. Hereinafter, the numerical value and classification set for learning are referred to as "reference values", and the numerical value and classification estimated using the state estimation formula are referred to as "estimated values" for distinction.

[0061] The state estimation formula is a formula that takes operation information as input and outputs (calculates) the mass of the hoisted load, the tension state, etc. Thus, multiple states (estimation items) of the hoisting machine 101 are estimated for one motor 220. In other words, the multiple estimation items share the operation information as input.

[0062] The operation information used as input may be the current sample (operation information), but in this embodiment, the time series information of a certain number of samples from the past to the present is used. This is because if the past operation information can be used, the expressiveness of the state estimation formula is improved, for example, a state estimation formula that takes into account the hysteresis of the motor 220 can be obtained. However, due to the constraints of the CPU's computing performance, the capacity of the storage device, etc., when the data capacity is limited, it is sufficient to at least obtain the operation information after leaving the ground.

[0063] Figure 6 This is a diagram showing an example (network 600) of a state estimation formula. Figure 6 The network structure of multi-task estimation in network 600 is shown in FIG. More specifically, network 600 is a fully connected neural network in which the number of nodes in input layer 610 is N, the number of shared layers 620 is "2", and the number of intrinsic layers 630 branching for each estimation item is "1". Operation information is provided to input layer 610 to calculate the value of each estimation item. In this case, the state estimation formula between each layer is (Formula 1).

[0064] u=Wx+b…(Formula 1)

[0065] Here, x is the input vector of each layer, and is the vector obtained by connecting the operation information in the input layer 610 (hereinafter referred to as the "operation information vector"). u is a vector representing the output of each layer. In addition, the u vector is a scalar value in the output layer of the hoisting mass, motor temperature, and hoisting swing amount, and is the probability of "tension" or "relaxation" in the output layer of the tension state. In addition, for example, in the case of estimating the hoisting mass, it can be the estimated value itself, or it can be the ratio of the estimated value of the hoisting mass corresponding to the rated load of the product. In the latter case, when u=0.5 in a product with a rated load of 1t, the estimated value of the hoisting mass is 500kg. W is a coefficient matrix representing the weights corresponding to each element of the operation information vector x. b is a vector representing the bias.

[0066] In addition, in the state estimation unit 513, low-pass filtering such as moving average filtering can be used as post-processing to remove time series noise after the state estimation.

[0067] However, when estimating multiple states, part of the calculation of the state estimation formula can be shared. This is because the estimation items are items that can be estimated based on the operating information of the same motor 220, and there is a relationship (such as a dependency) among the estimation items, so it can be realized. Let me explain in more detail. In the calculation of the tensioned state, a method such as outputting a "tensioned" judgment when the specified hoisting mass is exceeded can be adopted, so the hoisting mass and the tensioned state are in a relationship that can be estimated based on the same operating information. In other words, for the hoisting mass and the tensioned state, the physical background required for the calculation is common. Therefore, the state estimation formula can be divided into a shared layer 620 that can be shared in the calculation of multiple estimation items and an inherent layer 630 that is different for each estimation item.

[0068] In addition, when the model is different, the rated load, responsiveness, and characteristics of the transmission components such as the gearbox and bearings used by the motor are different, and the parameters such as the coefficient matrix and bias of each layer of the state estimation formula are different. The state estimation operation based on the operation information of the motor 220 can be divided into an operation at an abstract level that can be used universally in all models, and an operation that corrects the parameters and operation results of the operation according to the model. Therefore, the model difference is taken into account by learning by dividing the state estimation formula into a common estimation unit 640 common to all models and a model-specific estimation unit 650 that is different according to the model.

[0069] Next, the learning data will be described. The conditions for collecting the learning data (referred to as "collection conditions") may include the position deviation of the hanging weight 130 relative to the trolley 113 and the hoisting mode during lifting. In addition, the hoisting mode is a hoisting mode such as a high-speed hoisting mode and a low-speed hoisting mode, a hoisting speed, etc. The collected learning data preferably completely covers the off-the-ground mode envisioned in the operation of the hoist 101, so that the state can be estimated with higher accuracy.

