A method for data fusion and alignment of a molding press and an automation machine

Abstract

This invention discloses a method for data fusion and alignment between a molding machine and an automatic machine, belonging to the field of industrial automation control technology. It includes collecting the operating states of the molding machine and the automatic machine to generate synchronous time-series data; extracting key time points in the molding process to form molding process feature events; and extracting key time points in the automatic machine's operation to form automatic machine action feature events, resulting in a heterogeneous feature event sequence. Using the molding process feature events as anchor points, an alignment analysis window is defined on the time axis. The molding process feature events and the automatic machine action feature events are matched to generate aligned event pairs and calculate timing deviation parameters, generating collaborative control commands for adjusting the automatic machine's operation. This invention employs data synchronous acquisition based on a unified time reference, heterogeneous feature event extraction and matching, and timing deviation analysis to achieve precise alignment and dynamic collaborative control of the molding machine process and the automatic machine's actions.

Description

Technical Field

[0002] This invention relates to the field of industrial automation control technology, and in particular to a method for data fusion and alignment between a molding machine and an automatic machine. Background Technology

[0004] In the field of high-precision optical component manufacturing, aspherical glass lenses are widely used in high-end camera lenses, medical endoscopes, and automotive cameras due to their ability to correct aberrations and simplify optical system design. Glass molding technology is the mainstream process for producing these high-precision aspherical lenses. This process involves directly hot-pressing a glass preform under high temperature and pressure using an ultra-precision mold to create a lens with the final optical surface. In the entire automated production process, the molding machine is responsible for the core molding process, while the automated machine, typically a multi-axis industrial robot, performs key auxiliary operations such as loading / unloading and lens transfer. The high degree of coordination and timing between the molding machine's process cycle and the automated machine's loading / unloading actions directly determines the final optical quality of the lens and the overall efficiency of the production line.

[0005] Currently, in automated optical lens production lines, the commonly used technical solution for achieving collaborative operation between molding machines and automated machines is based on logical interaction using discrete input / output signals, combined with a fixed time delay to control the action sequence. Specifically, after the molding machine completes a heating, pressure holding, and cooling cycle and the mold opens, its control system sends a digital signal to the automated machine's controller indicating "mold opening complete" or "material removal permitted." Upon receiving this signal, the automated machine does not act immediately but waits for a fixed time delay preset by the process engineer. Only after this delay does it drive the robotic arm into the mold cavity to perform the material removal action. This fixed delay time is designed to provide a necessary cooling window for the newly formed hot lens.

[0006] However, when facing the stringent requirements of high-precision optical lens production, this fixed-delay-based collaborative approach lacks precision and adaptability. It cannot cope with the dynamic changes in the actual process time of the molding machine, such as slight differences in cooling rates caused by fluctuations in equipment operating temperature or changes in hydraulic system pressure. Secondly, inaccurate material picking timing poses a direct threat to lens quality. If material is picked up too early, the lens temperature is still above the glass transition temperature, and mechanical gripping will introduce stress, leading to out-of-tolerance lens surface accuracy and residual internal stress. If material is picked up too late, it not only reduces the production cycle time but may also increase the risk of dust adhesion due to excessive cooling of the lens surface. Finally, this solution lacks flexibility and the data is isolated. When changing to different specifications of lens molds or optimizing process parameters, experienced technicians must spend a lot of time repeatedly trying and resetting the delay parameters. Furthermore, when batch quality problems occur, because the molding machine's process data and the automatic machine's timing data are independent, it is difficult to trace and determine the root cause of the defects. Summary of the Invention

[0008] To address the aforementioned issues, this invention provides a method for data fusion and alignment between a molding machine and an automatic machine. By employing methods such as synchronous data acquisition based on a unified time reference, heterogeneous feature event extraction and matching, and time-series deviation analysis, it is possible to achieve precise alignment and dynamic collaborative control of the molding machine process and the automatic machine's actions.

[0009] The above objectives can be achieved through the following approach:

[0010] A method for data fusion and alignment between a molding machine and an automatic machine includes: collecting the operating status of the molding machine and the automatic machine, and generating synchronous time-series data by combining them with a unified time reference; based on the synchronous time-series data, extracting key time points in the molding process to form molding process feature events, and extracting key time points in the automatic machine's operation process to form automatic machine action feature events; combining the molding process feature events and the automatic machine action feature events to obtain a heterogeneous feature event sequence; based on the heterogeneous feature event sequence, using the molding process feature events as anchor points, dividing the time axis into alignment analysis windows, and matching the molding process feature events with one or more automatic machine action feature events within the alignment analysis windows to generate alignment event pairs; calculating timing deviation parameters based on the time relationship between corresponding events in the alignment event pairs, and generating collaborative control commands for adjusting the operation of the automatic machine based on the timing deviation parameters.

