Method, system and equipment for dynamically adjusting number of spare parts and storage medium

Abstract

The invention relates to the technical field of server equipment, and particularly provides a dynamic adjustment method and system for the number of spare parts, equipment and a storage medium, and the method comprises the steps: planning detection points corresponding to all parts on the equipment in advance, continuously collecting operation data through a sensor disposed at the detection points, and obtaining an operation data sequence of all the parts; inputting the operation data sequence of each component into a pre-trained long-short-term memory network model to obtain an operation time sequence feature of each component; basic information of equipment is acquired, and the basic information is quantized into feature vectors; the operation time sequence features of all the components and the feature vectors are input into a pre-trained support vector machine, a fault prediction result is obtained, and the fault prediction result comprises abnormal components; and the types and the corresponding number of the abnormal parts are counted, and the inventory number of the corresponding spare parts is dynamically adjusted based on the statistical result. The situation that the number of spare parts is unreasonable due to inaccurate estimation is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of server equipment, and in particular relates to a method, system, equipment and storage medium for dynamically adjusting the number of spare parts. Background Art

[0002] Equipment spare parts are core materials to ensure production continuity, and their inventory management has a direct impact on the company's operating costs and production efficiency. However, traditional spare parts management models mostly rely on experience judgment or a single quantitative storage strategy, which often leads to unnecessary inventory backlogs or emergency shortages.

[0003] Although some spare parts management methods try to predict inventory demand by statistically analyzing historical consumption data, this guidance method based on the law of consumption quantity is still too mechanical and fails to fully consider the actual operation needs of the equipment. Therefore, this management method is also prone to cause inventory backlogs or shortages. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method, system, device and storage medium for dynamically adjusting the number of spare parts to solve the above-mentioned technical problems.

[0005] In a first aspect, the present invention provides a method for dynamically adjusting the number of spare parts, comprising: Plan detection points corresponding to each component on the equipment in advance, use sensors deployed at the detection points to continuously collect operation data, and obtain operation data sequences of each component; Input the running data sequence of each component into the pre-trained long short-term memory network model to obtain the running sequence characteristics of each component; Obtain basic information of the device, including manufacturer information, model information, and usage time, and quantify the basic information into a feature vector; Inputting the running timing characteristics of each component and the characteristic vector into a pre-trained support vector machine to obtain a fault prediction result, wherein the fault prediction result includes abnormal components; The types and corresponding quantities of abnormal parts are counted, and the inventory quantity of corresponding spare parts is dynamically adjusted based on the statistical results.

[0006] In an optional implementation, detection points corresponding to each component are pre-planned on the device, and operating data is continuously collected using sensors deployed at the detection points to obtain an operating data sequence of each component, including: Generate component numbers for components of the device in advance, and use the component numbers to name corresponding operation data storage files; Pre-set the effective monitoring time of various components, and based on the effective monitoring time, use the time window technology to intercept the operation data sequence from the corresponding operation data storage file; Normalize the operation data series of each component respectively; The dynamic time warping algorithm is used to convert the operating data sequences of each component into operating data sequences of the same length.

[0007] In an optional implementation, the support vector machine training method includes: Construct all possible two-class classifiers in k-class samples. Each two-class classifier is trained with only two training samples in k-classes, and a total of k(k-1) / 2 two-class classifiers are constructed. When constructing the mth classifier among k classifiers (m∈y i ) The training samples of the mth class are labeled as a class y i =1 All training samples except class m are regarded as a class and labeled as y i m = -1, after optimization, the classification output function of the mth classifier can be established as: ; First, input the test data sample x into the classifier if the discriminant It is determined that the category 1 test is completed.

[0008] In an optional implementation, the types and corresponding quantities of abnormal components are counted, and the inventory quantity of corresponding spare parts is dynamically adjusted based on the statistical results, including: Record the inventory quantity of spare parts as the actual quantity, and the quantity of abnormal parts of the corresponding type as the required quantity; If the actual quantity exceeds the required quantity, the difference between the actual quantity and the required quantity is calculated, and the difference and the types of spare parts are uploaded to the data sharing platform, and other plant terminals borrow the spare parts through the data sharing platform; If the actual quantity does not exceed the required quantity, a corresponding spare parts replenishment task is generated, and based on the spare parts replenishment task, corresponding spare parts are borrowed from the sharing platform or a spare parts purchase order is generated.

[0009] In an optional embodiment, the method further comprises: Extracting the replenishment quantity of target spare parts from the spare parts replenishment task, where the replenishment quantity is the difference between the required quantity and the actual quantity; Query the supply data that matches the replenishment quantity of the target spare parts from the data sharing platform, and extract the supplier's location information and transaction costs from the supply data; Calculate logistics cost and logistics time based on the manufacturer's location information and local location, and use the sum of logistics cost and transaction cost as the loan cost; Determine whether the logistics time exceeds the set time threshold: If yes, a purchase order for spare parts is generated; If not, determine whether the loan cost exceeds the purchase cost. If the loan cost exceeds the purchase cost, generate a spare parts purchase order. If the loan cost does not exceed the purchase cost, execute the loan transaction through the data sharing platform.