[0070] In addition, regarding the pattern of the hoisting mass when learning data collection, it is preferred to gradually and evenly implement the weight from 0 kg to the overload weight (weight exceeding the rated load of the hoist 101). In addition, as the collection conditions, it is preferred to include at least one of the maximum load suspended by the hoist 101, the rated load, and no load. The reason is that the rated load is a representative design point of each model of the product, the maximum load is an operating point to be confirmed for product safety, and the no load is an operating point that can obtain data with the least influence of noise such as load error.

[0071] Next, the learning and adjustment process of the state estimation formula in the model development stage of the hoisting machine will be described.

[0072] Figure 7 This is a diagram showing the learning / adjustment process of the state estimation formula in the model development stage of the winch. Figure 8 is a functional structure diagram showing learning / adjustment, Fig. 9 The figure shows a hoisting machine system with a learning device for learning / adjustment. The standard state estimation formula is learned in the development stage of the standard model (S701, S702), and the standard state estimation formula is adjusted in the new model developed based on it (S703, S704).

[0073] In the case of learning / adjustment, Figure 5 The hoisting machine system shown in the figure is provided with a learning device 910, a learning data condition input unit 921, and a learning data setting unit 922. The learning device 910 has a learning data storage unit 911, a learning unit 912, and a learning result storage unit 913. In addition, the learning device 910 has hardware such as a CPU or the like, a main storage device 915 such as a semiconductor memory, an auxiliary storage device 916 such as a hard disk, and a communication device 917. The well-known operation of each hardware is described with appropriate omissions.

[0074] The functions of the learning device 910 (learning data storage unit 911, learning unit 912, learning result storage unit 913, etc.) can be realized by, for example, the computing device 914 reading the program stored in the auxiliary storage device 916 to the main storage device 915 and executing (software), or can be realized by hardware such as a dedicated circuit, or can be realized by combining software and hardware. In addition, one function of the learning device 910 can be divided into multiple functions, and multiple functions can also be combined into one function. In addition, part of the functions of the learning device 910 can be set as a separate function, or can be included in other functions.

[0075] The learning data storage unit 911 stores the learning data set by the learning data setting unit 922 of the operation terminal 120 in the auxiliary storage device 916 .

[0076] The learning unit 912 reads the learning data from the learning data storage unit 911 and learns the state estimation formula.

[0077] The learning result storage unit 913 stores the state estimation formula learned by the learning unit 912 in the auxiliary storage device 916 , and sends the state estimation formula to the state estimation unit 513 as appropriate.

[0078] The learning data condition input unit 921 is an interface for inputting learning data conditions such as the state of the motor 220 or the hanging weight 130, the individual identification number of the hoist 101, etc. during the learning stage. Here, the learning data condition is the operating condition (e.g., ground clearance condition) of the hoist 101 corresponding to the operating information acquired by the operating information acquisition unit 512, and is associated with at least one state information (reference value) of the hanging weight mass, tension state, motor temperature, and hanging weight swing amount. In addition, the learning data condition may include information such as the model name and individual identification number of the product. The interface of the learning data condition input unit 921 may be used in common with the interface of the operation input unit 522, or may be implemented as a separate interface.

[0079] The learning data setting unit 922 sets the learning data conditions input from the learning data condition input unit 921 and the operating information sent from the operating information acquisition unit 512 as one set of learning data, and sends it to the learning device 910 .

[0080] In addition, all or part of the functions of the learning device 910 can be implemented using the same machine body as the hoisting machine 101 and the operation terminal 120 or other computers or cloud servers that can communicate with them.

[0081] The following processing is performed in each step of the learning / adjustment process.

[0082] Step S701: The winch system collects learning data (referred to as "standard model learning data") during the lift-off test under conditions (standard collection conditions) that vary the mass of the hoisted weight, the positional deviation of the hoisted weight 130 relative to the trolley 113, and the lifting mode in a standard model. The standard model learning data is data including operation information (rotation speed, current value, etc.) and lift-off conditions (mass of the hoisted weight, positional deviation, etc.). In addition, the lift-off conditions are conditions input by the operator of the winch 101, conditions that automatically set part or all of the collection conditions, etc., including state information (reference value) when the lift-off occurs.

[0083] Step S702: The hoist system uses the standard machine model learning data collected in step S701 to learn a standard state estimation formula including a common estimation unit and a machine model specific estimation unit. For example, the coefficient matrix W and the bias b are optimized for the network 600 using the error back propagation method.