[0011] Optionally, generating synchronized timing data includes: generating a unified timestamp through a shared time synchronization unit in the control systems of the molding machine and the automatic machine; and using the unified timestamp to synchronously collect the operating status of the molding machine and the automatic machine to obtain synchronized timing data.

[0012] Optionally, obtaining the heterogeneous feature event sequence includes: based on the synchronous time-series data, acquiring the pressure peak time generated by the molding machine pressure data and the temperature inflection point time generated by the molding machine temperature data; using the pressure peak time and the temperature inflection point time to construct molding process feature events; based on the synchronous time-series data, acquiring the robotic arm end-effector arrival time or gripper state switching time generated by the automatic machine position or state data to construct automatic machine action feature events; and combining the molding process feature events and the automatic machine action feature events to obtain the heterogeneous feature event sequence.

[0013] Optionally, dividing the alignment analysis window on the time axis includes: obtaining the occurrence time of the molding process feature event in the heterogeneous feature event sequence; using the occurrence time as the starting boundary; and using the starting boundary to divide the alignment analysis window on the time axis according to a preset time length.

[0014] Optionally, the method further includes: before dividing the alignment analysis window, obtaining alignment event pairs within a historical period, and calculating a process stability index based on the historical alignment event pairs; and adjusting the time length according to the process stability index.

[0015] Optionally, generating aligned event pairs includes: performing correlation sliding calculation on the molding process feature event and the automaton action feature event within the alignment analysis window to obtain event correlation values; sorting the event correlation values ​​from largest to smallest, selecting the automaton action feature event corresponding to the first ranked event correlation value to obtain candidate action feature events; and combining the candidate feature events with the molding process feature event to obtain aligned event pairs.

[0016] Optionally, the method further includes: after the automaton completes the action according to the cooperative control instruction, collecting the adjusted operating status in the next production cycle and correcting the automaton action feature events; and using the corrected automaton action feature events to update the heterogeneous feature event sequence.

[0017] Optionally, the method further includes: obtaining product quality inspection results corresponding to the same production cycle as the alignment event pair; associating the alignment event pair with the product quality inspection results to establish a quality traceability data chain.

[0018] Based on the same inventive concept, this invention also provides a data fusion and alignment system for molding machines and automatic machines. The system includes: a data acquisition module for acquiring the operating status of the molding machine and the automatic machine, and generating synchronous time-series data by combining them with a unified time reference; a feature recognition module for extracting key time points in the molding process to form molding process feature events based on the synchronous time-series data, and extracting key time points in the automatic machine's operation process to form automatic machine operation feature events, and combining the molding process feature events and the automatic machine operation feature events to obtain a heterogeneous feature event sequence; an alignment matching module for dividing an alignment analysis window on the time axis based on the heterogeneous feature event sequence, using the molding process feature events as anchor points, and matching the molding process feature events with one or more automatic machine operation feature events within the alignment analysis window to generate aligned event pairs; and a control generation module for calculating timing deviation parameters based on the time relationship between corresponding events in the aligned event pairs, and generating cooperative control commands for adjusting the operation of the automatic machine based on the timing deviation parameters.

[0019] Compared with the prior art, the present invention has the following advantages:

[0020] 1. This invention significantly improves the accuracy and stability of multi-device collaborative control; by constructing a unified time base and collecting synchronous timing data, the problem of time asynchrony between heterogeneous devices is eliminated from the data source; furthermore, by extracting key feature events and performing alignment matching, the traditional collaborative method based on coarse signal interaction and fixed delay is upgraded to quantitative alignment based on precise process nodes and action nodes, so that the actions of the automatic machine can accurately respond to the actual process state of the molding machine, greatly improving the timing accuracy and repeatability consistency of collaborative actions;

[0021] 2. This invention enhances the adaptability and intelligence of automated systems. The method can calculate timing deviations in real time based on alignment events and automatically generate collaborative control commands to dynamically adjust the operation of automated machines. This means that the system can learn autonomously and compensate for timing drift caused by equipment wear, environmental changes, or process parameter adjustments. It can maintain the optimal collaborative cycle without human intervention, thereby reducing the dependence on operator experience, shortening production line debugging and product changeover time, and improving the overall flexible production capability.

[0022] 3. This invention establishes a deep correlation between process timing and product quality, providing powerful process optimization and quality traceability capabilities. By associating aligned event pairs with product quality inspection results in the same cycle, a clear quality traceability data chain is constructed. When quality problems occur, the specific production cycle can be quickly traced back, and the timing deviations of key events can be analyzed to determine if they are abnormal. This allows for precise identification of the root cause of the problem as equipment mismatch, providing direct data support for process optimization and fault diagnosis, and changing the previous situation of difficult quality traceability and ambiguous problem location.

[0023] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart illustrating a method for data fusion and alignment between a molding machine and an automatic machine according to an embodiment of the present invention.

