Intelligent factory equipment predictive maintenance system and method based on Internet of Things

Abstract

The invention discloses an intelligent factory equipment predictive maintenance system and method based on the Internet of Things, and relates to the technical field of processing equipment predictive maintenance, and the method comprises the following steps: collecting parameters of current intelligent factory equipment, building an equipment dual-index health curve model based on the attributes of different equipment, and building an intelligent factory equipment dual-index health curve model; collecting to-be-processed product order data, and establishing a to-be-processed product order database; based on historical processing data of the current intelligent factory, establishing a process parameter rule base, and according to order data in a to-be-processed product order database, performing quantitative analysis on a to-be-processed order of the current intelligent factory based on the process parameter rule base; and based on processing requirement quantitative analysis results of different to-be-processed orders and a double-index health curve model of different processing devices. According to the invention, through quantitative analysis of the multi-dimensional equipment health model and the order, predictive maintenance of the equipment and efficient matching of the processing equipment and the order are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of predictive maintenance of processing equipment, in particular to an intelligent factory equipment predictive maintenance system and method based on Internet of Things. BACKGROUND

[0002] In the intelligent manufacturing system, stable operation of processing equipment and efficient matching of orders are the core links to ensure the production efficiency of the intelligent factory. With the popularization of Internet of Things technology in the factory, it is more convenient to obtain equipment operation data.

[0003] Traditional processing equipment maintenance adopts periodic maintenance or after-failure maintenance mode. Periodic maintenance does not consider individual differences of equipment. Regardless of the actual wear and error state of the equipment, it is maintained and repaired according to a fixed cycle, which is easy to cause over-maintenance, resulting in waste of resources and interruption of production. After-failure maintenance needs to wait for the equipment to fail before processing, which not only affects the delivery of orders due to too long downtime, but also may exacerbate equipment wear and increase maintenance costs due to fault propagation. In addition, the traditional order allocation lacks a linkage mechanism with the health state of the equipment, and is mostly matched according to the equipment capacity or order priority, without considering the influence of the state of the equipment on the order processing requirements, which has the problems of low practicability and functionality. SUMMARY

[0004] In view of the problems in the related art, the present application proposes an intelligent factory equipment predictive maintenance system and method based on Internet of Things to overcome the above technical problems existing in the prior art.

[0005] To this end, the specific technical solutions adopted by the present application are as follows:

[0006] The intelligent factory equipment predictive maintenance method based on Internet of Things comprises the following steps:

[0007] S1, collecting current intelligent factory equipment parameters, establishing a double-index health curve model based on different properties of the equipment itself, and collecting product order data to be processed to establish a product order database to be processed;

[0008] S2, based on the historical processing data of the current intelligent factory, establishing a process parameter rule library, and based on the order data in the product order database to be processed, quantitatively analyzing the current intelligent factory order to be processed based on the process parameter rule library;

[0009] S3, based on the quantitative analysis results of different orders to be processed and the double-index health curve model of different processing equipment, predicting the wear trend and error accuracy change of the equipment involved in the current order, and performing predictive maintenance and equipment order re-matching on the order equipment.

[0010] As a preferred implementation, the S1 comprises the following steps:

[0011] S11, respectively installing sensors for different kinds of processing equipment in the intelligent factory to collect processing data during the operation of the processing equipment, and respectively constructing device double-index health curve models for different kinds of processing equipment, including a benchmark wear curve model and a benchmark error precision curve model;

[0012] S12, establishing a to-be-processed product order database and a processing equipment database through MySQL, filing and recording different processing equipment related data in the processing equipment database, collecting to-be-processed product order data and recording in the to-be-processed product order database, the to-be-processed product order data including product name, material type and hardness value of the processed part, processing steps, dimensional accuracy requirement, shape accuracy requirement, surface roughness requirement, production quantity, and delivery time.

[0013] As a preferred implementation, the S11 comprises the following steps:

[0014] S111, extracting the collected data of different processing equipment in the N time period after maintenance at the device maintenance time point as a cycle, including device state data and work load data, wherein the device state data is spindle motor power, current, torque feedback, and the work load data includes spindle speed, feed speed, cutting depth, cutting width, material type and hardness value of the processed part, and single processed part processing cycle, the benchmark wear curve model of the current processing equipment is obtained through principal component analysis combined with self-encoder for data processing, and a long short-term memory network, and the specific steps are as follows:

[0015] For the extracted feature vector X = [x1, x2, x3,..., x m ] T , each feature is standardized, wherein x ij , μ j , σ j represent the jth feature value of the ith sample, the mean of the jth feature, and the standard deviation of the jth feature, respectively, and z ij represents the standardized value, and the standardized data matrix is denoted as Z and the covariance matrix C is obtained through calculation, wherein n is the number of samples;

[0016] The eigenvalue decomposition is performed on the covariance matrix C to obtain eigenvalues λ1≥λ2≥λ3≥…≥λ m and eigenvectors ω1, ω2, ω3,..., ω m corresponding to the eigenvalues, the first k principal components are selected so that the contribution rate of the cumulative variance CV(k) reaches 95%, wherein The standardized data Z is projected to the subspace spanned by the eigenvectors corresponding to the first k principal components to obtain the dimension-reduced data matrix Y = ZW k , wherein W k is a matrix composed of the first k eigenvectors.