[0084] Step S703: The winch system collects learning data (referred to as "learning data for each model") during the lift-off test under conditions (collection conditions for each model) that vary the mass of the hoisted weight, the position deviation of the hoisted weight 130 relative to the trolley 113, and the lifting mode in the new model developed based on the standard model. The collection conditions for each model may vary less than the standard collection conditions.

[0085] Step S704: The hoist system uses the machine model learning data collected in step S703 to tune only the machine model specific estimation unit. For example, the parameters of the common estimation unit 640 of the standard state estimation formula are directly used for the network 600, and only the coefficient matrix W and the bias b are optimized for the machine model specific estimation unit 650 using the error back propagation method.

[0086] In addition, the learning data used in the above learning / adjustment may also include temperature information. The responsiveness of the motor and the characteristics of lubricants such as grease used in the transmission part change with temperature. Therefore, when collecting learning data, by including data under conditions with different motor temperatures, a state estimation formula that is robust to temperature changes can be obtained. In addition, when the individual differences in motor characteristics of each product are small, steps S703 and S704 are not necessary.

[0087] Next, the effects of this embodiment will be described.

[0088] Generally speaking, the power output from the motor 220 in the winch 101 is affected by the losses of various components (components) in the transmission system, such as the bearing 230, the transmission belt, the gear chain, and the brake. Therefore, in order to perform state estimation using a physical model, the load of each winch 101 needs to be divided into factors such as factors determined by the mass of the hoisted weight, factors determined by the components of the transmission system, and factors determined by individual differences in motor characteristics.

[0089] On the other hand, in order to improve the accuracy of state estimation, it is necessary to adjust the coefficients specified for each factor for each product, and the setting of the coefficients takes time. In addition, since the physical model becomes more complicated, the calculation load of state estimation also becomes higher, so the cost of the control circuit board required to perform high-speed and high-precision calculations tends to increase, and it is not easy to achieve.

[0090] In contrast, when the state estimation formula obtained based on the learning data is used as described in this embodiment, by setting the coefficient matrix W and the bias b in (Formula 1) to appropriate matrices and vectors, the mass of the hanging weight, the tension state, the motor temperature, and the hanging weight swing amount can be estimated with high accuracy. In addition, the feature quantity extraction and parameter adjustment are automatically performed through learning, so the burden of research needs can be suppressed. In addition, because a complex physical model is not used, the calculation load is also reduced, which produces advantages such as being able to suppress the increase in the cost of the control circuit board.

[0091] The above are the effects that can be obtained when estimating one state (single state), but the effect is even better in a use case where multiple states are estimated from one piece of operating information as shown in this embodiment.

[0092] For example, in the state estimation in this embodiment, a neural network is used, and a part of the calculation of multiple estimation items is shared in the shared layer 620. This is because the multiple estimation items are items that can be estimated based on the operation information of the same motor 220, or items that are related to the estimation items (for example, a dependency relationship), so this is possible.

[0093] To explain more specifically, in the calculation of the tension state, a method such as outputting a "tension" judgment when the weight exceeds a specified value can be adopted, so the weight and the tension state are in a relationship that can be estimated based on the same operation information. In other words, the physical background required for the estimation of the weight and the tension state is the same.

[0094] As described above, by estimating information that can be estimated from the same operating information and information in a dependent relationship together using one state estimation formula, it is possible to suppress calculation processing and parameter setting to the minimum necessary and simultaneously estimate a plurality of states of the hoisting machine 101. Therefore, the more states are estimated for one motor 220, the better the effect of state estimation using learning.

[0095] Furthermore, by adjusting only the model-specific estimation unit that differs for each model, it is possible to perform high-precision estimation for each model and reduce the workload of learning the state estimation formula for a new model.

[0096] In addition, in this embodiment, it is assumed that the state estimation is performed using a neural network, but various supervised learning methods can also be used. For example, in this embodiment, the weight mass, the motor temperature, and the weight swing amount are estimated by regression calculation, and the tension state is estimated by classification calculation, but depending on the form of the estimated value obtained, various learning methods such as simple linear regression, random forest, support vector machine, and logistic regression can also be used.

[0097] In addition, in this embodiment, the neural network merged into one is regarded as one state estimation formula to estimate multiple states, but it can also be decomposed into multiple state estimation formulas.

