Gyroscope optical fiber coil manufacturing system

Abstract

A system (S) for manufacturing an optical fiber coil of a gyroscope, the system comprising: a mandrel (M) having respective actuating means (W) for forward and reverse rotation; a robot arm (RA) equipped at an end portion with a first tip (P1), where the first tip has an elongated and partially flexible shape to guide the positioning of the optical fiber during the winding of the optical fiber on the mandrel (M); one or more optical means (C) arranged to continuously acquire images of the coil during the winding process; processing means (CU) operatively connected to the robot arm (RA), the one or more optical means (C) and the spindle-rotating actuation means (W), where the processing means are configured to control the robot arm and define means for controlling the spindle-rotating actuation means (W) in dependence on the continuously acquired images in order to control the manufacturing of the optical fiber coil.

Description

Technical Field

[0001] The present invention relates to the field of manufacturing optical fiber coils for gyroscopes. Background Art

[0002] Fiber optic gyroscopes (FOGs) are widely used sensing devices: used in navigation and positioning systems, angular velocity sensors, stabilization devices, and backup systems for driving autonomous vehicles in remote areas where GPS is not accessible, etc.

[0003] High-performance FOGs must accurately detect the phase difference between two optical signals caused by rotation and minimize or compensate for the phase difference caused by other sources (such as temperature effects).

[0004] Existing gyroscope coil winding techniques require high manual skills to achieve full precision. Attempts have been made to automatically wind quadrupole gyroscope coils, but these attempts have only been successful for very low-performance coils.

[0005] Before applying glue to each layer of the coil, the technician corrects it in case of errors.

[0006] Typically, the technician uses a plastic nib with a flexible tip to guide the optical fiber during the winding process to avoid overlap. In addition, since these are optical fibers with a very limited cross-section, the technician needs to put in a considerable amount of visual effort to track the process and intervene in a timely manner, unwinding and rewinding the optical fiber in case of defects.

[0007] For this purpose, the technician controls the forward and reverse rotational movements of the mandrel around which the coil is wound.

[0008] Therefore, the technician must take a break to rest his eyesight and regain the correct attention.

[0009] Since quadrupole winding is slow and laborious work, the number of technicians who can manufacture high-performance coils is limited.

[0010] All of these impose significant limitations on the achievable production volume, resulting in a very high production cost for gyroscopes.

[0011] Unless explicitly excluded in the following detailed description, the content described in this section will be considered as part of the detailed description. Summary of the Invention

[0012] The object of the present invention is to automate the process of winding an optical fiber to form a coil suitable for installation in a gyroscope.

[0013] The basic idea of the present invention is to implement a robotic arm on which a first tip with a flexible tip made of plastic material is fixed, and this first tip is currently being used by a winding technician. The task of the robotic arm is to guide the optical fiber during the winding process. One or more optical devices are arranged to continuously acquire images of the coil during the winding process. A computer is operably connected to the robotic arm, at least one optical device, and a device for controlling the rotation of the mandrel. A neural network runs on the computer, and this neural network is set to control the robotic arm and the device for controlling the rotation of the mandrel based on the continuously acquired images.

[0014] The system also includes a training and supervision station to accommodate the winding technician, including:

[0015] - At least one display operably connected to one or more optical devices for projecting the continuously acquired images,

[0016] - A human / machine interface device, including

[0017] ο A second tip with a tactile interface, similar to the first tip, equipped with force and / or deformation and position sensors, and

[0018] ο A support system for the second tip, which is equipped with a device for acquiring the spatial position of the second tip, and this device is suitable for detecting the rotational translation of the second tip applied by the winding technician, and an interface with the second tip, equipped with a force sensor.

[0019] - A human / machine interface device for acquiring commands regarding the control of the rotation of the mandrel.

[0020] Preferably, the system also includes a glue dispenser and a glue dispenser control device, and this glue dispenser control device is operably connected to the said computer and is also connected to the human / machine interface device associated with the training station.