[0017] The input Y is mapped to a low-dimensional bottleneck layer representation h = f enc (Y) by an encoder, and the bottleneck layer representation h is reconstructed to the output Y' = f dec (h) by a decoder, the input Y obtained after dimension reduction by principal component analysis is obtained by the reconstructed output Y' of the trained autoencoder, and the health index HI = ||Y-Y'|| 2 is obtained. The HI(t) points varying with time t are obtained for different processing equipment, and the baseline wear curve model of the current processing equipment is obtained by training of the long short-term memory network.

[0018] S112, taking the equipment maintenance time point as a period, the data of the workpieces of different processing equipment in the N time period after maintenance are collected and extracted, including the size tolerance, shape tolerance, surface roughness, material type and hardness of the workpiece, and the single workpiece processing cycle, which are processed by principal component analysis combined with the autoencoder, and the baseline error precision curve model of the current processing equipment is obtained by combining the long short-term memory network.

[0019] As a preferred embodiment, S21, based on the processing data of historical orders, a process parameter rule library is established, based on the order data in the order database of the product to be processed, the processing requirements of the current intelligent factory order are quantitatively analyzed by the process parameter rule library, and the spindle motor power, current, torque feedback, spindle speed, feed speed, cutting depth, cutting width, and single workpiece processing cycle in the current order processing process are obtained according to the product name, specification and model, processing steps, size accuracy requirement, shape accuracy requirement, surface roughness requirement, production quantity, and delivery time in the order of the product to be processed.

[0020] S22, based on the size accuracy requirement, shape accuracy requirement, and surface roughness requirement of the current product to be processed order, the size tolerance, shape tolerance, and surface roughness of the current product to be processed are obtained according to the mechanical processing standard specification.

[0021] As a preferred embodiment, the S3 comprises the following steps:

[0022] S31, quantitatively analyze the processing requirements of different orders to be processed, and obtain the baseline wear curve and the baseline error accuracy curve of different processing equipment for the current order to be processed by combining the double-index health curve model of different processing equipment, and pre-determine the processing equipment according to the error accuracy percentage in the baseline error accuracy curve and the processing requirements of the current order to be processed;

[0023] S32, based on the use time point of the pre-determined processing equipment, obtain the wear degree of the current equipment by combining the baseline wear curve, determine the wear degree of the equipment by combining the equipment maintenance threshold, and based on the determination result, combine the error accuracy percentage of the equipment to predictively maintain and order rematch the processing equipment in combination with the processing requirements of the remaining orders to be processed.

[0024] As a preferred embodiment, the S31 comprises the following steps:

[0025] S311, according to the quantitative analysis results of the processing requirements of different orders to be processed, substitute the characteristic data of the current order to be processed into the baseline error accuracy curve model of different equipment, obtain the error accuracy percentage output in the baseline error accuracy curve of different equipment for the current order to be processed, and combine the baseline accuracy of the current processing equipment to obtain the size tolerance, shape tolerance and surface roughness error value at different times = error accuracy percentage at different times x baseline accuracy, wherein the baseline accuracy includes size tolerance accuracy, shape tolerance accuracy and surface roughness accuracy;

[0026] S312, combine the processing requirements of the current order to be processed to obtain the processing accuracy threshold of the current order to be processed, including size tolerance range threshold, shape tolerance range threshold and surface roughness range threshold, determine the current processing equipment, and traverse the size tolerance, shape tolerance and surface roughness error value at each time in the baseline error accuracy curve. When there is any time of size tolerance error value size tolerance range threshold or shape tolerance error value shape tolerance range threshold or surface roughness error value surface roughness range threshold, record the current abnormal time and mark the current processing equipment.

[0027] As a preferred embodiment, the S32 comprises the following steps:

[0028] S321, according to the quantitative analysis results of the processing requirements of different orders to be processed, substitute the characteristic data of the current order to be processed into the baseline wear curve model of different equipment, combine the marked processing equipment with the recorded abnormal time, and obtain the wear degree of the current processing equipment that does not meet the production of the current order to be processed by substituting the abnormal time into the baseline wear curve;

[0029] S322、According to the wear degree threshold of different equipment, the current processing equipment is judged, when the wear degree < the wear degree threshold, it represents that the current processing equipment can still produce processing, when the wear degree ≥ the wear degree threshold, it represents that the current processing equipment needs to be maintained, and the current processing equipment is marked for maintenance.

