A product quality sampling method based on deep learning

Abstract

This application discloses a product quality sampling method based on deep learning. The target coefficient is used to characterize the degree of adaptation between workers and production lines, and the degree of adaptation between raw materials or intermediate products and production lines. The target coefficient is applied to the subsequent sampling steps so that this degree of adaptation can play a certain guiding role in the formulation, modification and implementation of the sampling strategy. Then, the sampling results can be matched with the order scale and manual error through technical means, and then the predictive execution of random inspections can be achieved in the generation link. Furthermore, this application also adopts an LSTM prediction model. The LSTM prediction model is trained based on the historical data of the production line. The LSTM prediction model can grasp the equipment conditions of the production line and the adaptation conditions between the equipment and raw materials. Then, the sampling results can be matched with equipment factors and raw material factors through technical means. The resulting sampling strategy is more comprehensive and more in line with the actual situation.

Description

Technical Field

[0001] The present application relates to the field of electronic digital data processing technology, and in particular to a method of digital computing or data processing specifically suitable for specific applications, specifically a product quality sampling method based on deep learning. Background Art

[0002] Various indications indicate that the order size of individual distributors is generally not very large. This leads to the fact that the product quality sampling methods used in related technologies, which rely entirely on statistical laws, can lead to significant errors when performing sampling on small-scale products such as a few hundred or a few thousand. Compared to large-scale orders, smaller orders generally have lower profits. If product quality problems are not discovered in a timely manner during the production process, it may cause significant losses. Moreover, for small-scale orders, real economy companies usually do not build a separate production line for it, but instead rent an existing production line and hire temporary workers. This means that the degree of adaptability between the production line and the product, as well as the workers' familiarity with the production line, will more or less affect product quality, inevitably leading to hidden dangers and requiring more caution in the technical means used in product quality sampling.

[0003] In view of this, how to carry out sampling with predictive capabilities for product quality issues for such small-scale orders has become an urgent issue to be resolved.

[0004] For example, the patent application with publication number CN115543260A, titled "A Method for Random Sampling of Waste Paper Quality Inspection" (Main Classification Number: G06F7 / 58), achieves the informatization and automation of waste paper sampling through the coordination of isolation and sampling rules, ensuring fairness, impartiality, and transparency in waste paper sampling. This demonstrates the potential of digital data processing technology in the field of product sampling and demonstrates the broad potential for technological advancement in this area. Summary of the Invention

[0005] The embodiments of the present application provide a product quality sampling method based on deep learning to at least partially solve the above technical problems.

[0006] The embodiments of this application adopt the following technical solutions:

[0007] In a first aspect, an embodiment of the present application provides a product quality sampling method based on deep learning, the method comprising:

[0008] When it is detected that a target production line has received an order task, a full inspection is performed on a specified number of intermediate products produced by each activated process node of the target production line to obtain a target coefficient corresponding to each process node; the target coefficient is positively correlated with the production efficiency of the corresponding intermediate product and is positively correlated with the yield rate of the intermediate product;

[0009] Using an LSTM prediction model pre-trained based on historical data, the inspection standards used in the order task, the data of the raw materials used, and the number of target products included in the order task, the expected yield of the product corresponding to each process node is predicted;

[0010] When the expected yield of the product is not less than a preset yield threshold, weighting the expected yield of the product by using the target coefficient to obtain the target yield corresponding to each process node;

[0011] In the case that the target yields corresponding to the respective process nodes show a trend of gradual change along the execution sequence of the process steps, a preset sampling strategy is adopted to perform random inspections on the intermediate products and target products.

[0012] In an optional embodiment of this specification, the method further includes:

[0013] When the target yield of a process node shows a sharp downward trend, the process node and the next process node are designated as risky process nodes;

[0014] Improve the sampling rate of the risky process nodes.