[0027] Figure 2 This is a schematic diagram of the timing data of the molding machine and the automatic machine in an embodiment of the present invention.

[0028] Figure 3 This is a schematic diagram of the structure of a data fusion and alignment system for a molding machine and an automatic machine according to an embodiment of the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Reference Figure 1One embodiment of the present invention proposes a data fusion and alignment method for molding machines and automatic machines. By employing methods such as synchronous data acquisition based on a unified time reference, heterogeneous feature event extraction and matching, and time sequence deviation analysis, the method can achieve precise alignment and dynamic collaborative control of molding machine processes and automatic machine actions.

[0032] The method described in this embodiment specifically includes:

[0033] The system collects the operating status of the molding machine and the automatic machine, and combines this data with a unified time reference to generate synchronized time-series data.

[0034] Based on the synchronous time series data, key time points in the molding process are extracted to form molding process feature events, and key time points in the automatic machine operation process are extracted to form automatic machine operation feature events. Combining the molding process feature events and the automatic machine operation feature events, a heterogeneous feature event sequence is obtained.

[0035] Based on the heterogeneous feature event sequence, taking the molding process feature event as the anchor point, an alignment analysis window is divided on the time axis, and within the alignment analysis window, the molding process feature event is matched with one or more automatic machine action feature events to generate alignment event pairs;

[0036] Based on the time relationship between corresponding events in the alignment event pair, the timing deviation parameter is calculated, and a cooperative control command for adjusting the operation of the automaton is generated according to the timing deviation parameter.

[0037] Specifically, by establishing a unified time benchmark, data is synchronously collected from two independently operating devices, the molding machine and the automatic machine, thus constructing a bias-free unified time coordinate system at the data source. Based on this, massive amounts of continuous raw operational data undergo dimensionality reduction and semantic processing to extract discrete feature events representing core process states and key physical actions. These events, from different devices and with different physical meanings, are integrated into a temporally ordered heterogeneous feature event sequence. Subsequently, using key molding process events as anchor points, intelligent matching is performed on the time axis to find the automatic machine action events most logically and temporally related, thereby establishing a clear correspondence between the two. Finally, by quantifying the actual temporal relationship between these matched events, the timing deviation from the ideal coordinated state is calculated, and specific control commands are generated based on this deviation. This achieves precise and dynamic adjustment of the automatic machine's action timing, forming a complete data-driven closed-loop feedback control loop.

[0038] Optionally, the generation of synchronization timing data includes:

[0039] In the control systems of molding machines and automatic machines, a unified timestamp is generated through a shared time synchronization unit;

[0040] Using the unified timestamp, the operating status of the molding machine and the automatic machine are synchronously collected to obtain synchronous time-series data.

[0041] Specifically, such as Figure 2 As shown, to achieve data fusion and alignment between the molding machine and the automatic machine, it is first necessary to generate a set of synchronized time-series data that is completely synchronized in time. The core of this process lies in establishing a unified time reference to ensure that the data collected from the two independent devices can correspond precisely in the time dimension. Operationally, the first step is to integrate or connect a shared time synchronization unit in the control systems of the molding machine and the automatic machine. This time synchronization unit, as an authoritative time source, can generate and distribute a unified timestamp. This unified timestamp is a high-precision time marker that provides a unique and common time reference for all subsequent data acquisition activities. The second step is to use this unified timestamp generated by the shared time synchronization unit to synchronously collect data on the operating status of the molding machine, such as pressure and temperature changes within the mold cavity, and the operating status of the automatic machine, such as the spatial position of the robotic arm and the opening and closing state of the end effector gripper. Synchronous acquisition means that at any given time point, the system will simultaneously record the status data of both the molding machine and the automatic machine and assign the same unified timestamp to this set of data. In this way, the final result is synchronous time-series data, which is a collection of data points. Each data point contains the operating status of the molding machine and the automatic machine at the same moment, thus ensuring time consistency at the data source.

[0042] Optionally, obtaining the heterogeneous feature event sequence includes:

[0043] Based on the synchronous time-series data, the pressure data of the molding machine is obtained to generate the pressure peak time, and the temperature data of the molding machine is obtained to generate the temperature inflection point time. Using the pressure peak time and the temperature inflection point time, the characteristic events of the molding process are constructed.

[0044] Based on the synchronous timing data, the robot arm end-effector arrival time or gripper state switching time generated by the automaton position or state data is obtained, and automaton action feature events are constructed.

[0045] By combining the molding process feature events with the automatic machine action feature events, a heterogeneous feature event sequence is obtained.