[0030] As a preferred embodiment, the S322 comprises the following steps:

[0031] S3221, for the processing equipment with wear degree < wear degree threshold, the processing data is collected during the whole order processing, including equipment state data and work load data, and the size tolerance, shape tolerance, surface roughness of the processed parts, material type and hardness of the processed parts, and single processed part processing cycle are collected, new time series data points are constructed, data processing is carried out through principal component analysis combined with self-encoder, and current processing equipment temporary wear curve model and temporary error precision curve model are obtained through long short-term memory network;

[0032] S3222, based on the temporary wear curve model and the temporary error precision curve model of the current processing equipment, combined with the characteristic data of the remaining orders to be processed in the intelligent factory, the processing equipment judgment steps of S311, S312, S321 and S322 are repeated, and the current processing equipment is matched with the order again until the current processing equipment is marked for maintenance.

[0033] The intelligent factory equipment predictive maintenance system based on Internet of Things comprises a data acquisition module, a model establishment module and a processing equipment maintenance judgment module.

[0034] The data acquisition module is used for collecting processing data of different types of processing equipment in the intelligent factory, and the processing data is collected through sensors.

[0035] The model establishment module is used for constructing device double-index health curve models, including benchmark wear curve model and benchmark error precision curve model, for different types of processing equipment.

[0036] The processing equipment maintenance determination module establishes a process parameter rule library based on current intelligent factory historical processing data, performs quantitative analysis on the current intelligent factory order to be processed based on the order data in the order database of the product to be processed and the process parameter rule library, and predicts the wear trend and error precision change of the equipment involved in the order based on the quantitative analysis result of the processing requirements of different orders to be processed and the double-index health curve model of different processing equipment, so as to perform predictive maintenance on the order equipment and re-match the equipment order.

[0037] The beneficial effects of the present application are:

[0038] 1. The present application establishes a double-index health curve model for different processing equipment, combines different quantitative analysis results of orders to be processed, determines the processing equipment through the processing requirements of the order to be processed, and performs predictive maintenance or re-matches the equipment order based on the determination result, so as to evaluate the order to be processed based on the state of different equipment;

[0039] 2. The present application predicts the error state of the processing equipment in the processing process through the reference wear curve model and the reference error precision curve model, determines the error state of the equipment combined with the processing requirements of the order, predicts the wear degree of the equipment under the abnormal node, evaluates the subsequent state of the equipment under the abnormal node, maintains the equipment with high wear degree, re-matches the order for the equipment that can still be used, and can realize reasonable allocation of high requirement-moderate requirement-low requirement processing orders, while avoiding over-maintenance and resource waste caused by regular maintenance or after-service repair, thereby enhancing the practicability and functionality;

[0040] 3. The present application can determine whether the equipment needs to be maintained by predicting the wear degree of the equipment under the abnormal node, and the equipment is maintained when it is really needed, thereby avoiding unnecessary maintenance operation, improving the maintenance efficiency and effect, and simultaneously maintaining the equipment with high wear degree in time according to the equipment state evaluation result under the abnormal node, reducing the further damage of the equipment and the wear speed of the equipment, and when the order demand and the equipment state suddenly change, the system can quickly re-match the order-equipment, re-match the equipment-order, so as to respond to the internal and external changes. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0042] Figure 1This is a flowchart of a predictive maintenance method for smart factory equipment based on the Internet of Things, according to an embodiment of the present invention.

[0043] Figure 2 This is a block diagram of a predictive maintenance system for smart factory equipment based on the Internet of Things, according to an embodiment of the present invention. Detailed Implementation

[0044] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0045] According to embodiments of the present invention, a predictive maintenance system and method for smart factory equipment based on the Internet of Things are provided.

[0046] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments:

[0047] Example 1:

[0048] like Figure 1 As shown, according to an embodiment of the present invention, a predictive maintenance method for smart factory equipment based on the Internet of Things includes the following steps:

[0049] S1. Collect the parameters of the current smart factory equipment, establish a dual-index health curve model for the equipment based on the attributes of different equipment, and collect the order data of products to be processed to establish a database of orders of products to be processed.

[0050] S11. For different types of processing equipment in the smart factory, sensors are installed to collect processing data during the operation of the processing equipment. For different types of processing equipment, a dual-index health curve model is constructed, including a reference wear curve model and a reference error accuracy curve model.

[0051] S111. Using equipment maintenance time points as the cycle, extract the collected data from different processing equipment within N time periods after maintenance, including equipment status data and workload data. Equipment status data includes spindle motor power, current, and torque feedback; workload data includes spindle speed, feed rate, depth of cut, cutting width, workpiece material type and hardness value, and processing cycle per workpiece. Data processing is performed using principal component analysis combined with an autoencoder, and a long short-term memory network is used to obtain the reference wear curve model of the current processing equipment. The specific steps are as follows:

[0052] For the extracted feature vector X = [x1, x2, x3, ..., xm] containing m features m ] T Perform each feature Standardization processing, where x ij μ j σ j Let z represent the j-th feature value of the i-th sample, the mean of the j-th feature, and the standard deviation of the j-th feature, respectively. ij The standardized values ​​are represented by Z, and the standardized data matrix is ​​denoted as Z and passed through... The covariance matrix C is calculated, where n is the number of samples;