[0098] Through the above mechanism, the hoist system 100 can provide a state estimation formula that can estimate the state of the weight, tension state, etc. with high accuracy with easy parameter setting. In addition, the estimated value of the weight can be used to perform overload prevention control, life prediction, etc., or the estimated value of the tension state can be used to detect ground clearance. Similarly, other states can also be estimated, for example, the weight swing amount can be estimated, and the estimated value can be used to perform emergency stop in the case of dangerous swing, weight swing reduction control, etc.

[0099] Example 2

[0100] (Individual Difference Correction)

[0101] For Example 2, use Fig.10 , Fig.11 The configuration and basic operation of each element in this embodiment are the same as those in the first embodiment, so the description will focus on the features of this embodiment.

[0102] Embodiment 1 can provide the optimal state estimation formula for each model, but even for the same model, there will be differences in characteristics between individuals. This is because the processing errors, assembly errors, coil resistance of the motor, and lubrication conditions (grease conditions) of mechanical parts (gears, bearings) are slightly different in each individual. Therefore, by making corrections for each individual, the accuracy of state estimation can be further improved.

[0103] use Fig.10 Explain the principle of the correction method. Fig.10 This is a graph showing the relationship between the estimated value (state estimation formula output) estimated by the state estimation formula and the measured value of the actual state. This relationship can be measured for each product, and the state estimation formula output can be corrected based on the obtained relationship.

[0104] Fig.11 6 is a diagram showing an example of the structure of a state estimation unit to which a correction of a state estimation formula is applied. A state estimation output correction unit for correction is provided at the rear end of the state estimation formula 600. Here, correction operations 710 and 720 are performed based on an arithmetic expression determined based on the output of the state estimation formula for each estimation item and the measured value of the actual state and the parameters of the arithmetic expression.

[0105] The correction formula and its parameters may be adjusted according to the output of the state estimation formula. This is because, for example, since the characteristics vary depending on the motor temperature, the accuracy can be further improved by making corrections that reflect the estimated motor temperature.

[0106] The state estimation output correction parameter determination unit that determines the correction formula and its parameters based on the output of the state estimation formula and the measured value of the actual state can be set on the same body as the learning device 910, the hoist 101, and the operation terminal 120, or on other computers or cloud servers that can communicate. When set on a cloud server, the state estimation value obtained by the state estimation formula collected in the hoist, or the motor operation information that serves as the basis for the estimation, and the measured value of the actual state are sent to the cloud server, and the correction formula and parameters are determined on the cloud server. The hoist receives them and updates the state estimation output correction unit in the hoist. By using a cloud server, it is possible to perform correction formulas and parameter determination operations with a large amount of calculations, and it is possible to obtain correction formulas and parameters that can perform corrections with higher accuracy.

[0107] In addition, the correction formula and the parameters of the correction formula can be determined before the product is shipped or after the hoist is installed. In the case of determining after the hoist is installed, the influence of the assembly error of the final device and the influence of the dead weight of the rope determined by the difference in the installation height of the hoist can also be corrected.

[0108] Furthermore, the operator can also directly set the correction formula and the parameters of the correction formula. When it is determined based on the measurement results, there is a risk that the measurement results include errors due to the surrounding environment and usage conditions and cannot be accurately determined. By allowing the operator to directly set it, the influence of errors can be considered and a more preferred setting can be made.

[0109] In this way, even when there are differences in characteristics between individuals in the same model, the state can be estimated with high accuracy. In addition, by being able to make corrections for each individual, the likelihood of individual differences can be reduced during the design phase, and effects such as shortened design time and reduced costs can be expected.

[0110] Example 3

[0111] (Deterioration status correction)

[0112] For Example 3, use Fig.12 The configuration and basic operation of each element in this embodiment are the same as those of Embodiment 1 and Embodiment 2, so the description will be centered around the features of this embodiment.

[0113] During the operation phase after the winch is shipped from the factory, the characteristics change due to device degradation, and the deviation between the estimated state and the actual state increases. Therefore, correction is performed based on the device degradation state caused by operation.