[0015] In an optional embodiment of this specification, the method further includes:

[0016] The target coefficient is also negatively correlated with the quantity of target products included in the order task, and negatively correlated with the duration of the intermediate warehousing link in the execution plan process of the order task.

[0017] In an optional embodiment of this specification, the method further includes:

[0018] When the number of the target products is less than the preset first quantity threshold, the target coefficient when the number of the started process nodes is less than the preset second quantity threshold is less than the target coefficient when the number of the started process nodes is not less than the second quantity threshold.

[0019] In an optional embodiment of this specification, the method further includes:

[0020] When the number of target products is not less than the first quantity threshold, the target coefficient when the number of started process nodes is less than the second quantity threshold is greater than the target coefficient when the number of started process nodes is not less than the second quantity threshold.

[0021] In an optional embodiment of this specification, the method further includes:

[0022] The first quantity threshold is positively correlated with the lowest value of the production efficiency.

[0023] In an optional embodiment of this specification, the method further includes:

[0024] In the case that the expected yield of the product is less than a preset yield threshold, a prompt message is issued; the prompt message is used to inform the production plan of the order task to be adjusted.

[0025] In an optional embodiment of this specification, the LSTM prediction model is trained by the following steps:

[0026] Based on historical order tasks, a corresponding sample sequence and sample annotation are constructed; the sample sequence includes a number of sample nodes; the sample nodes correspond one-to-one to the process nodes initiated by the historical order tasks; the data contained in the sample nodes include: the equipment attribute data of the corresponding process node, the raw material attribute data, the applicable testing standards, and the number of products included in the historical order tasks; the sample annotation represents the product yield of the sample node;

[0027] The preset LSTM model is trained based on the sample sequence and sample annotations to obtain the LSTM prediction model.

[0028] In an optional embodiment of this specification, the method further includes:

[0029] The specified number is positively correlated with the number of started process nodes, and negatively correlated with the distance between the key process node started in the target production line and the last process node; the key process node is the process node with the lowest yield in the target production line determined based on historical data.

[0030] In an optional embodiment of this specification, the method further includes:

[0031] The number of products included in the historical order tasks to which the historical data that can be used for sample construction belongs is not less than the preset third quantity threshold;

[0032] Among them, the third quantity threshold is negatively correlated with the number of process nodes in the production line where the historical order task is started; and the third quantity threshold when the historical order task is suspended due to equipment failure is greater than the third quantity threshold when no suspension occurs.

[0033] In a second aspect, an embodiment of the present application further provides a product quality sampling device based on deep learning, which is used to implement the method in the first aspect.

[0034] In a third aspect, an embodiment of the present application further provides an electronic device, including:

[0035] processor; and

[0036] A memory arranged to store computer executable instructions which, when executed, cause the processor to perform the method steps of the first aspect.

[0037] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, which stores one or more programs. When the one or more programs are executed by an electronic device including multiple applications, the electronic device executes the method steps described in the first aspect.

[0038] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects:

[0039] The method in this specification realizes product quality sampling for small-scale orders. During the execution of the method in this application, the degree of adaptation of workers to the production line, and the degree of adaptation of raw materials or intermediate products to the production line are characterized by the target coefficient. The target coefficient is applied to the subsequent sampling step so that this degree of adaptation can play a certain guiding role in the formulation, modification and implementation of the sampling strategy. Then, the sampling results can be matched with the order scale and manual error by technical means, and then it can be achieved in the generation link. Predictive execution of spot checks and timely discovery of problems provide conditions for solving problems. Furthermore, this application also adopts an LSTM prediction model. The LSTM prediction model is trained based on the historical data of the production line. The LSTM prediction model can grasp the equipment status of the production line and the adaptation of the equipment and raw materials. Then, the sampling results can be matched with equipment factors and raw material factors by technical means. The sampling strategy obtained is more comprehensive and more in line with the actual situation. The method of this application realizes the application of electronic digital data processing technology in the product quality sampling method based on deep learning by technical means. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0041] Figure 1 A process diagram of a product quality sampling method based on deep learning provided in an embodiment of this specification;