[0021] The dependent claims describe the preferred embodiments of the present invention and form part of this specification. Description of the Drawings

[0022] Other objects and advantages of the present invention will become clear from the following detailed description of examples of its embodiments (and their variations) and the drawings given purely for purposes of explanation and without limitation, where:

[0023] Figure 1 An example of a winding system according to the present invention is shown;

[0024] Figure 2 And Figure 3 Shows the details of the system according to Figure 1 ;

[0025] Figure 4shows Figure 1 a flowchart of the information flow in the system shown

[0026] The same reference numerals and letters in the drawings represent the same elements or components or functions

[0027] It should also be noted that the terms "first", "second", "third", "higher", "lower", etc. may be used here to distinguish various elements. Unless clearly stated or inferred from the text, these terms do not imply a spatial, sequential or hierarchical order of the elements being modified

[0028] It is emphasized that the use of neural networks or inference engines or fuzzy logic is completely equivalent

[0029] The elements and features shown in various preferred embodiments including the drawings may be combined with each other without departing from the scope of protection of the present application as described below Detailed Description of the Invention

[0030] Figure 1 shows an example of a gyroscope fiber optic coil winding system

[0031] The system includes

[0032] - a mandrel M having associated means W for actuating forward and reverse rotation to wind or unwind the fiber optic coil; generally, the mandrel itself is a known winding device and is directly controlled by the winding technician when the coil is directly operated by the winding technician

[0033] - a robotic arm RA having at its end portion a first tip P1 having an elongated and partially flexible shape for guiding the positioning of the fiber optic during winding around the mandrel M; preferably, the interface between the tip and the last joint of the robotic arm includes a force sensor arranged to measure the force and torque exchanged between the first tip and the robotic arm as a reaction to the contact between the first tip and the fiber optic; alternatively, the first tip is equipped with at least one strain gauge

[0034] - one or more optical devices C arranged to continuously acquire images of the coil during winding; preferably, the optical detection device includes a camera; the optical detection device is preferably fixed to the robotic arm to fully frame the tip of the first tip, but it may also be associated with a fixed point in the environment in which the system is located

[0035] - a computer CU operatively connected to the robotic arm RA, one or more optical devices C and the actuating means W for the rotation of the mandrel. The computer runs a learning neural network which is set to

[0036] · control the robotic arm, and

[0037] · Means for limiting the actuation device W for controlling the spindle rotation based on continuously acquired images.

[0038] The system further includes a supervision and training station for accommodating TCH winding technicians:

[0039] - At least one display DS for projecting images continuously acquired by one or more optical devices C,

[0040] - Human / machine interface device, including:

[0041] ο A second tip P2, similar to the first tip P1, and a second tip holder AQ, which is equipped with means for acquiring the spatial position of the second tip, the means being adapted to detect the rotational translation of the second tip applied by the winding technician, and wherein the support is equipped with an actuator to make the second tip tactile,

[0042] ο A human / machine interface device CM for controlling the spindle rotation actuation device W.

[0043] The computer CU is configured to acquire images from one or more optical devices C and use them to control the robotic arm RA. At the same time, the images are displayed on at least one display DS so that, if necessary, the TCH winding technician can monitor the behavior of the robotic arm.

[0044] It is worth emphasizing that this solution will make the posture of the winding technician more comfortable even without the autonomous ability of the computer CU to control coil manufacturing. The winding technician can sit comfortably at his supervision station instead of standing near the coil.

[0045] In case of unsatisfactory behavior, the winding technician can directly control the actuation device W for spindle rotation and the robotic arm RA respectively using the following devices:

[0046] - A human / machine interface device CM, which can be, for example, a keyboard, a mouse or a joystick, etc.,

[0047] - The second tip P2.

[0048] The computer CU acquires the movement applied by the winding technician to the second tip through the second tip holder AQ and controls the robotic arm RA to reproduce the same movement in real time.