[0053] Performing eigenvalue decomposition on the covariance matrix C yields the eigenvalues ​​λ1≥λ2≥λ3≥…≥λ m And the eigenvectors ω1, ω2, ω3, ..., ω corresponding to the eigenvalues. m We select the first k principal components such that the contribution rate of the cumulative variance CV(k) reaches 95%, where Projecting the standardized data Z onto the subspace spanned by the eigenvectors corresponding to the first k principal components yields the dimensionality-reduced data matrix Y = ZW. k W k It is a matrix composed of the first k eigenvectors;

[0054] It should be noted that N is usually set to 30 days, but can be adjusted according to the specific equipment situation. Data needs to be re-recorded after each equipment maintenance time point to provide data support for the model building of the next cycle, so as to realize the update of the dual-index health curve model. The eigenvalue represents the variance of each principal component, and the eigenvector represents the direction of the principal component. In the covariance matrix, the diagonal elements are the variance of the i-th feature itself, and the off-diagonal elements are the covariance between the i-th feature and the j-th feature, which is used to measure the degree of linear correlation between them. The magnitude of the eigenvalue directly represents the amount of original data information carried by the direction of the corresponding principal component. The larger the eigenvalue, the more important the direction of the principal component is, thus screening out features with high correlation to wear. The extracted equipment status features are aligned with the process parameters during order processing according to the timestamp to form production records, thereby obtaining HI.

[0055] The input Y is mapped to a low-dimensional bottleneck layer representation h = f by an encoder. enc (Y), the bottleneck layer representation h is reconstructed into the output Y' = f by the decoder. dec (h) The input Y, after dimensionality reduction through principal component analysis, is reconstructed using a trained autoencoder to obtain the output Y', thus yielding the health index HI = ||Y - Y'||. 2For different processing equipment, the HI(t) point that changes with time t is obtained, and the reference wear curve model of the current processing equipment is obtained by training a long short-term memory network.

[0056] It should be noted that when processing the encoder and decoder outputs, the encoder needs to be implemented using a neural network. For an encoder with layer l, the output h of layer s is... s =σ(W s h s-1 +b s ), where W s b s These represent the weight matrix and bias vector of the s-th layer, respectively, where σ is the ReLU function, and h... 0 =Y, the decoder is also implemented using a neural network, where the output y of layer a is... a =σ(W a y a-1 +b a ), where W a b a Let and represent the weight matrix and bias vector of layer a, respectively, and y 0 =h, Where L dec Let be the number of layers in the decoder. Using mean squared error as the loss function, we use SGD to minimize the loss function to update the parameters of the autoencoder.

[0057] In the process of training the Long Short-Term Memory (LSTM) network and creating the benchmark wear curve model, the feature data related to equipment wear obtained through principal component analysis and autoencoder processing are organized into a sequence suitable for LTM input, arranged chronologically. Each time step contains multiple feature dimensions, such as vibration signals, process signals, and spindle speed. The dataset is divided into training, validation, and test sets, with the training set accounting for 70%, the validation set for 15%, and the test set for 15%. Model training is then performed, with the input layer having the same dimension as the feature data at each time step. The output layer predicts the degree of equipment wear, which is a scalar, using mean squared error as the loss function. The Adam algorithm is used to update the model's weights and biases. The test set data is used to finally evaluate the optimized model, resulting in the benchmark wear curve model.

[0058] S112. Using the equipment maintenance time point as the period, extract the data collected from the processed parts of different processing equipment within N time periods after maintenance, including the dimensional tolerance, shape tolerance, surface roughness, material type and hardness of the processed parts, and the processing cycle of a single processed part. Through principal component analysis combined with autoencoder processing and long short-term memory network, obtain the reference error accuracy curve model of the current processing equipment.

[0059] It should be noted that the data collected by the machining parts of different machining equipment in the N time period after maintenance in S112 is extracted by principal component analysis combined with self-encoder processing, the feature vector containing the extracted features needs to be standardized, and the covariance matrix is calculated and obtained, the covariance matrix is subjected to eigenvalue decomposition, the error accuracy point changing with time is obtained through the encoder and the decoder, and then the current machining equipment is obtained by long short-term memory network training. The reference wear curve model of the output layer is the error accuracy percentage of the predicted equipment, which is a scalar. The reference error accuracy curve model and the reference wear curve model are obtained for different machining equipment in the current intelligent factory, which facilitates the output of the reference error accuracy curve and the reference wear curve of different equipment in the subsequent order processing process.

[0060] S12, establish a product order database to be processed and a machining equipment database through MySQL, file and record the data related to different machining equipment in the machining equipment database, collect product order data to be processed and record it in the product order database to be processed, and the product order data to be processed includes product name, material type and hardness value of machining parts, machining steps, size accuracy requirement, shape accuracy requirement, surface roughness requirement, production quantity, and delivery time.