[0042] Figure 2 This is a schematic diagram of the structure of an electronic device in an embodiment of this specification. DETAILED DESCRIPTION

[0043] The present invention will be further described in detail below with reference to the accompanying drawings by way of specific embodiments. Similar elements in different embodiments are numbered with associated similar elements. In the following embodiments, many detailed descriptions are provided to enable the present application to be better understood. However, those skilled in the art will readily appreciate that some of the features may be omitted under different circumstances, or may be replaced by other elements, materials, or methods. In some cases, some operations related to the present application are not shown or described in the specification. This is to avoid overwhelm the core of the present application with excessive descriptions, and for those skilled in the art, it is not necessary to describe these related operations in detail. They will fully understand the related operations based on the description in the specification and the general technical knowledge in the art.

[0044] In addition, the features, operations, or characteristics described in the specification may be combined in any appropriate manner to form various embodiments. Furthermore, the steps or actions in the method description may be reordered or adjusted in a manner readily apparent to those skilled in the art. Therefore, the various sequences in the specification and drawings are provided solely for the purpose of clearly describing a particular embodiment and are not intended to be mandatory, unless otherwise specified.

[0045] The serial numbers assigned to components herein, such as "first," "second," etc., are used solely to distinguish the objects being described and do not convey any sequential or technical meaning. References to "connection" and "coupling" herein, unless otherwise specified, include both direct and indirect connections (couplings).

[0046] The following describes in detail the technical solutions provided by various embodiments of the present application in conjunction with the accompanying drawings.

[0047] like Figure 1 As shown, the deep learning-based product quality sampling method in this specification includes the following steps:

[0048] S100: When it is detected that a target production line receives an order task, a full inspection is performed on a specified number of intermediate products produced by each activated process node of the target production line to obtain a target coefficient corresponding to each process node.

[0049] The target production line in this manual is the production line used to execute the order. Since the order is small, the number of target products may only be a few thousand or even a few hundred. There's no need to build a separate production line for it. The target production line is rented from another company or used to produce other products.

[0050] The target products included in the order tasks in this specification can be determined according to actual conditions, for example, socks, shoes, underwear, etc. In the case that the target product is underwear, the target production line may include but is not limited to one or more of the following equipment: cutting equipment, sewing equipment, mold cup shaping equipment, sponge processing equipment, punching and trimming equipment, pressing equipment, jacquard equipment, and packaging equipment. There is a corresponding relationship between equipment and process nodes, and a certain process node corresponds to at least one device. Since order tasks have a certain degree of flexibility, it is not necessarily guaranteed that all equipment in the target production line will be enabled during the execution of the order task. For example, if the target product is solid-colored underwear without patterns, the jacquard equipment will not be enabled, and the process node to which it belongs will not be enabled. Unless otherwise specified, the process nodes mentioned below refer to the enabled process nodes.

[0051] In an optional embodiment of this specification, the specified number can be an empirical value. Since the target production line has just started and workers have just begun to carry out production, this is also a period when problems are more likely to be exposed. A full inspection of intermediate products can help to discover problems to a greater extent. The specified number is the number of outputs inspected for each process node. For example, if the specified number is 10, then the first 10 intermediate products or target products output by process node a, process node b, and process node c will all be fully inspected.