[0114] Regarding the degradation state of the device, specifically, the degradation degree of the motor caused by the wear of the motor bearings and the degradation of the lubricant, or the degradation degree of the transmission system components caused by the wear of the components such as bearings and gears connected to the transmission system and the degradation of the lubricant can be used as an indicator. These degradation degrees are correlated with the characteristic quantities of the motor's operating information. The corresponding relationship between the motor's operating information and the degradation degree (degradation state) of the device is calculated in advance, and the degradation state of the device is estimated based on the motor's operating information. The characteristic changes caused by the degradation of the device are the same as the degree of characteristic changes of each individual, and the state estimation output is corrected in consideration of the influence of degradation.

[0115] Fig.121 is a diagram showing an example of the structure of a state estimation unit to which a correction of a degradation state is applied. The degradation state is estimated by a degradation state estimation unit 800 that estimates the degradation state based on the operation information of the motor, and the correction formula and parameters corresponding to the estimated degradation state are called from the state estimation output correction parameter storage unit 830 that has previously accumulated the correspondence between the degradation state and the correction formula and parameters of the state estimation output, and the state estimation output correction unit 700 is updated. The state estimation output correction parameter storage unit can be defined as a database indexed by the degradation state, or a function for the degradation state. When it is determined that the correction formula and parameters called from the state estimation output correction parameter storage unit are not suitable, and the error between the state estimation value and the actual state increases due to device degradation, the correspondence between the degradation state and the correction formula and parameters of the state estimation output is newly accumulated to improve the accuracy of the correction based on the degradation state. Specifically, the correction formula and parameters are determined by the state estimation formula 600 based on the output of the state estimation formula 600 at the moment when the error of the estimated state is judged to be large and the measured value of the actual state, and the determined result and the degradation state estimated by the degradation state estimation unit 800 are added to the state estimation output correction parameter storage unit 830.

[0116] In addition, the state estimation output correction parameter storage unit can be set on the same body as the learning device 910, the hoist 101, and the operation terminal 120, or on other computers or cloud servers that can communicate. When set on a cloud server, the degradation state estimated by the degradation state estimation unit on the hoist or the motor operation information that serves as the basis for the estimation is sent to the cloud server, and the correction formula and parameters corresponding to the sent degradation state are called from the state estimation output correction parameter storage unit and sent to the hoist. The hoist updates the state estimation output correction unit with the received data. In addition, the actual state measured in the hoist can also be sent to the cloud server, and the correction formula and parameters are determined on the cloud server and the corresponding relationship with the degradation state is added to the storage unit. By using a cloud server, the accumulated data of the corresponding relationship between the degradation state and the correction formula and parameters can be increased, and a more optimal correction formula and parameters can be used for high-precision estimation.

[0117] In this way, even if the device is degraded during operation, the state can be estimated with high accuracy. In addition, by correcting the degraded state, the likelihood of degradation can be reduced in the design stage, and effects such as shortening the design time and reducing costs can be expected. Furthermore, the degraded state can be corrected successively to maintain the device state, and effects such as suppressing the deterioration of power consumption caused by degradation can be expected.

[0118] Example 4

[0119] (Other correction methods for individual differences and deterioration conditions)

[0120] Embodiment 4 will be described. The configuration and basic operation of each element in this embodiment are the same as those of Embodiments 1 to 3, so the description will focus on the features of this embodiment.

[0121] As a method for correcting individual differences and degradation states, similar to Example 1, only the model-specific estimation unit of the state estimation formula is adjusted for each individual and each degradation state showing representative characteristics. For each individual and each degradation state with different characteristics, the hoisting system collects learning data (referred to as "each individual / each state learning data") during the lift-off test under conditions (each individual / each state collection condition) that change the mass of the hoisted weight, the position deviation of the hoisted weight 130 relative to the trolley 113, and the lifting mode. The individual / each state collection conditions can vary less than the standard collection conditions. Then, using the collected individual / each state learning data, only the model-specific estimation unit is adjusted. For example, for the network 600, the parameters of the common estimation unit 640 of the standard state estimation formula are directly used, and only the optimization learning of the coefficient matrix W and the bias b is performed on the model-specific estimation unit 650 using the error back propagation method. In this way, the optimal state estimation formula for each individual and each degradation state can be obtained.