[0046] Specifically, on the molding machine side, the molding machine pressure and temperature data are first separated from the synchronous timing data. For the molding machine pressure data, by analyzing its change curve over a complete working cycle, the time point when the pressure value reaches a local maximum is identified. This time point is the pressure peak moment, which usually corresponds to the critical process node where the material is fully compacted. For the molding machine temperature data, by analyzing its rate of change, the key points where the slope of the temperature curve changes are found, such as the inflection point from the rapid heating stage to the holding stage. This moment is the temperature inflection point. Combining these identified pressure peak moments and temperature inflection point moments constructs the molding process characteristic events that characterize the changes in the core process state. On the automaton side, the automaton position or state data is separated from the synchronous timing data. By continuously monitoring the spatial coordinates of the robotic arm, when the position data of its end effector enters the preset target tolerance range and remains stable, this moment is recorded as the robotic arm end effector arrival moment, marking the completion of a precise positioning operation. Simultaneously, by monitoring the status signals of components such as the grippers, when a signal value undergoes a clear reversal, such as changing from "open" to "closed," this moment is recorded as the gripper state switching moment, representing the execution of a gripping or releasing action. Combining these robotic arm end-effector arrival moments with gripper state switching moments constructs automaton motion feature events characterizing key physical actions. Finally, the molding process feature events and automaton motion feature events generated from the above two branches are merged and sorted on a unified timeline according to their respective timestamps. This results in a completely new data sequence containing a set of events from different devices, describing different physical processes, but strictly ordered in time—this is the final generated heterogeneous feature event sequence.

[0047] Optionally, dividing the alignment analysis window on the time axis includes:

[0048] Obtain the occurrence time of the molding process characteristic event in the heterogeneous characteristic event sequence;

[0049] The time of occurrence is taken as the starting boundary;

[0050] Using the starting boundary, an alignment analysis window is divided on the time axis according to a preset time length.

[0051] Specifically, to divide the timeline into aligned analysis windows, this method first requires utilizing the heterogeneous feature event sequence generated in the previous step. The first step is to traverse this heterogeneous feature event sequence, identify and extract molding process feature events as the analysis benchmark, such as the pressure peak moment or the temperature inflection point moment, and accurately obtain the occurrence time of the event.

[0052] The second step is to directly specify the occurrence time of the acquired molding process characteristic events as the starting boundary of the alignment analysis window. This starting boundary marks a clear starting point on the timeline, and subsequent analysis will revolve around this point in time.

[0053] Third, based on this initial boundary, the system applies a preset time length to determine the range of the window. This preset time length is a duration pre-set based on historical experience or process specifications; it represents the upper limit of the time that the associated automaton actions should complete after a critical event in the molding process occurs. By combining the initial boundary with this preset time length, a range with a clear start and end point can be divided on the timeline; this range is the alignment analysis window. This process can be expressed by the following relationship:

[0054] ,

[0055] in, This represents the final alignment analysis window, which is a closed interval on the time axis. This represents the time of occurrence of the selected molding process feature event, and the value is obtained directly from the timestamp of the corresponding event in the heterogeneous feature event sequence. This represents the preset time length, which is a pre-configured system parameter.

[0056] Optionally, the method further includes:

[0057] Before dividing the alignment analysis window, obtain the alignment event pairs in the historical cycle, and calculate the process stability index based on the historical alignment event pairs;

[0058] The time length is adjusted according to the process stability index.

[0059] Specifically, the first step is to acquire and analyze historical production data. The first step involves accessing alignment event pairs stored in the system's historical records. These alignment event pairs are combinations of successfully matched molding process characteristic events and automated machine action characteristic events from previous production cycles. The second step is to calculate a process stability index based on these historical alignment event pairs. Specifically, for each historical alignment event pair, the time difference between the occurrence time of the automated machine action characteristic event and the occurrence time of the molding process characteristic event is calculated. By statistically analyzing the time difference data obtained from multiple historical cycles, a numerical value quantifying the tightness and consistency of process coordination can be obtained, i.e., the process stability index. A commonly used process stability index is the standard deviation of historical time differences, calculated as follows:

[0060] ,

[0061] in, This represents a process stability index; the smaller the value, the more stable the process timing. This represents the total number of historical alignment event pairs used for calculation; this value is obtained from statistics in the historical database. This represents the time difference between the occurrence time of the automaton action feature event and the occurrence time of the molding process feature event in the i-th historical alignment event pair. This value is obtained by subtracting the timestamps in the historical alignment event pair. Represents all historical time differences The arithmetic mean of the time parameters. All variables involved in the calculation in this formula are either time-based or dimensionless. The third step is to adjust the preset time length used to divide the alignment analysis window based on the calculated process stability index S. The principle of adjustment is that when the process stability index... A smaller time difference indicates less fluctuation in the historical coordinated actions and a stable process. In this case, the time length can be appropriately shortened to improve the efficiency and accuracy of subsequent matching. Conversely, when the process stability index... A large time difference indicates that the time difference of historical coordinated actions fluctuates greatly and the process is unstable. In this case, it is necessary to appropriately increase the time length to expand the search range and ensure that automaton action feature events that may have large delays can be captured.