[0010] In a second aspect, the present invention provides a system for dynamically adjusting the number of spare parts, including: An operation monitoring module is used to pre-plan detection points corresponding to each component on the equipment, and continuously collect operation data using sensors deployed at the detection points to obtain an operation data sequence of each component; The feature extraction module is used to input the operation data sequence of each component into the pre-trained long short-term memory network model to obtain the operation sequence characteristics of each component; An information processing module is used to obtain basic information of the device, including manufacturer information, model information, and usage time, and quantify the basic information into a feature vector; A fault prediction module, used to input the running sequence characteristics of each component and the feature vector into a pre-trained support vector machine to obtain a fault prediction result, wherein the fault prediction result includes abnormal components; The quantity adjustment module is used to count the types and corresponding quantities of abnormal parts, and dynamically adjust the inventory quantity of corresponding spare parts based on the statistical results.

[0011] In an optional embodiment, the operation monitoring module includes: A data naming unit, used to generate component numbers for components of the equipment in advance, and use the component numbers to name corresponding operation data storage files; A sequence interception unit, used to pre-set the effective monitoring time of various components, and intercept the operation data sequence from the corresponding operation data storage file using the time window technology based on the effective monitoring time; A normalization processing unit, used for respectively performing normalization processing on the operation data sequence of each component; The length processing unit is used to convert the operation data sequence of each component into an operation data sequence of the same length by using a dynamic time warping algorithm.

[0012] In an optional implementation, the support vector machine training method includes: Construct all possible two-class classifiers in k-class samples. Each two-class classifier is trained with only two training samples in k-classes, and a total of k(k-1) / 2 two-class classifiers are constructed. When constructing the mth classifier among k classifiers (m∈y i ) The training samples of the mth class are labeled as a class y i =1 All training samples except class m are regarded as a class and labeled as y i m = -1, after optimization, the classification output function of the mth classifier can be established as: ; First, input the test data sample x into the classifier if the discriminant It is determined that the category 1 test is completed.

[0013] In a third aspect, a device is provided, comprising: A memory, used for storing a dynamic adjustment program of the number of spare parts; A processor is used to implement the steps of the method for dynamically adjusting the number of spare parts as provided in the first aspect when executing the dynamic adjustment program for the number of spare parts.

[0014] In a fourth aspect, a computer-readable storage medium is provided, on which a dynamic adjustment program for the number of spare parts is stored. When the dynamic adjustment program for the number of spare parts is executed by a processor, the steps of the dynamic adjustment method for the number of spare parts provided in the first aspect are implemented.

[0015] The beneficial effect of the present invention is that the method, system, device and storage medium for dynamically adjusting the number of spare parts provided by the present invention accurately predict the actual demand through a specific fault prediction model, dynamically adjust the inventory quantity based on the actual spare parts quantity demand, and effectively avoid inventory backlog and shortage problems. Compared with the traditional management method based on manual experience, it reduces the situation of unreasonable spare parts quantity caused by inaccurate estimation.

[0016] In addition, the invention has a reliable design principle, a simple structure and a very broad application prospect. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 is a schematic flow chart of a method according to an embodiment of the present invention.

[0019] Figure 2 is a schematic block diagram of a system according to an embodiment of the present invention.

[0020] Figure 3 A schematic diagram of the structure of a device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0023] The key terms appearing in the present invention are explained below.

[0024] Long Short-Term Memory (LSTM) is a carefully designed recurrent neural network (RNN) architecture that aims to overcome the gradient vanishing and gradient exploding problems faced by traditional RNNs when processing long sequence data. In the hierarchical structure of LSTM, each time step receives an input signal and the cell state from the previous time step. Through the fine-tuned forget gate, input gate, and output gate mechanism, LSTM can intelligently decide which information should be forgotten, which new information should be included, and which information should be passed to the hidden state. This process allows the hidden state to gradually accumulate and reflect the local feature information of the sequence.

[0025] In order to accurately capture these local features, we focus on the hidden state of the LSTM layer at a specific time step. This hidden state not only incorporates the information of the current time step, but also includes the accumulated information of the previous time step. Thanks to the unique gating mechanism of LSTM, it can effectively capture the key local features in the sequence. Therefore, this hidden state can be used as a local feature representation of the data at that position.

[0026] Local feature extraction based on LSTM has the following significant advantages: Excellent ability to capture sequence information: LSTM can keenly capture long-term dependencies in sequence data. Its hidden state is rich in sequence information, providing a solid foundation for the extraction of local features.

[0027] Adaptive feature extraction: With its unique gating structure, LSTM can flexibly extract key features in the sequence while effectively filtering out irrelevant information, thereby significantly improving the accuracy of feature extraction.

[0028] High flexibility: LSTM can process sequence data of different lengths, and its network structure and parameters can be flexibly adjusted according to the requirements of specific tasks and data sets, showing wide applicability.

[0029] Therefore, the LSTM model can capture the timing characteristics of equipment operation data.

[0030] Support Vector Machine (SVM) is a supervised learning model first proposed by Corinna Cortes, Vapnik and others in 1995. It has many unique advantages in solving small sample, nonlinear and high-dimensional pattern recognition, and can be extended to other machine learning problems such as function fitting.

[0031] The basic idea of ​​SVM is to find a hyperplane to segment the samples. This hyperplane can maximize the interval between data points belonging to two different classes, that is, to find the maximum interval hyperplane. In order to deal with nonlinear problems, SVM introduces the kernel function method to map the vectors in the low-dimensional space to the high-dimensional space, thereby constructing the maximum interval hyperplane in the high-dimensional space. This method overcomes the problems of dimensionality curse and nonlinear separability without increasing the complexity of calculation.