[0122] The obtained state estimation formula is accumulated in the learning result accumulation unit. When the state estimation is performed for individuals and degradation states other than those used in the learning, the state estimation formula of individuals / degradation states with similar characteristics is extracted from the accumulated state estimation formulas, and the state estimation is performed. Alternatively, the state estimation formula of individuals / degradation states with similar characteristics may be extracted and the state estimation is performed, and the average value obtained by weighting the results according to the similarity of the characteristics is used as the state estimation value.

[0123] Thus, even if there are individual and state-deteriorated characteristic differences, the state estimation can be performed with high accuracy. The above-described embodiments are directed to a hoisting machine, but the present invention can also be applied as a state estimation system for a device equipped with a motor other than a hoisting machine.

[0124] In addition, the present invention is not limited to the above-mentioned embodiments, and includes various modified examples. The above-mentioned embodiments are described in detail to explain the present invention in an easy-to-understand manner, and are not limited to all structures that must be described. In addition, a part of the structure of a certain embodiment can be replaced with the structure of other embodiments, and the structure of other embodiments can be added to the structure of a certain embodiment. In addition, for a part of the structure of each embodiment, other structures can be added, deleted, or replaced.

[0125] Description of Reference Numerals

[0126] 100…Winch system, 101…Winch, 120…Operation terminal, 910…Learning device.

Claims

1. A winch having an electric motor and controlling the electric motor to lift and lower a load, characterized in that: include: An operation information acquisition unit for acquiring operation information of the motor; and a state estimation unit that estimates the state of the hoist based on the operation information acquired by the operation information acquisition unit and a state estimation formula learned to be able to estimate the state of the hoist, The state estimation formula includes a common estimation part irrespective of the model of the hoisting machine and a model-specific estimation part which is different for each model of the hoisting machine.

2. The winch according to claim 1, characterized in that: A cable capable of installing the hanging weight is provided, The state estimation unit estimates at least one of the mass of the weight mounted on the cable, the acceleration of the weight, the tension of the cable, the rotational load of the motor, the tension state of the cable, the temperature of the motor, and the weight swing amount of the weight as the state of the hoist.

3. The winch according to claim 1, characterized in that: The operation information of the motor includes at least one of the following information: a voltage value applied to the motor, a command frequency of a control unit of the motor, a rotation speed of the motor, a drive current value input to the motor, an excitation current value in vector control of the motor, a torque current value in vector control of the motor, an input torque value to the motor, a command speed input to the motor, a slip value of the rotation speed of the motor corresponding to the command speed input to the motor, a temperature of the motor, a degradation degree of the motor, and a degradation degree of a component connected to a transmission system of the motor.

4. The winch according to claim 3, characterized in that: The operation information of the electric motor used by the state estimation unit to estimate the state of the hoisting machine is time series information.

5. The winch according to claim 1, characterized in that: The state estimation unit includes a state estimation output correction unit that corrects, for each hoisting machine, a state estimation value obtained using the state estimation formula.

6. A winch system, comprising a winch having an electric motor and controlling the electric motor to lift and lower a load, wherein the winch system comprises: An operation information acquisition unit for acquiring operation information of the motor; a state estimation unit that estimates the state of the hoist based on the operation information acquired by the operation information acquisition unit and a state estimation formula learned to be able to estimate the state of the hoist; and a learning device for learning a state estimation formula for estimating the state of the hoisting machine, The state estimation formula is composed of a common estimation part that is independent of the model of the hoisting machine and a model-specific estimation part that is different for each model of the hoisting machine. The learning device comprises: a learning data storage unit that stores learning data that associates the operation information of the electric motor with information indicating a state of the hoisting machine; a learning unit that learns the state estimation formula using the learning data accumulated by the learning data accumulation unit; and a learning result storage unit for storing the state estimation formula learned by the learning unit, The learning unit learns the state estimation formula of the standard model based on standard learning data that associates the operating information in the standard model with the state, and adjusts the state estimation formula based on learning data of each model that associates the operating information in models other than the standard model with the state.

7. A winch system, comprising a winch having an electric motor and controlling the electric motor to lift and lower a load, wherein the winch system comprises: An operation information acquisition unit for acquiring operation information of the motor; a state estimation unit that estimates the state of the hoist based on the operation information acquired by the operation information acquisition unit and a state estimation formula learned to be able to estimate the state of the hoist; and a learning device for learning a state estimation formula for estimating the state of the hoisting machine, The state estimation unit includes a state estimation output correction unit that corrects a state estimation value obtained using the state estimation formula for each of the hoisting machines.