[0062] Optionally, the generation of alignment event pairs includes:

[0063] Within the alignment analysis window, a correlation sliding calculation is performed on the molding process feature events and the automatic machine action feature events to obtain event correlation values;

[0064] The event correlation values ​​are sorted from largest to smallest, and the automaton action feature event corresponding to the first ranked event correlation value is selected to obtain the candidate action feature event.

[0065] The candidate feature events are combined with the molding process feature events to obtain aligned event pairs.

[0066] Specifically, within this alignment analysis window, for each molding process feature event serving as the anchor point, a correlation sliding calculation is performed with each automaton action feature event contained within the window to quantify the strength of their correlation. This calculation aims to find the events that are most closely related in time. For any automaton action feature event within the window, its event correlation value with the anchor molding process feature event can be calculated using the following formula:

[0067] ,

[0068] In this formula, This represents the calculated event correlation value. It is a dimensionless numerical value with a range of 0 to 1. The closer the value is to 1, the stronger the temporal correlation between the two events. The time of occurrence of the characteristic event of the automaton action currently being calculated is also obtained from its timestamp. The scaling factor is a positive value, a pre-set system parameter used to adjust the sensitivity of the time difference to the correlation value. Its dimension is the negative square of time to ensure the exponential part is dimensionless. This calculation is performed on all automaton action feature events within the window one by one. The second step involves collecting all calculated event correlation values ​​and sorting them in descending order. After sorting, the event correlation value at the top of the list, i.e., the largest value, is selected. The automaton action feature event corresponding to this largest event correlation value is identified as a candidate action feature event. The third step combines this selected candidate action feature event with the molding process feature event initially used as the anchor point. This combination constitutes a complete, strongly correlated aligned event pair.

[0069] Optionally, the method further includes:

[0070] After the automatic machine completes its action according to the coordinated control command, the adjusted operating status is collected in the next production cycle to correct the automatic machine's action characteristic events;

[0071] Update the heterogeneous feature event sequence using the corrected automaton action feature events.

[0072] Specifically, after the automaton completes an adjustment action based on the generated coordinated control instructions, this method does not end immediately. Instead, it uses this adjustment as the starting point for a new round of analysis. In the next production cycle, the system restarts the data acquisition process. Specifically, it performs real-time, timestamped synchronous acquisition of the automaton's adjusted operating state, such as the position trajectory of its robotic arm or the state changes of its grippers, obtaining a new set of synchronous timing data reflecting the control effect. Based on this newly acquired synchronous timing data, the system re-executes feature extraction calculations. It analyzes this data to determine new, adjusted key time points of the automaton's actions, such as the new end-effector position time or gripper state switching time. This newly determined event is the corrected automaton action feature event, which objectively records the actual action timing after the implementation of the coordinated control instructions. Finally, it locates the heterogeneous feature event sequence of the current production cycle. It uses this newly generated corrected automaton action feature event, reflecting the actual execution effect, to replace or update the original, unadjusted, or previously predicted automaton action feature events in the sequence. Through this operation, the heterogeneous feature event sequence is updated in real time, enabling it to accurately record the actual collaborative state after the control command takes effect.

[0073] Optionally, the method further includes:

[0074] Obtain the product quality inspection results corresponding to the same production cycle as the alignment event;

[0075] The alignment event pairs are associated with the product quality inspection results to establish a quality traceability data chain.

[0076] Specifically, obtain the product quality inspection results. Specifically, for the products produced within the same production cycle as a certain alignment event pair, the system needs to retrieve the quality inspection data of the product from the downstream quality inspection station or system. This process relies on a unique production cycle identifier, such as a production batch number or a product serial number, to ensure a one-to-one correspondence between the process described by the alignment event pair and the product quality inspection results. The second step is to associate the obtained product quality inspection results with the corresponding alignment event pair. This is a data structuring process where the system creates a new data record that contains two core pieces of information. One is the alignment event pair representing the process execution timing, which details the precise coordination of the key molding process feature events and the automaton action feature events in this production cycle. The other is the product quality inspection result representing the final output quality, such as the dimensional accuracy, surface finish, or the determination of whether the product is qualified. By binding these two information entities in the same data record and using the production cycle identifier as an index, the system successfully connects the process parameters with the result data. After completing this operation for a series of production cycles, the collection of these associated records constitutes the quality traceability data chain.