[0032] Support vector machines have the following advantages when dealing with equipment fault diagnosis: Using small sample data sets for diagnosis: The SVM algorithm can perform effective machine learning in small sample conditions, which makes it particularly suitable for equipment fault diagnosis because equipment fault data is often difficult to obtain in large quantities.

[0033] Based on the principle of minimum structural risk: The SVM algorithm is based on the principle of structural risk minimization and is easy to achieve near-optimal model structure and parameters, thereby improving the accuracy and reliability of fault diagnosis.

[0034] Strong generalization ability: The SVM algorithm has good generalization ability and can handle unknown or new fault types, which is of great significance for the uncertainty problem in equipment fault diagnosis.

[0035] The computational overhead is low: Although the SVM algorithm requires complex calculations when processing high-dimensional data, since its final decision function is determined by only a few support vectors, it has certain advantages in computational overhead compared to other machine learning algorithms.

[0036] The results are easy to explain: The results of the SVM algorithm are easy to interpret, and support vectors can be analyzed to understand which features have a significant impact on the fault diagnosis results, which helps engineers better understand the causes and mechanisms of equipment failures.

[0037] The method for dynamically adjusting the number of spare parts provided in the embodiment of the present invention is executed by a computer device. Accordingly, the system for dynamically adjusting the number of spare parts runs in the computer device.

[0038] Figure 1 is a schematic flow chart of a method according to an embodiment of the present invention. Figure 1 The execution subject may be a dynamic adjustment system for the number of spare parts. According to different requirements, the order of the steps in the flow chart may be changed, and some may be omitted.

[0039] like Figure 1 As shown, the method includes: S1. Plan detection points corresponding to each component on the equipment in advance, and use sensors deployed at the detection points to continuously collect operation data to obtain operation data sequences of each component.

[0040] During the initial configuration of the equipment, technicians need to accurately plan the inspection points corresponding to each key component on its structure in advance. These inspection points are precisely laid out to ensure that they can fully cover the operating conditions of the equipment. Subsequently, high-performance sensors are deployed at each planned inspection point. These sensors have continuous and high-precision data collection capabilities, and can continuously capture and record real-time data of each component during operation, thereby forming a series of detailed operating data sequences, providing a solid foundation for subsequent analysis.

[0041] S2. Input the operation data sequence of each component into the pre-trained long short-term memory network model to obtain the operation sequence characteristics of each component.

[0042] The collected operation data sequence of each component is used as input and passed to a pre-trained long short-term memory network (LSTM) model. With its unique structural design, the LSTM model can effectively capture the time dependency and long-term trend in the data, thereby accurately extracting the operation sequence characteristics of each component. These characteristics not only reflect the current working status of the component, but also contain its historical operation mode and potential change trend.

[0043] S3. Obtain basic information of the device, including manufacturer information, model information, and usage time, and quantify the basic information into a feature vector.

[0044] In order to more comprehensively assess the failure risk of equipment, it is necessary to obtain its basic information from the equipment management system, including but not limited to the manufacturer, specific model, and length of time in use. This information is essential for understanding the inherent characteristics and expected life of the equipment. Subsequently, quantitative technology is used to convert this unstructured basic information into feature vectors, which accurately express the basic properties of the equipment in numerical form, providing a key supplement for subsequent analysis.

[0045] S4. Input the running timing characteristics of each component and the characteristic vector into a pre-trained support vector machine to obtain a fault prediction result, which includes abnormal components.

[0046] The running timing features of each component extracted from the LSTM model are combined with the feature vector of the quantified basic information of the equipment and passed as input to a pre-optimized support vector machine (SVM) model. With its powerful classification ability and generalization performance in high-dimensional space, the SVM model can comprehensively evaluate the health status of each component and output accurate fault prediction results. These results clearly indicate which components are abnormal, providing clear guidance for taking maintenance measures in a timely manner.

[0047] S5. Count the types and corresponding quantities of abnormal parts, and dynamically adjust the inventory quantity of corresponding spare parts based on the statistical results.

[0048] Based on the fault prediction results output by the SVM model, a detailed statistical analysis is performed on the types and quantities of abnormal parts. This step not only helps identify the types of parts with frequent failures, but also reveals potential equipment maintenance needs. On this basis, enterprises can dynamically adjust the inventory quantity of corresponding spare parts based on statistical results to ensure that they can respond quickly when equipment fails and replace damaged parts in a timely manner, thereby minimizing losses caused by downtime, while optimizing inventory management and avoiding unnecessary waste of resources.

[0049] In an embodiment of the present invention, based on step S1, a possible embodiment is given below to illustrate its specific implementation scheme in a non-limiting manner.

[0050] By installing various sensors on the equipment, such as electrical sensors, temperature sensors, vibration sensors, pressure sensors, etc., the operating data of the equipment is collected.

[0051] In order to facilitate subsequent data processing, the data collected by the sensor needs to be preprocessed: S101. Generate component numbers for components of the device in advance, and use the component numbers to name corresponding operation data storage files.

[0052] In the process of equipment maintenance and management, it is crucial to assign a unique part number to each key component. This number can not only serve as the identity of the component, making it easy to quickly locate the specific component in subsequent maintenance and data analysis, but also serve as the basis for naming the operation data storage file. By using the part number to name the corresponding operation data storage file, the data organization structure can be ensured to be clear and easy to manage. In this way, when you need to query or analyze the operation data of a certain component, you only need to quickly find the corresponding file based on the part number, which greatly improves the efficiency of data processing.