[0061] It should be noted that different types of machining equipment in the intelligent factory are equipped with sensors to collect machining data during the operation of the machining equipment. The installation of sensors and parameter collection needs to be combined with the specific conditions of different machining equipment, including the wear of machining tools, part precision changes, and other machining parameters related to different equipment. At the same time, sensors are installed at key positions such as tools and spindles to collect representative equipment parameters of different machining equipment such as spindle speed, feed speed, current, power, real-time temperature, and pressure. By collecting different product order parameters to be processed and establishing and statistical database files, data support can be provided for subsequent equipment-order association, which facilitates predictive maintenance of different orders for different states of equipment.

[0062] S2, based on the historical machining data of the current intelligent factory, establish a process parameter rule library, and based on the order data in the product order database to be processed, quantitatively analyze the current intelligent factory order to be processed based on the process parameter rule library;

[0063] S21, based on the historical order processing data, the process parameter rule library is established, based on the order data in the order database of the product to be processed, the machining requirements of the current intelligent factory order are quantitatively analyzed through the process parameter rule library, and the spindle motor power, current, torque feedback, spindle speed, feed speed, cutting depth, cutting width, single machining piece machining cycle in the current order processing process are obtained according to the product name, specification, machining step, size precision requirement, shape precision requirement, surface roughness requirement, production quantity, delivery time in the order of the product to be processed;

[0064] It should be noted that in the process of establishing the process parameter rule library, detailed data of past processed products need to be extracted from the production record system of the intelligent factory, including product name, specification, machining step, corresponding spindle motor power, current, torque feedback, spindle speed, feed speed, cutting depth, cutting width and other process parameters, to ensure that the data covers a variety of different types of products and processing scenarios, and at the same time, it is necessary to communicate with senior process engineers and technical experts inside the intelligent factory, and to formulate process parameter mapping rules based on experience, for the same or similar product name, specification and machining step, the statistical results of historical data, the recommended values of industry standards and expert experience are comprehensively considered to determine the reasonable process parameter value range or specific value, and for complex processing conditions, rules based on multiple factors can be formulated to determine the process parameters according to the material, size and processing precision requirements of the product.

[0065] S22, based on the size precision requirement, shape precision requirement, surface roughness requirement of the current product to be processed order, the size tolerance, shape tolerance and surface roughness of the current product to be processed are obtained according to the mechanical processing standard specification.

[0066] It should be noted that according to the size precision requirement, shape precision requirement, surface roughness requirement of the current product to be processed order, the relevant mechanical processing standard specification is searched, including national standard GB, international standard ISO, etc., to obtain the size tolerance, shape tolerance and surface roughness of the current product to be processed, wherein the mechanical processing standard specification needs to be consistent with the standard at the time of establishing the reference error precision curve model, different precision requirements correspond to different tolerance grades and surface roughness grades in the standard specification, according to the shape characteristics of the product to be processed, such as flatness, cylindricity, roundness, etc., the shape tolerance value is determined according to the shape precision requirement and the corresponding standard specification, for example, the shape precision requirement is "flatness 0.01", and the shape tolerance value is 0.01, according to the grade of the surface roughness requirement, the corresponding specific value is searched in the surface roughness standard table, for example, the surface roughness requirement is "Ra 1.6", and the surface roughness value is 1.6 μm.

[0067] Example 2:

[0068] S3, based on the quantitative analysis results of the processing requirements of different orders to be processed and the double-index health curve model of different processing equipment, the wear trend and error accuracy change of the equipment involved in the current order are predicted, and predictive maintenance and order re-matching of the equipment are performed;

[0069] S31, according to the quantitative analysis results of the processing requirements of different orders to be processed, and combining the double-index health curve model of different processing equipment, the reference wear curve and the reference error accuracy curve of different processing equipment under the current order to be processed are obtained, and the processing equipment is pre-judged according to the error accuracy percentage in the reference error accuracy curve and the processing requirements of the current order to be processed;

[0070] S311, according to the quantitative analysis results of the processing requirements of different orders to be processed, the characteristic data of the current order to be processed is substituted into the reference error accuracy curve model of different equipment, the error accuracy percentage output in the reference error accuracy curve of different equipment for the current order to be processed is obtained, and the size tolerance, shape tolerance and surface roughness error value at different times are obtained by combining the reference accuracy of the current processing equipment, wherein the reference accuracy includes size tolerance accuracy, shape tolerance accuracy and surface roughness accuracy.

[0071] It should be noted that the reference accuracy needs to be obtained through the equipment specification of the current processing equipment, and the reference accuracy is obtained by referring to the equipment specification of the current processing equipment, which specifically includes size tolerance accuracy, shape tolerance accuracy and surface roughness accuracy. The characteristic data is the input data of the long short-term memory network.

[0072] S312, combining the processing requirements of the current order to be processed, the processing accuracy threshold of the current order to be processed is obtained, including the size tolerance range threshold, the shape tolerance range threshold and the surface roughness range threshold, the current processing equipment is judged, and the size tolerance, shape tolerance and surface roughness error value at each time in the reference error accuracy curve are traversed. When there is any time of size tolerance error value size tolerance range threshold or shape tolerance error value shape tolerance range threshold or surface roughness error value surface roughness range threshold, record the current abnormal time and mark the current processing equipment.