[0049] At the same time, the neural network implemented on the same computer performs self-training:

[0050] - Learning that the winding technician considers the fiber configuration on the coil to be unsatisfactory, and

[0051] - Learning the actions required to correct this unsatisfactory configuration.

[0052] Obviously, in order to allow the winding technician to better train the neural network, it is recommended that the second tip be tactile, reproducing the same resistance torque perceived by the robotic arm RA that activates the first tip.

[0053] The support system of the second tip is preferably of the pantograph type, also known as DELTA, which in any case allows the operator to hold the haptic device as if holding an ordinary tip, thus not forcing the operator to use an unfamiliar grip. In other words, the operator must feel able to use his skills as if he were operating directly on the optical fiber, without the intervention of the monitoring station and the robotic arm.

[0054] The robotic arm is preferably anthropomorphic with six degrees of freedom, but can also be of a delta configuration with five degrees of freedom.

[0055] During the direct control of the winding operation, the neural network receives the following inputs:

[0056] - Images continuously acquired using at least one optical device;

[0057] - The position defining the kinematic chain of the robotic arm;

[0058] - The magnitude of the reaction force sensed by the first tip when the tip is in contact with the optical fiber;

[0059] - The angular position and rotational speed of the mandrel;

[0060] - Coil winding state data such as the number of layers, the number of turns per layer, the optical fiber tension, the coil size, the mandrel size, and the optical fiber type.

[0061] During the training of the direct control of the winding operation by the winding technician, in addition to receiving the previous information, the neural network also receives the inputs:

[0062] - The spatial position of the second tip,

[0063] - Commands issued by the winding technician via the human / machine interface CM.

[0064] The neural network is preferably of the deep supervised learning type. Supervised learning allows the operator to signal the correct movements and correct results to the neural network in a robust manner during the learning phase, and by leaving error reports and system guidance during error recovery, greatly reduces the possibility of incorrect learning. One of the main features of deep learning is its excellent image analysis ability. An example of a dataset useful for learning is a dataset whose features include the winding state, such as: images of optical fibers wound on a mandrel or film and related features such as the mutual distance between the optical fibers, the height of the current coil relative to other turns of the same layer, in order to visualize the correct or incorrect positioning of the optical fibers. On the other hand, it involves the acceleration and / or speed and / or position of the mandrel, the force and torque applied to the tip, the commands sent by the haptic interface, when the robotic arm is controlled by the operator, obtaining the position, orientation, speed and angular acceleration of the haptic tip, the amount of glue released from the dispenser, the optical fiber tension, the cross-section and type of the optical fiber and mandrel part, the coil size and the number of turns per layer.

[0065] All these features are related to the operator's assessment of the winding quality through the haptic interface during the winding operation, which is usually referred to as "online" learning, for example, in the case of supervised learning, this is useful for training the neural network inference engine.

[0066] Figure 4 An example flowchart showing the information flow method according to the present invention is shown.

[0067] The system has been defined as an assembly that includes a mandrel M with associated actuating means W for forward and reverse rotation, a robotic arm RA equipped with a tip P1 at its end portion, where the first tip has an elongated and partially flexible shape for guiding the positioning of the optical fiber during the winding of the optical fiber on the mandrel M, one or more optical devices C arranged to continuously acquire images of the coil during winding, and a processing device CU.

[0068] The Figure 1 As always, it starts from the start block.

[0069] Step 1: Activate the system;

[0070] Step 2: The processing device acquires information about:

[0071] - Mandrel position,

[0072] - Mandrel speed,

[0073] - One or more sensors, such as force sensors, encoders, strain gauges, etc., in addition,

[0074] - The trajectory of the tip, etc.,

[0075] - The video stream from the camera, and preferably also

[0076] - The amount of glue layer-by-layer dispensed from the dispenser onto the coil,

[0077] - The type of optical fiber, the number of layers, and the number of turns per layer,

[0078] - The tension of the optical fiber during the winding process;