[0053] S102. Preset the effective monitoring time of various components, and based on the effective monitoring time, use the time window technology to intercept the operation data sequence from the corresponding operation data storage file.

[0054] In a specific example, there is an industrial equipment that contains multiple key components, such as motors, sensors, and controllers. In order to monitor the operating status of these components, sensors are installed on the equipment and a data acquisition system is set up to continuously record the operating data of the components. Now, this data is used to predict the failure of the components so that maintenance can be carried out in time.

[0055] Setting of effective monitoring time Motor: The motor is the core component of the equipment, and its operating status directly affects the performance of the entire equipment. According to the working characteristics and operating cycle of the motor, the effective monitoring time of the motor is set to the first 8 hours of operation every day (assuming that the equipment runs 12 hours a day, the last 4 hours are in a low-load state, and the data reference value is low).

[0056] Sensors: Sensors are used to monitor various parameters of equipment, such as temperature, pressure, and vibration. Since sensors need to work continuously, their effective monitoring time is set to 24 hours a day.

[0057] Controller: The controller is responsible for the control and scheduling of the equipment, and its operating status is crucial to the stability and efficiency of the equipment. Set the effective monitoring time of the controller to the entire operating time of the day (i.e. 12 hours).

[0058] Application of time window technology After setting the effective monitoring time, use the time window technology to intercept the operation data sequence from the operation data storage file. The following are the specific steps: Data preparation: First, export the operating data of each component from the data acquisition system and store them as files in chronological order. For example, the operating data of the motor can be stored as "motor operating data_YYYYMMDD.csv", where YYYY represents year, MM represents month, and DD represents date.

[0059] Set time window: Set the time window for each component according to the set effective monitoring time. For example, for motors, set the time window to 00:00 to 08:00 every day; for sensors and controllers, set the time window to all day.

[0060] Data capture: Use programming languages ​​(such as Python) or data processing software (such as Excel) to read the operation data file and capture the required data segment according to the set time window. For example, for the motor, read the "Motor Operation Data_YYYYMMDD.csv" file every day and capture the data from 00:00 to 08:00.

[0061] S103. Normalize the operation data sequence of each component respectively.

[0062] Data preprocessing: Preprocess the intercepted data, such as removing invalid data, filling missing values, and performing normalization to ensure the quality and consistency of the data.

[0063] S104. Use a dynamic time warping algorithm to convert the operating data sequences of each component into operating data sequences of the same length.

[0064] Select a suitable target length, that is, the length you want all sequences to have after conversion. This length can be determined based on specific requirements, such as wanting all sequences to have the same number of sampling points, or to have the same length as a specific sequence.

[0065] The DTW algorithm is commonly used to compute the similarity between two time series, but is slightly repurposed here to align the series to the same length. However, it is important to note that DTW itself does not directly change the length of the series; instead, it provides an alignment path that can be used to resample or interpolate the series to reach a target length.

[0066] For each pair of sequences to be aligned, the scipy.spatial.distance.cdist function is used to calculate the DTW distance matrix between them and implement the logic of resampling or interpolation. The DTW distance matrix records the cumulative distance of all possible paths from one point in the sequence to another.

[0067] Find the optimal path from the upper left corner to the lower right corner in the DTW distance matrix (i.e. the path with the smallest cumulative distance). This path represents the best alignment between the two sequences.

[0068] Resample or interpolate the sequence using the found optimal alignment path. This can be done by selecting or computing a corresponding value at each point on the path. For example, the values ​​of the original sequence can be interpolated based on the positions of the points on the path in the new sequence.

[0069] Example Results Through the above steps, the operation data sequence of each component within the effective monitoring time is obtained. These data sequences include both the normal operation status of the component (such as stable operation of the motor, normal readings of the sensor, etc.), and may also include abnormal or fault conditions (such as overheating of the motor, abnormal readings of the sensor, etc.). Next, these data sequences are used to train and test the fault prediction model in order to detect and predict the fault conditions of the component in a timely manner.

[0070] In an embodiment of the present invention, based on step S2, a possible embodiment is given below to illustrate its specific implementation scheme in a non-limiting manner.

[0071] First, set the network structure and parameters of the LSTM model: Number of layers: The number of layers in the LSTM network can be selected based on the complexity of the task and the size of the dataset. Generally, deeper networks are able to capture more complex features, but may also lead to overfitting.

[0072] Number of hidden units: The number of hidden units determines the size of the hidden state in the LSTM layer. A larger number of hidden units can capture more information, but it also increases the complexity and computational cost of the model.

[0073] Learning rate and optimizer: The choice of learning rate and optimizer is crucial for training LSTM networks. A smaller learning rate may lead to a slow training process, while a larger learning rate may lead to unstable training. Commonly used optimizers include SGD, Adam, etc.

[0074] Then, execute the training strategy and regularization: Batch size and number of iterations: The choice of batch size and number of iterations affects the stability and efficiency of the training process. Smaller batch sizes can reduce memory usage and increase training speed, but may lead to unstable training. More iterations can improve model performance, but also increase computational costs.

[0075] Regularization method: To prevent overfitting, regularization methods such as dropout and L2 regularization are used. Dropout can randomly discard the output of a part of the hidden units during the training process, thereby reducing the model's dependence on training data. When selecting evaluation indicators, it is necessary to consider the characteristics and requirements of the task. For example, in classification tasks, indicators such as accuracy and recall are used to evaluate the performance of the model. When deploying the trained LSTM model to practical applications.