[0077] Based on the same inventive concept, as Figure 3 shown, the present invention also provides a data fusion and alignment system for a molding press and an automaton, and the system includes:

[0078] A data acquisition module, configured to acquire the operating states of the molding press and the automaton, and generate synchronized timing data by combining a unified time reference;

[0079] A feature recognition module, configured to extract key time points in the molding process to form molding process feature events based on the synchronized timing data, and extract key time points in the automaton action process to form automaton action feature events, and obtain a heterogeneous feature event sequence by combining the molding process feature events and the automaton action feature events;

[0080] An alignment matching module, configured to divide an alignment analysis window on the time axis with the molding process feature events as anchor points based on the heterogeneous feature event sequence, and match the molding process feature events with one or more automaton action feature events within the alignment analysis window to generate alignment event pairs;

[0081] A control generation module, configured to calculate a timing deviation parameter based on the time relationship between the corresponding events in the alignment event pairs, and generate a collaborative control instruction for adjusting the operation of the automaton according to the timing deviation parameter.

[0082] To verify the feasibility of this invention in practice, it was applied to a high-precision optical lens production line. This production line is dedicated to producing high-quality aspherical glass lenses. Its core processes include hot pressing preforms using a glass molding machine (GMP), and automated operations such as loading, unloading, and transfer are completed by multi-axis automated machines (such as six-axis robots).

[0083] In traditional production models, the coordinated operation of molding machines and automated machines relies primarily on fixed delay parameters and simple I / O signal interactions. After long-term operation, the mechanical response time can drift due to wear, temperature changes, and other factors, rendering the fixed delays no longer optimal. Secondly, when changing to different lens mold specifications or adjusting process parameters, experienced engineers must spend considerable time manually readjusting the coordination time, resulting in low efficiency and difficulty in ensuring consistency. Finally, when problems such as lens breakage or deterioration in surface quality occur, it is difficult to trace whether the issue stems from abnormal molding processes or improper timing of the automated machine's pick-up and drop (such as grabbing the lens before it has fully cooled).

[0084] In this embodiment, the production line utilizes the method proposed in this invention to intelligently upgrade the glass molding machine (hereinafter referred to as the molding machine) and the six-axis automatic machine responsible for loading and unloading (hereinafter referred to as the automatic machine). The specific implementation process and verification results of this invention will be described in detail below.

[0085] To achieve data fusion and alignment between the molding machine and the automated machine, a Network Time Protocol (NTP) client is deployed in both the PLC control system of the molding machine and the robot controller of the automated machine. This client periodically synchronizes with the same NTP time server within the production line's local area network. The NTP server serves as a shared time synchronization unit, ensuring that both control systems generate unified timestamps with millisecond-level accuracy. Using this unified timestamp, the operating status of the molding machine (mold cavity pressure, upper / lower mold temperature, mold position) and the operating status of the automated machine (end-effector TCP coordinates, joint angles, end-gripper opening / closing status signals) are synchronously collected at a frequency of 100Hz via an industrial Ethernet protocol (such as Profinet or EtherCAT). The final result obtained in this way is synchronized time-series data, a collection of data points, each containing the operating status of both the molding machine and the automated machine at the same moment, providing a high-fidelity data foundation for subsequent analysis.

[0086] On the molding machine side, the molding machine pressure and temperature data are first separated from the synchronous timing data. By analyzing the pressure curve, the moment when the pressure value reaches the preset molding pressure is identified and denoted as the "pressure peak moment," which marks the final determination of the lens shape. By analyzing the temperature curve, the moment when the temperature drops and crosses the glass transition temperature (Tg point) is identified and denoted as the "temperature inflection point moment," which is crucial to the influence of internal stress in the lens. Combining these "pressure peak moments" and "temperature inflection point moments" constructs the molding process characteristic events that characterize the changes in the core process state.

[0087] On the automaton side, by continuously monitoring the TCP coordinates of the robotic arm's end effector, when it moves within a preset tolerance range (e.g., ±0.1mm) of the material handling position of the molding machine, this moment is recorded as the "robotic arm end effector arrival moment". Simultaneously, by monitoring the sensor signals of the gripper, when the signal value changes from "0" (open) to "1" (closed), this moment is recorded as the "gripper state switching moment". Combining these "robotic arm end effector arrival moments" and "gripper state switching moments" constructs the automaton motion characteristic events that characterize key physical actions.

[0088] Finally, the molding process feature events and the automatic machine action feature events generated by the two branches are merged and sorted on a unified time axis according to their respective timestamps to obtain a time-ordered heterogeneous feature event sequence containing events such as "pressure peak", "temperature inflection point", "robotic arm in place", and "gripper closure".

[0089] The "temperature inflection point" (representing the moment when the lens has cooled to a safe gripping temperature) is used as the anchor point for the molding process characteristic event. The system uses the occurrence time of this anchor point as the time... As the starting boundary, and based on the material taking time distribution obtained from historical data statistics, a preset time length is set. An alignment analysis window is defined on the timeline, with a time interval of 5 seconds. .