8. The winch system according to claim 7, characterized in that: A state estimation output correction parameter determination unit is provided for determining a calculation formula and a parameter of the state estimation output correction unit based on a correspondence relationship between a state estimation value obtained by the state estimation formula and a measurement value of an actual state of the hoist.

9. The winch system according to claim 8, characterized in that: The state estimation output correction parameter determination unit is provided in a computer connected to the hoist via a network. The hoist sends the state estimation value obtained by the state estimation formula or the operation information of the motor which is the basis of the state estimation value, and the measured value of the actual state of the hoist to the computer, The computer determines the operation formula and parameters of the state estimation output correction unit by using the state estimation output correction parameter determination unit, and sends the determined result to the hoist. The hoisting machine updates the calculation formula and parameters of the state estimation output correction unit based on the transmitted result.

10. The winch system according to claim 7, characterized in that: have: a degradation state estimation unit that estimates the degradation state of the hoisting machine from the operation information of the electric motor; and a state estimation output correction parameter storage unit that stores the correspondence between the estimated value of the deterioration state of the hoisting machine estimated by the degradation state estimation unit and the calculation formula and parameter of the state estimation output correction unit, Based on the correspondence relationship accumulated in the state estimation output correction parameter accumulation unit, the calculation expression and the parameter of the state estimation output correction unit are updated to correspond to the degradation state estimated value estimated by the degradation state estimation unit.

11. The winch system according to claim 10, characterized in that: The deterioration state of the hoisting machine includes at least one of a deterioration degree of the electric motor and a deterioration degree of a component connected to a power transmission system of the electric motor.

12. The winch system according to claim 10, characterized in that: When it is determined that the error between the estimated state value estimated by the state estimation unit and the actual state of the hoisting machine is large, Obtaining a correspondence between the estimated state value and the measured value of the actual state of the winch, determining the operation expression and the parameter of the state estimation output correction part based on the correspondence relationship acquired by the state estimation output correction parameter determination part that determines the operation expression and the parameter of the state estimation output correction part, The degradation state estimated value estimated by the degradation state estimating unit and the determined calculation formula and parameter of the state estimation output correction unit are added to the state estimation output correction parameter storage unit.

13. The winch system according to claim 12, characterized in that: The state estimation output correction parameter storage unit is provided in a computer connected to the hoist via a network. the hoisting machine sends the estimated degradation state value of the hoisting machine estimated by the degradation state estimating unit or the operation information of the motor serving as a basis for the estimated degradation state value to the computer; The computer acquires the calculation formula and parameter of the state estimation output correction unit corresponding to the sent degradation state estimation value based on the correspondence between the degradation state estimation value accumulated in the state estimation output correction parameter accumulation unit and the calculation formula and parameter of the state estimation output correction unit, and sends the acquired result to the hoist. The hoist updates the calculation formula and parameters of the state estimation output correction unit based on the sent result, When it is determined that the error between the estimated state value estimated by the state estimation unit and the actual state of the hoist is large, the hoist sends the estimated state value of degradation and the estimated state value or the operation information of the motor as the basis for the estimation, and the measured value of the actual state of the hoist to the computer, The computer determines an operation expression and parameters of the state estimation output correction unit based on the transmitted information using the state estimation output correction parameter determination unit, and adds the degradation state estimation value and the determined operation expression and parameters of the state estimation output correction unit to the state estimation output correction parameter accumulation unit.

14. A state estimation system for estimating the state of a device having an electric motor, characterized in that: include: An operation information acquisition unit for acquiring operation information of the motor; and a state estimation unit that estimates the state of the device based on the operation information acquired by the operation information acquisition unit and a state estimation formula learned to be able to estimate the state of the device, The state estimation formula is composed of a common estimation part that is independent of the model of the device and a model-specific estimation part that is different for each model of the device.

15. A state estimation system for estimating a state of a device having an electric motor, characterized in that: include: An operation information acquisition unit for acquiring operation information of the motor; and a state estimation unit that estimates the state of the device based on the operation information acquired by the operation information acquisition unit and a state estimation formula learned to be able to estimate the state of the device, The state estimation unit includes a state estimation output correction unit that corrects the state estimation value obtained using the state estimation formula for each of the devices.

Images