[0076] Convert the input sequence (such as text, time series data, etc.) into a numerical form so that the LSTM network can process it. This is usually achieved through word embedding or other feature extraction techniques. Word embedding is a method of converting words or phrases into fixed-length vectors that capture the semantic relationship between words. After the numerical representation of each element is combined into a sequence, it can be used as the input of the LSTM network. The LSTM network processes each element in the order of the sequence and updates its cell state and hidden state.

[0077] In an embodiment of the present invention, based on step S3, a possible embodiment is given below to illustrate its specific implementation scheme in a non-limiting manner.

[0078] In a specific example, the process of converting basic device information into a feature vector is as follows: S301. Obtaining basic device information Manufacturer: Find the manufacturer information of the equipment from the equipment management system. For example, the manufacturer of a certain equipment is "XYZ Industrial Equipment Company".

[0079] Specific models: Similarly, the model of the device can also be found in the management system. For example, the model of the device is "ABC-1234".

[0080] Time in use: The management system usually has the installation date or commissioning date of the equipment. By calculating the difference between the current date and the installation date, you can get the length of time the equipment has been in use. For example, the equipment has been in use for 3 years and 6 months.

[0081] Quantitative Technology and Information Conversion Manufacturer quantification: To convert the manufacturer information into a numerical form, a unique identifier (ID) is assigned to each manufacturer. For example, the ID for "XYZ Industrial Equipment Company" might be 1.

[0082] If more refined quantification is required, consider additional information such as the manufacturer's reputation, market share, etc., and assign corresponding weights and values ​​to these factors.

[0083] Specific model quantification: The device model can also be converted to a numerical form, for example, by assigning a unique ID to each model, or by generating a hash value as part of the feature vector based on the combination of letters and numbers in the model.

[0084] Additionally, the association between the model and the failure rate is considered and a risk score is assigned to the model based on this association.

[0085] Quantification of the time in use: The length of time in use can be directly expressed as a numerical value, for example, 3 years and 6 months can be converted to 3.5 years (or 42 months).

[0086] In order to express the use of equipment more precisely, consider dividing the duration into different intervals and assigning a value or weight to each interval. For example, divide the service life of the equipment into "early", "middle" and "late" stages, and assign different risk factors to each stage.

[0087] Feature vector construction After obtaining basic information such as the manufacturer, specific model, and length of time the equipment has been in use, and performing quantitative processing, this information is combined into a feature vector. For example: Feature vector: [1, hash value of ABC-1234, 3.5] Or, if factors such as the manufacturer's reputation and market share are taken into account, the feature vector may be more complex: Feature vector: [1, reputation score of XYZ Industrial Equipment Company, market share of XYZ Industrial Equipment Company, risk score of ABC-1234, 3.5, risk factor of equipment use phase].

[0088] In an embodiment of the present invention, based on step S4, a possible embodiment is given below to illustrate its specific implementation scheme in a non-limiting manner.

[0089] Support vector machine (SVM) is a binary classification model based on supervised learning, but it can also be extended to multi-classification problems. Its training process mainly includes the following steps: Construct a two-class classifier: In order to distinguish any two types of samples among k types of samples, it is necessary to construct all possible two-type classifiers.

[0090] Each two-class classifier is trained using only two training samples from the k classes.

[0091] Therefore, for k types of samples, a total of k(k-1) / 2 two-class classifiers need to be constructed.

[0092] Train two classifiers: When constructing the mth classifier (m belongs to the category set y i ), take the training sample of the mth class as the positive class (the class label is y i =1).

[0093] The training samples of all categories except the mth category are regarded as negative categories (the category label is y i m =-1).

[0094] Through an optimization algorithm (such as the sequential minimum optimization algorithm SMO), the optimal hyperplane (or decision boundary) is found to maximize the classification interval of the two types of samples on the hyperplane.

[0095] After optimization, the classification output function of the mth classifier can be established. This function is usually a linear function or a linear combination of kernel functions, which is used to determine the category of the test data sample. The specific classification output function is: ; Classification of test data samples: Input the test data sample x into all constructed two-class classifiers.

[0096] For each classifier, the category of the test data sample x is determined according to the discriminant (i.e., the value of the classification output function).

[0097] If the value of the discriminant is greater than a certain threshold (usually 0), it is judged as category 1 (i.e., positive class); otherwise, it is judged as category -1 (i.e., negative class).

[0098] In multi-classification problems, a "one-to-many" or "one-to-one" strategy is usually used to determine the final category of the test data sample. For the "one-to-many" strategy, the category corresponding to the classifier with the largest discriminant value is selected as the final category; for the "one-to-one" strategy, the final category is determined by a voting mechanism or by comparing the difference in discriminant values.

[0099] Processing of multi-classification problems: Although SVM was originally designed for binary classification problems, it can be extended to multi-classification problems through the "one-vs-many" or "one-vs-one" strategies mentioned above.

[0100] In the fault prediction of the present invention, it is necessary to classify and predict according to different fault types (ie, multiple categories) of the equipment. At this time, multiple two-class classifiers can be constructed to distinguish different fault types, and the final fault type can be determined by comparing the discriminant value strategy.

[0101] Model optimization and evaluation: During the training process, the performance of the model can be optimized by adjusting the SVM parameters (such as penalty parameter C, kernel function parameters, etc.).