[0090] Within this window, the system searches for the most relevant automaton action. This includes the "gripper state transition moment (closing)" that should occur afterward. The system determines the optimal match through correlation sliding calculation, using the event correlation value. Through formula Calculation, where It is the moment when a "gripper closure" event occurs within the window. It is a preset scale factor. The system will select the "gripper closure" event with the highest event correlation value, combine it with the "temperature inflection point time" to generate an aligned event pair, for example {event: "temperature inflection point", time: T1} <-> {event: "gripper closure", time: T2}.

[0091] Based on the alignment event pairs, the system calculates the timing deviation parameters. This deviation This reflects whether the automatic machine's material handling action is ahead of or behind the ideal process time. The system compares this to a preset optimal target time delay. Comparison, optimal target latency Based on experience, a time of 1.5 seconds is set to ensure that the lens is sufficiently cooled but not overcooled.

[0092] The deviation control generation module generates a collaborative control instruction. For example, if the calculated timing deviation parameter is 2.0 seconds, which is 0.5 seconds slower than the target value, the system will generate an instruction to adjust the trigger delay of the relevant task in the automatic machine controller, causing it to start the material picking action 0.5 seconds earlier in the next production cycle. In the next cycle, the system will collect data again to verify the adjustment effect and update the heterogeneous feature event sequence with the corrected automatic machine action feature events, forming a closed-loop feedback system for continuous learning and optimization.

[0093] In addition, the system will associate the alignment event pairs of each production cycle with the quality inspection results of the lenses produced in that cycle (e.g., the PV value of the lens surface shape error measured by an interferometer), and establish a complete quality traceability data chain using the lens's unique ID as an index.

[0094] To verify the beneficial effects of this invention, a one-month comparative test was conducted on the production line. One production line was selected as the experimental group, deploying the method described in this invention; another production line producing the same lenses was selected as the control group, using the traditional fixed-delay collaborative method. Data shows that, through the system of this invention, the production collaboration efficiency and product quality stability of the experimental group were significantly improved.

[0095] As shown in Table 1, in the experimental group, the timing deviation (standard deviation) between key events in the molding process and the actions of the automated machine decreased significantly from the initial 280ms and stabilized at around 35ms, indicating extremely high consistency in the coordinated actions. This reduced the average production cycle time from 45.2 seconds in the control group to 42.5 seconds, improving efficiency by approximately 6%.

[0096] Table 1 Comparison of Production Cycle and Timing Deviation

[0097] index Control group (traditional method) Experimental group (applying this invention) Improvement / Optimization Average production cycle 45.2 seconds 42.5 seconds Efficiency improved by 6.0% Timing deviation (standard deviation) 280 milliseconds 35 milliseconds Accuracy improved by 87.5% Initial time setting Approximately 2-3 hours (manual) Approximately 15 minutes (automatic learning) Debugging efficiency improved by >90%

[0098] As shown in Table 2, the more precise timing of material retrieval avoids quality problems caused by retrieving materials too early (when the lens is not sufficiently cooled, which can easily lead to thermal stress) or too late (when the lens is too cold, which may cause impurities to adhere to the surface). The product yield (the proportion of the surface PV value that is better than the standard) in the experimental group increased from 95.5% to 98.2%, demonstrating significant improvement.

[0099] Table 2 Comparison of Product Yield Rate

[0100] index Control group (traditional method) Experimental group (using this invention) Improvement effect Yield before testing 95.3% 95.5% - Yield after testing 95.6% 98.2% The yield rate improved by 2.6 percentage points. Defect rate due to timing issues Approximately 2.1% Approximately 0.4% The defect rate decreased by 81%.

[0101] As shown in Table 3, in the third week of testing, a batch of lenses in the experimental group exhibited abnormal PV values. Through the quality traceability data chain, the system quickly located the corresponding "alignment event pair" for this batch, where the timing deviations of the "temperature inflection point" and "gripper closure" showed a clustered large deviation. Further investigation revealed that this was caused by a slow response of a joint motor in the automated machine. The entire traceability and diagnostic process took only 10 minutes, while in the control group, similar problems typically required several hours of manual investigation.

[0102] Table 3 Comparison of Quality Traceability Cases

[0103] project Control group (traditional method) Experimental group (using this invention) Advantages Comparison Anomaly location method Manually check process parameters and equipment logs Automatically associate alignment event pairs with quality data Automation, precision Diagnosis time Average 2.5 hours Average 10 minutes Speed ​​increased by 15 times Root cause analysis Relying on engineers' experience may lead to misjudgment. Data-driven approach directly addresses the root causes of timing deviations. Objective and reliable

[0104] As can be seen from the data in Tables 1 to 3 above, after applying the method of this invention, the experimental group experienced significant improvements in the coordination accuracy, production efficiency, and product quality between the equipment. The system can automatically adapt to minor changes in equipment status and continuously optimize coordination time, transforming the complex debugging process that originally required manual intervention into an autonomous, self-optimizing intelligent system. These data fully demonstrate the advanced nature, effectiveness, and practical value of this invention in solving the coordination problem between molding machines and automated machines in the field of optical lens processing.