[0052] The target coefficients in this specification correspond one-to-one with process nodes. Process nodes correspond one-to-one with intermediate products or target products. The target coefficient is positively correlated with the production efficiency and yield of the corresponding intermediate product. It is a value between 0 and 1 that represents the compatibility between the worker and the production line, and the raw material and the target production line. Production efficiency reflects a worker's proficiency in operating the equipment. Due to the small size of orders, workers are likely to be temporary hires. Since small orders have lower profits, real economy companies are unlikely to hire skilled workers with higher labor costs. Generally, without large-scale, continuous training opportunities, it is difficult for workers to quickly master a particular piece of equipment. For example, jacquard equipment requires a very high level of proficiency. This means that a worker's proficiency in the equipment may not improve significantly and continuously throughout the entire order execution process, so it is necessary to consider the worker's situation. Furthermore, the target coefficient also represents the compatibility between the raw material or an intermediate product and the corresponding equipment. If the compatibility is low, the product yield will naturally be lower. For example, if a garment steamer's temperature is too high or rises too quickly, it can negatively impact synthetic fabrics and discolor silk. Laser cutting can also damage materials containing silk or wool. Because the target production line has historically lacked consistent production of the target product, and the real economy company lacks experience producing the same product as the order, laser cutting can easily result in defects if wool is mixed into the fabric.

[0053] As for the method of determining production efficiency and product yield of full inspection, the technical means in the relevant technology are applicable to this manual when conditions permit.

[0054] In an optional embodiment of the present specification, the target coefficient is also negatively correlated with the number of target products contained in the order task (the fewer the number of target products, the shorter the running-in period between the workers and the target production line, and the more lasting the negative impact of the workers' unskilledness), and is negatively correlated with the duration of the intermediate warehousing link in the execution plan process of the order task (the longer the duration of the warehousing link, the discontinuous operation of the production line, which may cause large errors in the prediction results of the LSTM prediction model).

[0055] Furthermore, in a further optional embodiment of this specification, when the number of target products is less than a preset first threshold (which may be an empirical value), the target coefficient when the number of activated process nodes is less than a preset second threshold (which may be an empirical value) is less than the target coefficient when the number of activated process nodes is not less than the second threshold. Because the samples used in the LSTM prediction model training process are mostly collected for the entire production line, if the production process of the target product involves a small number of devices, on the one hand, there is the possibility that the production situation will differ significantly from the samples used in the training model, resulting in distorted model output; on the other hand, due to the short running-in period between workers and the target production line, quality stability is difficult to achieve. Therefore, greater caution should be exercised.

[0056] If the number of target products is not less than the first threshold, the target coefficient is set when the number of activated process nodes is less than the second threshold, and is greater than the target coefficient when the number of activated process nodes is not less than the second threshold. If the number of target products is not too small, workers and the target production line can achieve a certain degree of adaptation. The fewer the process nodes, the simpler the target products, and the lower the likelihood of potential problems. This shows that adjusting the target coefficient based on the number of target products and process nodes can amplify the adaptation effect between workers and production lines.

[0057] Optionally, the first quantity threshold is positively correlated with the lowest value of the production efficiency. The lower the production efficiency, the less skilled the workers are, which, combined with the small number of target products, can amplify the feature.

[0058] In an optional embodiment of the present specification, the specified number is positively correlated with the number of started process nodes (to detect product quality problems caused by the correlation between process nodes), and is negatively correlated with the distance between the key process node started in the target production line and the last process node (the distance is the number of separated process nodes); the key process node is the process node with the lowest yield in the target production line determined based on historical data (the low yield of the key process node may be caused by the accumulation of problems in the previous process nodes. The later the key process node is, the more likely it is that there are undetected problems in the previous process nodes).

[0059] S102: Using an LSTM prediction model pre-trained based on historical data, and the inspection standards used in the order task, the data of the raw materials used, and the number of target products included in the order task, predict the expected yield of the products corresponding to each process node.

[0060] The LSTM prediction model described in this manual is used to predict the yield of intermediate products or target products at a process node when production on a production line reaches steady state. The LSTM prediction model corresponds one-to-one with each production line. If a production line changes, the model needs to be retrained.