[0102] The generalization ability of the model is evaluated through methods such as cross-validation to ensure that the model can also perform well on unseen data.

[0103] The trained SVM model is applied to the actual equipment fault prediction, and the health status and potential fault type of the equipment are judged according to the output results of the model.

[0104] Combined with the basic information of the equipment (such as manufacturer, model, usage time, etc.) and operating sequence characteristics, the prediction results are interpreted and analyzed to provide decision support for equipment maintenance.

[0105] In an embodiment of the present invention, based on step S5, a possible embodiment is given below to illustrate its specific implementation in a non-limiting manner.

[0106] In a specific example, the following is a detailed spare parts inventory management and allocation process: S501. Records of inventory quantity and demand quantity: First, the inventory quantity of spare parts is recorded as the actual quantity, which is the actual stock of spare parts currently in the warehouse.

[0107] At the same time, the number of abnormal parts of the corresponding type needs to be recorded, which represents the current demand for spare parts.

[0108] S502. Comparison of inventory quantity and demand quantity: Next, the actual quantity is compared with the required quantity.

[0109] If the actual quantity exceeds the required quantity, it means that the current inventory is sufficient to meet the demand and there is surplus available for allocation.

[0110] If the actual quantity does not exceed the required quantity, it means that the inventory is insufficient and spare parts need to be replenished.

[0111] S503. Allocation process when inventory is sufficient: When the inventory is sufficient, calculate the difference between the actual quantity and the required quantity. This difference represents the number of spare parts that can be allocated.

[0112] Then, this difference and the type of spare parts are uploaded to the data sharing platform so that other plant terminals can borrow these spare parts.

[0113] Other factory terminals can view the available spare parts information through the data sharing platform and borrow them according to their own needs.

[0114] S504. Replenishment process when inventory is insufficient: When the inventory is insufficient, the corresponding spare parts replenishment tasks are generated.

[0115] From this replenishment task, extract the replenishment quantity of the target spare parts, which is the difference between the required quantity and the actual quantity.

[0116] Next, the supply data that matches the target spare parts replenishment quantity is queried from the data sharing platform. These supply data include the location information and transaction costs of the supply manufacturers.

[0117] Calculate logistics cost and time based on supplier location information and local location.

[0118] Then, the sum of logistics cost and transaction cost is taken as the loaned cost for cost assessment.

[0119] S505. Comparison between loan cost and purchase cost: In the cost evaluation stage, the first step is to determine whether the logistics time exceeds the set time threshold.

[0120] If the logistics time exceeds the time threshold, it means that the time cost of borrowing spare parts is too high, and purchasing is more likely to be chosen.

[0121] If the logistics time does not exceed the time threshold, the loan cost is further compared with the procurement cost.

[0122] If the borrowing cost exceeds the purchasing cost, choose purchasing as well.

[0123] If the borrowing cost does not exceed the purchasing cost, the borrowing transaction will be executed through the data sharing platform to borrow spare parts from the supplier.

[0124] S506. Generation of purchase order: When purchasing is selected, a spare parts purchase order is generated and sent to the appropriate supplier for purchase.

[0125] This process describes in detail the various steps of spare parts inventory management and allocation, including the recording and comparison of inventory quantities, allocation and replenishment processes, cost evaluation and procurement decisions, etc., which helps companies better manage spare parts inventories and ensure the continuous operation of production lines.

[0126] In some embodiments, the system for dynamically adjusting the number of spare parts may include a plurality of functional modules composed of computer program segments. The computer program of each program segment in the system for dynamically adjusting the number of spare parts may be stored in a memory of a computer device and executed by at least one processor to perform (see Figure 1 Description) The function of dynamically adjusting the number of spare parts.

[0127] In this embodiment, the system for dynamically adjusting the number of spare parts can be divided into multiple functional modules according to the functions it performs, such as Figure 2 As shown. The functional modules of the system 200 may include: an operation monitoring module 210, a feature extraction module 220, an information processing module 230, a fault prediction module 240 and a quantity adjustment module 250. The module referred to in the present invention refers to a series of computer program segments that can be executed by at least one processor and can complete fixed functions, which are stored in a memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0128] An operation monitoring module is used to pre-plan detection points corresponding to each component on the equipment, and continuously collect operation data using sensors deployed at the detection points to obtain an operation data sequence of each component; The feature extraction module is used to input the operation data sequence of each component into the pre-trained long short-term memory network model to obtain the operation sequence characteristics of each component; An information processing module is used to obtain basic information of the device, including manufacturer information, model information, and usage time, and quantify the basic information into a feature vector; A fault prediction module, used to input the running sequence characteristics of each component and the feature vector into a pre-trained support vector machine to obtain a fault prediction result, wherein the fault prediction result includes abnormal components; The quantity adjustment module is used to count the types and corresponding quantities of abnormal parts, and dynamically adjust the inventory quantity of corresponding spare parts based on the statistical results.

[0129] Optionally, as an embodiment of the present invention, the operation monitoring module includes: A data naming unit, used to generate component numbers for components of the equipment in advance, and use the component numbers to name corresponding operation data storage files; A sequence interception unit, used to pre-set the effective monitoring time of various components, and intercept the operation data sequence from the corresponding operation data storage file using the time window technology based on the effective monitoring time; A normalization processing unit, used for respectively performing normalization processing on the operation data sequence of each component; The length processing unit is used to convert the operation data sequence of each component into an operation data sequence of the same length by using a dynamic time warping algorithm.