[0105] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0106] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method of die press and automation data fusion and alignment, the method comprising: The method includes: The system collects the operating status of the molding machine and the automatic machine, and combines this data with a unified time reference to generate synchronized time-series data. Based on the synchronous time series data, key time points in the molding process are extracted to form molding process feature events, and key time points in the automatic machine operation process are extracted to form automatic machine operation feature events. Combining the molding process feature events and the automatic machine operation feature events, a heterogeneous feature event sequence is obtained. Based on the heterogeneous feature event sequence, taking the molding process feature event as the anchor point, an alignment analysis window is divided on the time axis, and within the alignment analysis window, the molding process feature event is matched with one or more automatic machine action feature events to generate alignment event pairs; Based on the time relationship between corresponding events in the alignment event pair, the timing deviation parameter is calculated, and a cooperative control command for adjusting the operation of the automaton is generated according to the timing deviation parameter.

2. A method of die press and transfer machine data fusion and alignment as claimed in claim 1, wherein, The generation of synchronized time-series data includes: In the control systems of molding machines and automatic machines, a unified timestamp is generated through a shared time synchronization unit; Using the unified timestamp, the operating status of the molding machine and the automatic machine are synchronously collected to obtain synchronous time-series data.

3. The method of claim 1, wherein, The obtained heterogeneous feature event sequence includes: Based on the synchronous time-series data, the pressure data of the molding machine is obtained to generate the pressure peak time, and the temperature data of the molding machine is obtained to generate the temperature inflection point time. Using the pressure peak time and the temperature inflection point time, the characteristic events of the molding process are constructed. Based on the synchronous timing data, the robot arm end-effector arrival time or gripper state switching time generated by the automaton position or state data is obtained, and automaton action feature events are constructed. By combining the molding process feature events with the automatic machine action feature events, a heterogeneous feature event sequence is obtained.

4. The method for data fusion and alignment between a molding machine and an automatic machine according to claim 1, characterized in that, The process of dividing the alignment analysis window on the time axis includes: Obtain the occurrence time of the molding process characteristic event in the heterogeneous characteristic event sequence; The time of occurrence is taken as the starting boundary; Using the starting boundary, an alignment analysis window is divided on the time axis according to a preset time length.

5. The method for data fusion and alignment between a molding machine and an automatic machine according to claim 4, characterized in that, The method further includes: Before dividing the alignment analysis window, obtain the alignment event pairs in the historical cycle, and calculate the process stability index based on the historical alignment event pairs; The time length is adjusted according to the process stability index.

6. The method for data fusion and alignment between a molding machine and an automatic machine according to claim 1, characterized in that, The generated alignment event pair includes: Within the alignment analysis window, a correlation sliding calculation is performed on the molding process feature events and the automatic machine action feature events to obtain event correlation values; The event correlation values ​​are sorted from largest to smallest, and the automaton action feature event corresponding to the first ranked event correlation value is selected to obtain the candidate action feature event. The candidate feature events are combined with the molding process feature events to obtain aligned event pairs.

7. The method for data fusion and alignment between a molding machine and an automatic machine according to claim 1, characterized in that, The method further includes: After the automatic machine completes its action according to the coordinated control command, the adjusted operating status is collected in the next production cycle to correct the automatic machine's action characteristic events; Update the heterogeneous feature event sequence using the corrected automaton action feature events.

8. The method for data fusion and alignment between a molding machine and an automatic machine according to claim 1, characterized in that, The method further includes: Obtain the product quality inspection results corresponding to the same production cycle as the alignment event; The alignment event pairs are associated with the product quality inspection results to establish a quality traceability data chain.

9. A data fusion and alignment system for molding machines and automatic machines, characterized in that, The system includes: The data acquisition module is used to collect the operating status of the molding machine and the automatic machine, and generate synchronous time-series data by combining them with a unified time base. The feature recognition module is used to extract key time points in the molding process based on the synchronous time series data to form molding process feature events, and to extract key time points in the automatic machine operation process to form automatic machine operation feature events. The module combines the molding process feature events and the automatic machine operation feature events to obtain a heterogeneous feature event sequence. The alignment matching module is used to divide the time axis into alignment analysis windows based on the heterogeneous feature event sequence, with the molding process feature event as the anchor point, and match the molding process feature event with one or more automatic machine action feature events within the alignment analysis window to generate alignment event pairs. The control generation module is used to calculate the timing deviation parameter based on the time relationship between corresponding events in the alignment event pair, and generate a cooperative control command for adjusting the operation of the automaton according to the timing deviation parameter.

Images