[0073] S32, based on the use time point of the processing equipment pre-judged, the wear degree of the current equipment is obtained by combining the reference wear curve, the wear degree of the equipment is judged by combining the equipment maintenance threshold, and based on the judgment result, the error accuracy percentage of the equipment is combined with the processing requirements of the remaining orders to be processed, and the processing equipment is subjected to predictive maintenance and order re-matching;

[0074] S321, according to the quantitative analysis result of the processing requirements of different orders to be processed, the characteristic data of the current order to be processed is substituted into the baseline wear curve model of different equipment, the marked processing equipment is combined with the recorded abnormal moment, the abnormal moment is substituted into the baseline wear curve to obtain the wear degree of the current processing equipment when it does not meet the production of the current order to be processed;

[0075] S322, according to the wear degree threshold of different equipment, the current processing equipment is judged, when the wear degree < wear degree threshold, it represents that the current processing equipment can still produce and process, when the wear degree ≥ wear degree threshold, it represents that the current processing equipment needs to be maintained, and the current processing equipment is marked for maintenance.

[0076] It should be noted that the wear degree threshold of different equipment is different, and needs to be combined with the service life and equipment state of different equipment, and set by consulting experts in the current field based on experience method. The initial wear degree threshold is usually set to 55%, that is, when the wear degree exceeds 55%, predictive maintenance is needed. For equipment in good condition, the threshold can be increased, and for equipment in poor condition, the threshold can be reduced. The specific wear degree threshold needs to be set in combination with the actual application scene.

[0077] S3221, for the processing equipment with wear degree < wear degree threshold, processing data is collected during the whole order processing process, including equipment state data and work load data, as well as size tolerance, shape tolerance, surface roughness of processed parts, material type and hardness of processed parts, and single processed part processing cycle. New time series data points are constructed, data processing is carried out through principal component analysis combined with self-encoder, and current processing equipment temporary wear curve model and temporary error accuracy curve model are obtained through long short-term memory network;

[0078] It should be noted that the equipment state data and work load data are consistent with the type of data collected in S111, wherein the equipment state data is spindle motor power, current and torque feedback, and the work load data includes spindle speed, feed speed, cutting depth, cutting width, material type and hardness value of processed parts, and single processed part processing cycle. Through the construction of temporary wear curve model and temporary error accuracy curve model, the processing adaptation degree of the next order can be evaluated by comprehensively considering the processing influence of the last order, so as to realize the closed-loop production of prediction-execution-update-re-prediction.

[0079] S3222, based on the temporary wear curve model and the temporary error accuracy curve model of the current processing equipment, the characteristic data of the remaining orders to be processed in the intelligent factory is combined, and the processing equipment judgment steps of S311, S312, S321 and S322 are repeated to match the orders again for the current processing equipment until the current processing equipment is marked for maintenance.

[0080] It should be noted that in order to match the order again by the current processing equipment, the characteristic data of all orders to be processed in the current smart factory is traversed, and the temporary wear curve model and the temporary error precision curve model are used to screen the orders to be processed that meet the current processing equipment for production and processing.

[0081] Embodiment 3:

[0082] As shown in Figure 2 The predictive maintenance system of the smart factory equipment based on the Internet of Things includes a data acquisition module, a model establishment module, and a processing equipment maintenance determination module.

[0083] The data acquisition module installs sensors for different types of processing equipment in the smart factory to collect processing data during the operation of the processing equipment. Meanwhile, a MySQL is used to establish a database of orders to be processed and a database of processing equipment. The data related to different processing equipment is recorded in the database of processing equipment, and the data of orders to be processed is collected and recorded in the database of orders to be processed. The data of orders to be processed includes product name, material type and hardness value of processing parts, processing steps, size precision requirement, shape precision requirement, surface roughness requirement, production quantity, and delivery time.

[0084] The model establishment module establishes a double-index health curve model for different types of processing equipment in combination with the collected data, including a benchmark wear curve model and a benchmark error precision curve model.

[0085] The processing equipment maintenance determination module establishes a process parameter rule library based on the historical processing data of the current smart factory. According to the order data in the database of orders to be processed, the current orders to be processed in the smart factory are quantitatively analyzed based on the process parameter rule library. Based on the quantitative analysis results of different orders to be processed and the double-index health curve models of different processing equipment, the wear trend and error precision change of the equipment involved in the current order are predicted, and the predictive maintenance of the order equipment and the re-matching of the equipment order are performed.

[0086] To sum up, the application establishes double-index health curve models for different processing equipment respectively, combines different to-be-processed order quantitative analysis results, judges the processing equipment according to the processing requirements of the to-be-processed order, and performs predictive maintenance or equipment order re-matching on the processing equipment based on the judgment result, realizes the evaluation of the to-be-processed order based on the state of different equipment, predicts the error state of the processing equipment in the processing process through the benchmark wear curve model and the benchmark error precision curve model, judges the error state of the equipment according to the processing requirements of the order, predicts the wear degree of the equipment under the abnormal node, evaluates the subsequent state of the equipment under the abnormal node, maintains the equipment with high wear degree, re-matches the order for the equipment that can still be used, can realize the reasonable allocation of high requirement-moderate requirement-low requirement processing orders, and can avoid over-maintenance and resource waste caused by regular maintenance or after-maintenance, and enhances the practicability and functionality.