[0130] Optionally, as an embodiment of the present invention, the training method of the support vector machine includes: Construct all possible two-class classifiers in k-class samples. Each two-class classifier is trained with only two training samples in k-classes, and a total of k(k-1) / 2 two-class classifiers are constructed. When constructing the mth classifier among k classifiers (m∈y i ) The training samples of the mth class are labeled as a class y i =1 All training samples except class m are regarded as a class and labeled as y i m = -1, after optimization, the classification output function of the mth classifier can be established as: ; First, input the test data sample x into the classifier if the discriminant It is determined that the category 1 test is completed.

[0131] Optionally, as an embodiment of the present invention, counting the types and corresponding quantities of abnormal components, and dynamically adjusting the inventory quantity of corresponding spare parts based on the statistical results, includes: Record the inventory quantity of spare parts as the actual quantity, and the quantity of abnormal parts of the corresponding type as the required quantity; If the actual quantity exceeds the required quantity, the difference between the actual quantity and the required quantity is calculated, and the difference and the types of spare parts are uploaded to the data sharing platform, and other plant terminals borrow the spare parts through the data sharing platform; If the actual quantity does not exceed the required quantity, a corresponding spare parts replenishment task is generated, and based on the spare parts replenishment task, corresponding spare parts are borrowed from the sharing platform or a spare parts purchase order is generated.

[0132] Optionally, as an embodiment of the present invention, it further includes: Extracting the replenishment quantity of target spare parts from the spare parts replenishment task, where the replenishment quantity is the difference between the required quantity and the actual quantity; Query the supply data that matches the replenishment quantity of the target spare parts from the data sharing platform, and extract the supplier's location information and transaction costs from the supply data; Calculate logistics cost and logistics time based on the manufacturer's location information and local location, and use the sum of logistics cost and transaction cost as the loan cost; Determine whether the logistics time exceeds the set time threshold: If yes, a purchase order for spare parts is generated; If not, determine whether the loan cost exceeds the purchase cost. If the loan cost exceeds the purchase cost, generate a spare parts purchase order. If the loan cost does not exceed the purchase cost, execute the loan transaction through the data sharing platform.

[0133] Figure 3 The dynamic adjustment method of the number of spare parts provided for the embodiment of the present application can be applied to equipment. It will be appreciated by those skilled in the art that the equipment structure involved in the embodiment of the present invention does not constitute a limitation on the equipment, and the equipment may include more or less components than shown, or combine certain components, or arrange different components. In an embodiment of the present invention, the equipment includes but is not limited to laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The equipment may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples, and are not intended to limit the implementation of the embodiments of the present application described and / or required herein.

[0134] The device 300 may include: a processor 310, a memory 320 and a communication unit 330. These components communicate via one or more buses. Those skilled in the art will appreciate that the server structure shown in the figure does not limit the present invention, and it may be a bus structure or a star structure, and may include more or fewer components than shown in the figure, or combine certain components, or arrange components differently.

[0135] The memory 320 may be used to store the execution instructions of the processor 310, and the memory 320 may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. When the execution instructions in the memory 320 are executed by the processor 310, the device 300 is enabled to perform some or all of the steps in the following method embodiments.

[0136] The processor 310 is the control center of the storage device, and uses various interfaces and lines to connect various parts of the entire electronic device. It runs or executes software programs and / or modules stored in the memory 320, and calls data stored in the memory to perform various functions of the electronic device and / or process data. The processor can be composed of an integrated circuit (IC), for example, it can be composed of a single packaged IC, or it can be composed of a plurality of packaged ICs with the same or different functions. For example, the processor 310 can include only a central processing unit (CPU). In an embodiment of the present invention, the CPU can be a single computing core or multiple computing cores.

[0137] The communication unit 330 is used to establish a communication channel so that the storage device can communicate with other devices, receive user data sent by other devices or send user data to other devices.

[0138] The present invention also provides a computer storage medium, wherein the computer storage medium may store a program, and when the program is executed, the program may include some or all of the steps in each embodiment provided by the present invention. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0139] Those skilled in the art can clearly understand that the technology in the embodiments of the present invention can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution in the embodiments of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, and other media that can store program codes, including several instructions for enabling a computer device (which can be a personal computer, a server, or a second device, a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention.

[0140] In this specification, the same or similar parts between the various embodiments can be referred to each other. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description in the method embodiment.

[0141] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of systems or modules, which can be electrical, mechanical or other forms.

[0142] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0143] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0144] Although the present invention has been described in detail with reference to the accompanying drawings and in combination with the preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, a person of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and these modifications or substitutions shall be within the scope of the present invention. Any person of ordinary skill in the art may easily think of changes or substitutions within the technical scope disclosed by the present invention, and these shall be within the scope of protection of the present invention.

Claims

1. A method for dynamically adjusting the number of spare parts, characterized in that: include: Plan detection points corresponding to each component on the equipment in advance, and use sensors deployed at the detection points to continuously collect operation data to obtain operation data sequences of each component; Input the running data sequence of each component into the pre-trained long short-term memory network model to obtain the running sequence characteristics of each component; Obtain basic information of the device, including manufacturer information, model information, and usage time, and quantify the basic information into a feature vector; Inputting the running timing characteristics of each component and the characteristic vector into a pre-trained support vector machine to obtain a fault prediction result, wherein the fault prediction result includes abnormal components; The types and corresponding quantities of abnormal parts are counted, and the inventory quantity of corresponding spare parts is dynamically adjusted based on the statistical results.