[0061] The LSTM prediction model learns the characteristics of the attribute data of each device included in the target production line (for example, sewing devices a and b are normal, while sewing device c is prone to thread entanglement); the interaction between the characteristics of the raw material attribute data and the characteristics of the equipment when the production line reaches steady state (the production line as a whole reaches the optimal production state and the negative impact of human labor is negligible) (for example, sewing device a is suitable for various specifications of thread, but the diameter of the thread suitable for sewing devices b and c should not exceed XX, otherwise it will malfunction, etc.); and the yield rate under different applicable testing standards (for example, a company's corporate standards, national standards, EU standards, etc.).

[0062] Since the data used in training the LSTM prediction model is mostly based on the production line's own conditions and the coordination between the production line and raw materials, its prediction results are difficult to reflect the impact of human factors.

[0063] LSTM (Long Short-Term Memory) is a special recurrent neural network (RNN) architecture designed primarily for processing sequential data. Although LSTM itself is a human-designed neural network model, its design and training process does incorporate principles and ideas related to natural laws.

[0064] 1. Time series dependencies. Manifestations of natural laws: Many phenomena in nature exhibit time series dependencies. For example, weather changes, biological behavior patterns, and the grammatical structure of language are all influenced by past states. This temporal causal relationship is a key manifestation of natural laws. Manifestations in LSTMs: LSTMs are specifically designed to process sequential data and can capture the dependencies between individual time steps in a sequence. Through recurrent connections, the current state of an LSTM depends not only on the current input but also on the state at the previous moment. This design enables LSTMs to understand temporal dependencies in sequential data, similar to the biological brain. For example, they can understand the grammatical structure of sentences in natural language processing and predict future trends based on historical data in time series forecasting.

[0065] 2. Memory and Forgetting Mechanisms. A natural law manifests itself: When processing information, the biological brain filters it based on its importance, memorizing important information and forgetting unimportant information. This mechanism helps the brain efficiently process and store information, avoiding interference from irrelevant information. Its manifestation in LSTM: The core of LSTM is the cell state, which is similar to the memory unit in the biological brain. LSTM uses gating mechanisms (input gate, output gate, and forget gate) to control the inflow, outflow, and forgetting of information. The forget gate functions similarly to the forgetting mechanism in the biological brain. Based on the current input and the previous state, it determines which information is no longer important and removes it from the cell state. The input gate is responsible for writing new important information into the cell state, similar to the biological brain's memory process of new information.

[0066] 3. Adaptive learning capability. Reflection of natural laws: Biological systems possess strong adaptive learning capabilities, able to adjust their behavior and structure in response to changes in the environment. This ability is a result of biological evolution and a key manifestation of natural laws. This is reflected in LSTMs: During training, LSTMs adjust network weights using the backpropagation algorithm to learn patterns in the data. This learning process is adaptive because the network automatically adjusts weights based on the characteristics of the training data to better fit the data. For example, when processing different types of sequence data, LSTMs automatically adjust the parameters of the gating mechanism to adapt to the characteristics of the data, thereby achieving adaptive learning.

[0067] 4. Information screening and integration. A natural law manifests itself: When processing information, biological systems employ complex screening and integration mechanisms to extract useful information while ignoring irrelevant information. This mechanism helps improve system efficiency and accuracy. This is also evident in LSTMs: The LSTM gating mechanism is essentially a process of information screening and integration. The input gate determines which information is important based on the current input and the previous state, integrating it into the cell state. The forget gate filters out and removes no longer important information. This screening and integration process enables LSTMs to efficiently process sequential data, avoiding interference from irrelevant information and thereby improving model performance.

[0068] 5. Energy and information conversion. Manifestation of natural laws: The conversion of energy and information is ubiquitous in natural systems. For example, organisms consume energy to process and store information, and the transmission and processing of information in turn affects the organism's energy consumption. This is also evident in LSTMs: Although LSTMs are computational models, their training process also involves the conversion of energy and information. During training, computing resources (which can be compared to energy) are used to process training data (information). Optimization algorithms adjust network weights to learn patterns in the data. This energy and information conversion process enables LSTMs, like natural systems, to achieve complex information processing by consuming certain resources.