[0087] The above merely describes the preferred embodiments of the present application and is not used to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predictive maintenance of smart factory equipment based on Internet of Things, characterized in that, The method comprises the following steps: S1, collecting current intelligent factory equipment parameters, establishing a device double-index health curve model based on different properties of the device itself, and collecting product order data to be processed to establish a product order database to be processed; S2, based on the historical processing data of the current intelligent factory, establishing a process parameter rule library, and based on the order data in the product order database to be processed, quantitatively analyzing the current intelligent factory order to be processed based on the process parameter rule library; S3, based on the quantitative analysis results of different processing orders and the double-index health curve model of different processing equipment, predicting the wear trend and error accuracy change of the equipment involved in the order, and performing predictive maintenance and equipment order re-matching on the order equipment. 2.The IoT-based smart factory equipment predictive maintenance method of claim 1, wherein, The S1 comprises the following steps: S11, for different types of processing equipment in the intelligent factory, respectively installing sensors to collect processing data during the operation of the processing equipment, and for different types of processing equipment, respectively constructing a device double-index health curve model, including a baseline wear curve model and a baseline error accuracy curve model; S12, establishing a product order database to be processed and a processing equipment database through MySQL, filing and recording different processing equipment related data in the processing equipment database, collecting product order data to be processed and recording in the product order database to be processed, the product order data to be processed including product name, material type and hardness value of the processed part, processing step, size accuracy requirement, shape accuracy requirement, surface roughness requirement, production quantity, delivery time. 3.The IoT-based smart factory equipment predictive maintenance method of claim 2, wherein, The S11 comprises the following steps: S111, taking the equipment maintenance time point as the period, extracting the collected data of different processing equipment in the N time period after maintenance, including equipment state data and work load data, wherein the equipment state data is spindle motor power, current and torque feedback, and the work load data includes spindle speed, feed speed, cutting depth, cutting width, material type and hardness value of the processed part, and single processed part processing cycle, the data is processed through principal component analysis combined with self-encoder, and the baseline wear curve model of the current processing equipment is obtained through long short-term memory network, and the specific steps are: For the extracted feature vector X = [x1, x2, x3, ..., xm] containing m features m ] T Perform each feature Standardization processing, where x ij μ j σ j Let z represent the j-th feature value of the i-th sample, the mean of the j-th feature, and the standard deviation of the j-th feature, respectively. ij The standardized values ​​are represented by Z, and the standardized data matrix is ​​denoted as Z and passed through... The covariance matrix C is calculated, where n is the number of samples; Eigenvalue decomposition is performed on the covariance matrix C to obtain eigenvalues λ1≥ λ2≥ λ3≥ … ≥ λ m and eigenvectors ω1, ω2, ω3,..., ω m corresponding to the eigenvalues, and the first k principal components are selected so that the contribution rate of the cumulative variance CV(k) reaches 95%, wherein The standardized data Z is projected into a subspace spanned by the eigenvectors corresponding to the first k principal components to obtain a reduced dimension data matrix Y = ZW k , wherein W k is a matrix composed of the first k eigenvectors. The input Y is mapped to a low-dimensional bottleneck layer representation h = f enc (Y) by an encoder, and the bottleneck layer representation h is reconstructed to the output Y' = f dec (h) by a decoder, the input Y after dimensionality reduction by principal component analysis is obtained by the reconstructed output Y' of the trained autoencoder, and the health index HI = ||Y-Y'|| 2 is obtained. The HI(t) points varying with time t are obtained for different processing equipment, and the baseline wear curve model of the current processing equipment is obtained by training a long short-term memory network. S112, taking the equipment maintenance time point as the period, extracting the collected data of different processing equipment in the N time period after maintenance, including the size tolerance, shape tolerance, surface roughness, material type and hardness of the processed part, and single processed part processing cycle, the data is processed through principal component analysis combined with self-encoder, and the baseline error accuracy curve model of the current processing equipment is obtained through long short-term memory network. 4.The IoT-based smart factory equipment predictive maintenance method of claim 3, wherein, The S2 comprises the following steps: S21, based on the historical order processing data, a process parameter rule library is established, based on the order data in the order database of the product to be processed, the machining requirements of the current intelligent factory order are quantitatively analyzed through the process parameter rule library, and the spindle motor power, current, torque feedback, spindle speed, feed speed, cutting depth, cutting width, and single machining piece machining cycle in the current order processing are obtained according to the product name, specification, machining step, size accuracy requirement, shape accuracy requirement, surface roughness requirement, production quantity, and delivery time in the order of the product to be processed. S22, based on the size accuracy requirement, shape accuracy requirement, and surface roughness requirement of the current product to be processed order, the size tolerance, shape tolerance, and surface roughness of the current product to be processed are obtained according to the machining standard specification. 5.The IoT-based smart factory equipment predictive maintenance method of claim 3, wherein, The S3 includes the following steps: S31, according to the quantitative analysis results of the machining requirements of different orders to be processed, the reference wear curve and the reference error accuracy curve of different machining equipment under the current order to be processed are obtained by combining the double-index health curve model of different machining equipment, and the machining equipment is pre-determined according to the error accuracy percentage in the reference error accuracy curve and the machining requirements of the current order to be processed; S32, based on the use time point of the pre-determined machining equipment, the wear degree of the current equipment is obtained by combining the reference wear curve, the wear degree of the equipment is determined by combining the equipment maintenance threshold, and based on the determination result, the error accuracy percentage of the equipment is combined with the machining requirements of the remaining orders to be processed, and the machining equipment is pre-maintained and re-matched with the order. 6.The IoT-based smart factory equipment predictive maintenance method of claim 5, wherein, The S31 includes the following steps: S311, according to the quantitative analysis results of the machining requirements of different orders to be processed, the characteristic data of the current order to be processed is substituted into the reference error accuracy curve model of different equipment, the error accuracy percentage output in the reference error accuracy curve of different equipment for the current order to be processed is obtained, and the size tolerance, shape tolerance, and surface roughness error value at different times are obtained by combining the reference accuracy of the current machining equipment, wherein the reference accuracy includes size tolerance accuracy, shape tolerance accuracy, and surface roughness accuracy; S312, in combination with the machining requirement of the current order to be processed, obtaining the machining precision threshold of the current order to be processed, including the size tolerance range threshold, the shape tolerance range threshold, the surface roughness range threshold, judging the current machining equipment, traversing the size tolerance, shape tolerance, surface roughness error value of each time in the reference error precision curve, when there is any time of size tolerance error value The size tolerance range threshold or the shape tolerance error value The shape tolerance range threshold or the surface roughness error value The surface roughness range threshold, record the current abnormal time and mark the current machining equipment. 7.The IoT-based smart factory equipment predictive maintenance method of claim 6, wherein, The S32 includes the following steps: S321, according to the quantitative analysis results of the machining requirements of different orders to be processed, the characteristic data of the current order to be processed is substituted into the reference wear curve model of different equipment, the wear degree of the current machining equipment that does not meet the production of the current order to be processed is obtained by substituting the abnormal time into the reference wear curve according to the recorded abnormal time of the marked machining equipment; S322, according to the wear degree threshold of different equipment, the current machining equipment is determined, when the wear degree is less than the wear degree threshold, it represents that the current machining equipment can still be produced and machined, and when the wear degree is greater than or equal to the wear degree threshold, it represents that the current machining equipment needs to be maintained, and the current machining equipment is marked for maintenance. 8.The IoT-based smart factory equipment predictive maintenance method of claim 1, wherein, The S322 includes the following steps: S3221, for the wear degree < wear degree threshold value of the processing equipment, the processing data is collected in the whole order processing process, including equipment state data and work load data, while collecting the size tolerance, shape tolerance, surface roughness of the processed parts, the material type and hardness of the processed parts, and the single processing part processing cycle, constructing new time series data points, processing data through principal component analysis combined with self-encoder, combining long short-term memory network to obtain the temporary wear curve model and temporary error precision curve model of the current processing equipment; S3222, based on the temporary wear curve model and temporary error precision curve model of the current processing equipment, combining the feature data of the remaining to-be-processed orders in the intelligent factory, repeating the processing equipment determination steps of S311, S312, S321 and S322, re-matching the orders of the current processing equipment until the current processing equipment is marked for maintenance.