2. The method according to claim 1, characterized in that Plan detection points corresponding to each component on the equipment in advance, use sensors deployed at the detection points to continuously collect operation data, and obtain operation data sequences of each component, including: Generate component numbers for components of the device in advance, and use the component numbers to name corresponding operation data storage files; Pre-set the effective monitoring time of various components, and based on the effective monitoring time, use the time window technology to intercept the operation data sequence from the corresponding operation data storage file; Normalize the operation data series of each component respectively; The dynamic time warping algorithm is used to convert the operating data sequences of each component into operating data sequences of the same length.

3. The method according to claim 1, characterized in that: The training method of the support vector machine includes: Construct all possible two-class classifiers in k-class samples. Each two-class classifier is trained with only two training samples in k-classes, and a total of k(k-1) / 2 two-class classifiers are constructed. When constructing the mth classifier among k classifiers (m∈y i ) The training samples of the mth class are labeled as a class y i =1 All training samples except class m are regarded as a class and labeled as y i m = -1, after optimization, the classification output function of the mth classifier can be established as: ； First, input the test data sample x into the classifier if the discriminant It is determined that the category 1 test is completed.

4. The method according to claim 1, characterized in that: Count the types and corresponding quantities of abnormal parts, and dynamically adjust the inventory quantity of corresponding spare parts based on the statistical results, including: Record the inventory quantity of spare parts as the actual quantity, and the quantity of abnormal parts of the corresponding type as the required quantity; If the actual quantity exceeds the required quantity, the difference between the actual quantity and the required quantity is calculated, and the difference and the types of spare parts are uploaded to the data sharing platform, and other plant terminals borrow the spare parts through the data sharing platform; If the actual quantity does not exceed the required quantity, a corresponding spare parts replenishment task is generated, and based on the spare parts replenishment task, corresponding spare parts are borrowed from the sharing platform or a spare parts purchase order is generated.

5. The method according to claim 4, characterized in that The method further comprises: Extracting the replenishment quantity of target spare parts from the spare parts replenishment task, where the replenishment quantity is the difference between the required quantity and the actual quantity; Query the supply data that matches the replenishment quantity of the target spare parts from the data sharing platform, and extract the supplier's location information and transaction costs from the supply data; Calculate logistics cost and logistics time based on manufacturer location information and local location, and use the sum of logistics cost and transaction cost as the loan cost; Determine whether the logistics time exceeds the set time threshold: If yes, a purchase order for spare parts is generated; If not, determine whether the loan cost exceeds the purchase cost. If the loan cost exceeds the purchase cost, generate a spare parts purchase order. If the loan cost does not exceed the purchase cost, execute the loan transaction through the data sharing platform.

6. A system for dynamically adjusting the number of spare parts, characterized in that: include: An operation monitoring module is used to pre-plan detection points corresponding to each component on the equipment, and continuously collect operation data using sensors deployed at the detection points to obtain an operation data sequence of each component; The feature extraction module is used to input the operation data sequence of each component into the pre-trained long short-term memory network model to obtain the operation sequence characteristics of each component; An information processing module is used to obtain basic information of the device, including manufacturer information, model information, and usage time, and quantify the basic information into a feature vector; A fault prediction module, used to input the running sequence characteristics of each component and the feature vector into a pre-trained support vector machine to obtain a fault prediction result, wherein the fault prediction result includes abnormal components; The quantity adjustment module is used to count the types and corresponding quantities of abnormal parts, and dynamically adjust the inventory quantity of corresponding spare parts based on the statistical results.

7. The method according to claim 6, characterized in that The operation monitoring module comprises: A data naming unit, used to generate component numbers for components of the equipment in advance, and use the component numbers to name corresponding operation data storage files; A sequence interception unit, used to pre-set the effective monitoring time of various components, and intercept the operation data sequence from the corresponding operation data storage file using the time window technology based on the effective monitoring time; A normalization processing unit, used for respectively performing normalization processing on the operation data sequence of each component; The length processing unit is used to convert the operation data sequence of each component into an operation data sequence of the same length by using a dynamic time warping algorithm.

8. The method according to claim 6, characterized in that The training method of the support vector machine includes: Construct all possible two-class classifiers in k-class samples. Each two-class classifier is trained with only two training samples in k-classes, and a total of k(k-1) / 2 two-class classifiers are constructed. When constructing the mth classifier among k classifiers (m∈y i ) The training samples of the mth class are labeled as a class y i =1 All training samples except class m are regarded as a class and labeled as y i m = -1, after optimization, the classification output function of the mth classifier can be established as: ； First, input the test data sample x into the classifier if the discriminant It is determined that the category 1 test is completed.

9. A device, characterized in that: include: A memory, used for storing a dynamic adjustment program of the number of spare parts; A processor is used to implement the steps of the method for dynamically adjusting the number of spare parts as described in any one of claims 1 to 5 when executing the dynamic adjustment program for the number of spare parts.

10. A computer-readable storage medium storing a computer program, characterized in that: The readable storage medium stores a program for dynamically adjusting the number of spare parts. When the program for dynamically adjusting the number of spare parts is executed by the processor, the steps of the method for dynamically adjusting the number of spare parts according to any one of claims 1 to 5 are implemented.

Images