[0069] 6. Stability and dynamism of complex systems. Natural laws manifest: Complex systems in nature (such as ecosystems and climate systems) exhibit both stability and dynamism. These systems can maintain stability within certain limits while also being able to respond to external changes. This is reflected in LSTMs: The LSTM's cell state and gating mechanism jointly ensure the network's stability and dynamism. Controlled by the gating mechanism, the cell state maintains a certain degree of stability, thereby preventing problems like vanishing gradients. Simultaneously, the dynamic adjustment of the gating mechanism enables the LSTM to flexibly process diverse sequence data. This combination of stability and dynamism enables the LSTM, like complex systems in nature, to maintain stability and adapt to changes when processing sequence data.

[0070] In summary, while the design and training process of the LSTM model is based on artificial algorithms and mathematical principles, it does, in some respects, embody principles and ideas from natural laws. These principles and ideas not only enable LSTM to efficiently process sequential data but also provide insights into the information processing mechanisms of natural systems.

[0071] In the related art, any technical means for training LSTM models, where applicable, are applicable to this specification. In an optional embodiment of this specification, the model training process may include: constructing a corresponding sample sequence and sample annotations based on historical order tasks; (there is a one-to-one correspondence between historical order tasks, sample sequences, and sample annotations) The sample sequence includes several sample nodes; each sample node corresponds one-to-one to the process node initiated by the historical order task; the data contained in the sample node includes: equipment attribute data, raw material attribute data, applicable testing standards, and the number of products included in the historical order task; and the sample annotation represents the product yield of the sample node. A feature vector can be constructed based on this data. Technical means for constructing feature vectors in the related art, where applicable, are applicable to this specification. Optionally, the number of products included in the historical order task to which the historical data used for sample construction belongs is no less than a preset third threshold. The third threshold is negatively correlated with the number of process nodes in the production line initiated by the historical order task, allowing for a more comprehensive evaluation of the entire production line and facilitating the reflection of the coordination between process nodes in the features. Furthermore, the third quantity threshold for when a historical order is paused due to equipment failure is greater than the third quantity threshold for when no pause occurs, allowing the LSTM prediction model to learn the impact of the number of products included in the order task on the yield rate. Because the samples used during model training correspond to historical order tasks with a relatively high number of products, this objectively reflects the true condition of the equipment while also enabling the model to learn the impact of product quantity on the yield rate. This allows the model to learn the impact of product quantity on the yield rate when the model is used offline, as small orders often contain a relatively small number of target products. This can lead to quantity-related confidence differences in the yield rate output by the model, allowing the output to reflect the production line's sensitivity to product quantity.

[0072] The preset LSTM model is trained based on the sample sequence and sample annotations to obtain the LSTM prediction model.

[0073] S104: When the expected yield of the product is not less than a preset yield threshold, the target coefficient is used to weight the expected yield of the product to obtain the target yield corresponding to each process node.

[0074] The number of projected product yields is determined by the number of process nodes. A projected product yield of no less than a preset yield threshold means that the projected yield of the intermediate or target product corresponding to each process node is no less than the preset yield threshold. For example, if the yield of an earlier process node is low, then a higher yield at a subsequent process node will also result in higher costs.

[0075] The weightings used in this specification are one-to-one. For example, the target coefficient for process node A is used to weight the projected product yield for process node A. Because the target coefficient reflects labor requirements and the compatibility of the production line with raw materials, and the projected product yield reflects the production line's inherent conditions and its optimal compatibility with different raw materials, this weighted combination makes the resulting target yield more comprehensive and more accurately reflects the actual achievable yield.

[0076] In an optional embodiment of this specification, if the expected yield of a product is less than a preset yield threshold, a prompt message is issued; the prompt message is used to inform the production plan of the order task to be adjusted. The specific adjustment method can be determined based on manual experience, such as replacing one fabric with another, or changing the original one ironing process to two.