9. An intelligent factory equipment predictive maintenance system based on Internet of Things, characterized in that, The system adopts the predictive maintenance method of intelligent factory equipment based on Internet of Things according to any one of claims 1-8, comprising a data acquisition module, a model establishment module and a processing equipment maintenance determination module; The data acquisition module is used for collecting processing data of different types of processing equipment in the intelligent factory, and establishing a to-be-processed product order database and a processing equipment database through MySQL. The related data of different processing equipment is recorded in the processing equipment database, and the to-be-processed product order data is collected and recorded in the to-be-processed product order database. The to-be-processed product order data includes product name, material type and hardness value of processing part, processing step, size accuracy requirement, shape accuracy requirement, surface roughness requirement, production quantity and delivery time. The model establishment module is used for constructing a double-index health curve model of the equipment, including a benchmark wear curve model and a benchmark error precision curve model, based on the collected data of different types of processing equipment. The processing equipment maintenance determination module is used for establishing a process parameter rule library based on the historical processing data of the current intelligent factory, performing quantitative analysis on the to-be-processed orders in the current intelligent factory based on the process parameter rule library and the order data in the to-be-processed product order database, predicting the wear trend and error precision change of the equipment involved in the current order based on the double-index health curve model of different processing equipment and the quantitative analysis result of different to-be-processed orders, and performing predictive maintenance and equipment order re-matching on the order equipment.

Images