[0079] Step 3: Integrate the previously obtained information to create a dataset that correlates all the above information on the same timeline;

[0080] Step 4: Memorize the information processed in the previous steps;

[0081] Step 5: Process the integrated data in batches to enable iteration;

[0082] Step 6: Gradually power the robotic arm control model e;

[0083] Step 7: Based on the control model generated in the previous steps, perform real-time control and analysis of the winding process;

[0084] Step 8: Check if the winding operation is completed. If so,

[0085] Step 9: Stop. Otherwise,

[0086] Step 10: Check if there is an alarm or signal. Such an alarm may be caused by a quantity outside the previously observed range, or due to the control model being unable to independently control the system for any reason. If so,

[0087] Step 11: Stop. Otherwise, return to Step 2.

[0088] It should be emphasized that Steps 1 to 6 must be executed continuously and restarted from Step 1 at least once to allow the system to process the information flow and generate at least the first model, which will then be further updated. At the same time, subsequent Steps 7 to 11 will be executed in real time. This is essentially a real-time execution process, and Steps 5 and 6, which are executed offline in particular, are part of this process to update the control model.

[0089] The present invention can be advantageously implemented by a computer program that includes encoding means for performing one or more steps of the method when the program is executed on a computer. Therefore, it can be understood that the scope of protection extends to the said computer program and further to a computer-readable device including recorded messages, said computer-readable device including program encoding means for performing one or more steps of the method when the program runs on a computer.

[0090] Structural changes to the described non-limiting examples are possible without departing from the protective purpose of the present invention, including all embodiments equivalent in content to the claims for those skilled in the art.

[0091] Based on the above description, those skilled in the art can achieve the purpose of the present invention without introducing further structural details.

Claims

1. A gyroscope fiber optic coil manufacturing system (S), comprising: - A mandrel (M), the mandrel having respective actuating means (W) for forward and reverse rotation; - A robotic arm (RA), the robotic arm being equipped at its end portion with a first tip (P1), wherein the first tip has an elongated and partially flexible shape for guiding the positioning of the optical fiber during the winding of the optical fiber around the mandrel (M); - One or more optical devices (C), the optical devices being arranged to continuously acquire images of the coil during the winding process; - A processing device (CU), the processing device being operatively connected to the robotic arm (RA), the one or more optical devices (C), and the actuating means (W) for the rotation of the mandrel; Wherein the processing device is configured to: · Control the robotic arm, and · Control the actuating means (W) for the rotation of the mandrel according to the continuously acquired images, To control the manufacture of the fiber optic coil.

2. The system according to claim 1, further comprising a supervision station, the supervision station comprising: - At least one display (DS) for displaying the images continuously acquired by the one or more optical devices (C), - A human / machine interface device (P2, CM), comprising: ○ A second tip (P2), ○ A second tip holder (AQ) system, equipped with: + A tactile interface and means for acquiring the spatial position of the second tip, capable of detecting the rotational translation of the second tip, and + A connection interface to the second tip, the connection interface being equipped with a force sensor arranged to measure the force acting on the second tip, ○ A human / machine interface device (CM) for controlling the actuating means (W) for the rotation of the mandrel.

3. The system according to claim 2, wherein, The supervision station is arranged to preferentially control the robotic arm (RA) and the actuating means (W) for the rotation of the mandrel with respect to the processing device (CU).

4. The system according to any one of claims 1-3, wherein, The processing device (CU) is configured by a learning neural network.

5. The system according to claim 4, wherein, The neural network is set to enter a learning mode when the supervision station directly controls the robotic arm (RA) and / or the actuating means (W) for the rotation of the mandrel.

6. The system according to claim 4 or 5, wherein, The neural network is configured to be trained by control signals generated by the means for acquiring the spatial position of the second tip and the human / machine interface device (CM) of the supervision station.

7. The system according to any one of claims 4-6, wherein, The neural network is configured to be trained by the coil image stream generated by the optical device during the winding process.

Images