[0077] S106: When the target yields corresponding to the respective process nodes show a trend of gradual change along the execution sequence of the process steps, a preset sampling strategy is used to perform random inspections on the intermediate products and the target products.

[0078] In actual production, it's difficult to rigorously inspect every process node, and some processes may even skip inspection altogether. Especially for smaller orders, the increased costs associated with rigorous spot checks can further reduce profits. The method described in this manual, however, involves limited full inspections (for example, covering a few dozen products), allowing for a comprehensive understanding of each process node early in production. Furthermore, full inspections can demonstrate the effectiveness of the connections between process nodes.

[0079] Gradual change means that the slope of the yield curve nodes (corresponding to process nodes, with process nodes on the horizontal axis and yield on the vertical axis) is no less than a preset slope threshold (an empirical value related to the expected profit margin of the order). If the target yield of each process node shows a gradual trend along the execution sequence, it indicates that human influence is controllable and stable, and does not pose a significant threat to overall production. Otherwise, if the yield of a particular process node continues to decline, showing a smaller slope, it indicates a higher risk at that process node.

[0080] In the related art, strategies that can be used to implement sampling can be used as preset sampling strategies in this specification when conditions permit, such as stratified sampling.

[0081] In an optional embodiment of the present specification, when the target yield of a certain process node shows a trend of sharp decline (the slope is less than the slope threshold, that is, the absolute value of the slope is large but negative), the process node and the next process node are regarded as risky process nodes; the sampling rate of the risky process nodes is increased to timely discover the risk of defective intermediate products and improve the risk monitoring effect.

[0082] The method in this specification implements product quality sampling for small-scale orders. During the execution of the method in this application, the degree of adaptation of workers to the production line, and the degree of adaptation of raw materials or intermediate products to the production line are characterized by the target coefficient, and the target coefficient is applied to the subsequent sampling steps so that this degree of adaptation can play a certain guiding role in the formulation, modification and implementation of the sampling strategy. Then, the sampling results can be matched with the order scale and manual error by technical means, and then it can be achieved in the generation link. Predictive execution of spot checks and timely discovery of problems provide conditions for solving problems. Furthermore, this application also adopts an LSTM prediction model. The LSTM prediction model is trained based on the historical data of the production line. The LSTM prediction model can grasp the equipment status of the production line and the adaptation of the equipment and raw materials. Then, the sampling results can be matched with equipment factors and raw material factors by technical means. The sampling strategy obtained thereby is more comprehensive and more in line with the actual situation.

[0083] Furthermore, this specification also provides a product quality sampling device based on deep learning, which can execute the method in any of the aforementioned embodiments and can achieve the same or similar technical effects, which will not be repeated here.

[0084] Figure 2 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. Figure 2 At the hardware level, the electronic device includes a processor and, optionally, an internal bus, a network interface, and memory. The memory may include internal memory, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk drive. Of course, the electronic device may also include other hardware required for its services.

[0085] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, Figure 2 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0086] The memory is used to store programs. Specifically, the program may include program code, which includes computer operating instructions. The memory may include internal memory and non-volatile memory, and provides instructions and data to the processor.

[0087] The processor reads the corresponding computer program from the non-volatile memory into the internal memory and then runs it, forming a product quality sampling device based on deep learning at the logical level. The processor executes the program stored in the memory and is specifically configured to perform any of the aforementioned deep learning-based product quality sampling methods.

[0088] The above application Figure 1The deep learning-based product quality sampling method disclosed in the illustrated embodiments can be applied to or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, the steps of the above method can be performed by hardware integrated logic circuits in the processor or by software instructions. The processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The methods, steps, and logic block diagrams disclosed in the embodiments of this application can be implemented or executed. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly executed by a hardware decoding processor or by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method.

[0089] The electronic device may also perform Figure 1 A product quality sampling method based on deep learning and implementation Figure 1 The functions of the illustrated embodiment will not be described in detail in the embodiments of the present application.

[0090] An embodiment of the present application also proposes a computer-readable storage medium, which stores one or more programs, and the one or more programs include instructions. When the instructions are executed by an electronic device including multiple applications, any one of the aforementioned deep learning-based product quality sampling methods is executed.

[0091] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0092] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0093] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0094] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0095] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0096] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0097] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can be implemented using any method or technology to store information. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change RAM (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media such as modulated data signals and carrier waves.

[0098] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0099] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0100] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A product quality sampling method based on deep learning, characterized in that: The method comprises: When it is detected that a target production line has received an order task, a full inspection is performed on a specified number of intermediate products produced by each activated process node of the target production line to obtain a target coefficient corresponding to each process node; the target coefficient is positively correlated with the production efficiency of the corresponding intermediate product and is positively correlated with the yield rate of the intermediate product; Using an LSTM prediction model pre-trained based on historical data, the inspection standards used in the order task, the data of the raw materials used, and the number of target products included in the order task, the expected yield of the product corresponding to each process node is predicted; When the expected yield of the product is not less than a preset yield threshold, weighting the expected yield of the product by using the target coefficient to obtain the target yield corresponding to each process node; In the case that the target yields corresponding to the respective process nodes show a trend of gradual change along the execution sequence of the process steps, a preset sampling strategy is adopted to perform random inspections on the intermediate products and target products.

2. The method according to claim 1, wherein: The method further comprises: When the target yield of a process node shows a sharp downward trend, the process node and the next process node are designated as risky process nodes; Improve the sampling rate of the risky process nodes.

3. The method according to claim 1, wherein: The method further comprises: The target coefficient is also negatively correlated with the quantity of target products included in the order task, and negatively correlated with the duration of the intermediate warehousing link in the execution plan process of the order task.

4. The method according to claim 1, wherein: The method further comprises: When the number of the target products is less than the preset first quantity threshold, the target coefficient when the number of the started process nodes is less than the preset second quantity threshold is less than the target coefficient when the number of the started process nodes is not less than the second quantity threshold.

5. The method according to claim 4, wherein: The method further comprises: When the number of target products is not less than the first quantity threshold, the target coefficient when the number of started process nodes is less than the second quantity threshold is greater than the target coefficient when the number of started process nodes is not less than the second quantity threshold.

6. The method according to claim 4, wherein: The method further comprises: The first quantity threshold is positively correlated with the lowest value of the production efficiency.

7. The method according to claim 1, wherein: The method further comprises: In the case that the expected yield of the product is less than a preset yield threshold, a prompt message is issued; the prompt message is used to inform the production plan of the order task to be adjusted.

8. The method according to claim 1, wherein: The LSTM prediction model is trained through the following steps: Based on historical order tasks, a corresponding sample sequence and sample annotation are constructed; the sample sequence includes a number of sample nodes; the sample nodes correspond one-to-one to the process nodes initiated by the historical order tasks; the data contained in the sample nodes include: the equipment attribute data of the corresponding process node, the raw material attribute data, the applicable testing standards, and the number of products included in the historical order tasks; the sample annotation represents the product yield of the sample node; The preset LSTM model is trained based on the sample sequence and sample annotations to obtain the LSTM prediction model.

9. The method according to claim 8, wherein: The method further comprises: The number of products included in the historical order tasks to which the historical data that can be used for sample construction belongs is not less than the preset third quantity threshold; Among them, the third quantity threshold is negatively correlated with the number of process nodes in the production line where the historical order task is started; and the third quantity threshold when the historical order task is suspended due to equipment failure is greater than the third quantity threshold when no suspension occurs.

10. The method according to claim 1, wherein: The method further comprises: The specified number is positively correlated with the number of started process nodes, and negatively correlated with the distance between the key process node started in the target production line and the last process node; the key process node is the process node with the lowest yield in the target production line determined based on historical data.

Images