TW202226088A - Method and system for anticipating delivery time - Google Patents
Method and system for anticipating delivery time Download PDFInfo
- Publication number
- TW202226088A TW202226088A TW109146820A TW109146820A TW202226088A TW 202226088 A TW202226088 A TW 202226088A TW 109146820 A TW109146820 A TW 109146820A TW 109146820 A TW109146820 A TW 109146820A TW 202226088 A TW202226088 A TW 202226088A
- Authority
- TW
- Taiwan
- Prior art keywords
- data
- production
- delivery
- model
- forecasting
- Prior art date
Links
Images
Landscapes
- General Factory Administration (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Description
揭露書公開一種預測交貨日期的方法,特別是指一種通過智能手段根據歷史數據與生產資訊預測交期的方法與系統。The disclosure document discloses a method for predicting the delivery date, especially a method and system for predicting the delivery date based on historical data and production information through intelligent means.
在產品製造過程,企業需要針對原物料、生產設備與人員等需求預備生產線,而在工廠端,達交率與生產選擇上很大程度影響著企業內部營運與財務狀況,若能精準控制上述關鍵課題將能有效掌控工廠運作、改善營運與財務狀況。然而,在目前工廠製作產品的過程往往受到諸多不可控因素導致而無法如預期交貨,也無法快速的決定產線變更時該如何應對。In the product manufacturing process, enterprises need to prepare production lines for raw materials, production equipment and personnel. On the factory side, the delivery rate and production selection greatly affect the internal operation and financial status of the enterprise. If the above-mentioned key factors can be accurately controlled The subject will be able to effectively control the operation of the factory, improve the operation and financial situation. However, the production process of products in the current factory is often caused by many uncontrollable factors and cannot be delivered as expected, and it is impossible to quickly decide how to respond when the production line is changed.
再者,因應少量多樣與即時化生產技術(Just In Time,JIT)的趨勢,企業回覆客戶的交期與客戶滿意度息息相關,現行產業常態為較短期(如72小時至1周)的交期回覆常常造成第一線業務的困擾。Furthermore, in response to the trend of just in time production technology (Just In Time, JIT) in a small amount of variety, the delivery time for companies to respond to customers is closely related to customer satisfaction. The current industry norm is a relatively short-term (such as 72 hours to 1 week) delivery time. Replies often cause trouble for front-line businesses.
為了能夠準確地提供客戶交貨日期,揭露書提出一種智能交期預測方法以及實現此方法的系統,智能交期預測系統中包括應用電腦技術的軟體與搭配硬體實現的功能模組,以執行智能交期預測方法,其中主要提供有提取實體(如公司、工廠)內的一企業資源規劃系統(ERP)數據的數據收集手段、利用機器學習演算法學習數據收集手段所取得的數據,並學習數據中各種信息的關聯性以形成預測交貨日期的預測模型的模型訓練手段。當數據收集手段接收到一生產需求的數據時,可通過所建立的預測模型預測交貨日期,並實現在此交貨日期的一或多種生產組合,形成決策,最終提供公司老闆決策。In order to accurately provide the customer delivery date, the disclosure document proposes an intelligent delivery forecasting method and a system for implementing the method. The intelligent delivery forecasting system includes software using computer technology and functional modules implemented with hardware to execute Intelligent delivery forecasting method, which mainly provides data collection means for extracting an enterprise resource planning system (ERP) data in entities (such as companies, factories), uses machine learning algorithms to learn the data obtained by the data collection means, and learns The correlation of various information in the data to form a model training method for predicting the delivery date. When the data collection means receives a production demand data, it can predict the delivery date through the established forecasting model, and realize one or more production combinations on this delivery date, form a decision, and finally provide the company boss with a decision.
進一步地,智能交期預測系統還可包括其他功能模組,如一數據清洗手段,用以清洗通過企業資源規劃系統取得的數據,以形成符合機器學習演算法學習所需的數據格式的數據;還可包括一自動優化手段,當模型訓練手段通過一或多個機器學習演算法學習數據得出多個預測模型時,可通過自動優化手段選擇或自行組成一表現較優的預測模型,作為預測交期的預測模型;亦可包括一自動學習手段,當通過數據收集手段取得新的數據時,可使得模型訓練手段以機器學習演算法學習新的數據,使得後續根據學習新的數據得出其中各種信息的關聯性優化預測模型。Further, the intelligent delivery forecasting system may also include other functional modules, such as a data cleaning method, which is used to clean the data obtained through the enterprise resource planning system, so as to form data in a data format conforming to the learning required by the machine learning algorithm; and It can include an automatic optimization method. When the model training method uses one or more machine learning algorithms to learn data to obtain multiple prediction models, it can select or form a better performance prediction model through the automatic optimization method. It can also include an automatic learning method. When new data is obtained through the data collection method, the model training method can use the machine learning algorithm to learn the new data, so that the subsequent learning of the new data can be used to obtain various kinds of data. Relevance of information optimizes predictive models.
更者,針對預測模型中可提供一權重設定功能,讓所述實體之決策者修正生產組合中物料規劃與/或排程規劃中的參數之權重改變生產順序;以及,系統還包括一生產組合選擇手段,其中利用另一機器學習演算法學習實體之決策者過去選擇之生產組合以及修正各生產組合中的參數之權重的歷史數據,形成一生產組合選擇模型,提供實體進行生產重劃。What's more, a weight setting function can be provided in the forecasting model, so that the decision maker of the entity can modify the weights of parameters in the material planning and/or scheduling planning in the production combination to change the production sequence; and, the system further includes a production combination. Selection means, wherein another machine learning algorithm is used to learn the production combinations selected by the decision maker of the entity in the past and to correct the historical data of the weights of the parameters in each production combination to form a production combination selection model, and provide the entity for production reassignment.
優選地,所述通過企業資源規劃系統取得的數據為實體過去生產組合的歷史數據,歷史數據至少包括至少一客戶訂單與至少一工廠出貨單,再以機器學習演算法通過歷史數據學習實體過去的生產組合,得出預測交貨日期的預測模型。更者,所述通過企業資源規劃系統取得的數據還包括工廠之生產數據與其中機台的生產時間。所述物料規劃的參數包括生產一產品所需的原料、零組件以及/或半成品的庫存;排程規劃的參數包括依照日期生產產品的交貨比例與數量。Preferably, the data obtained through the enterprise resource planning system is the historical data of the entity's past production combination, the historical data at least includes at least one customer order and at least one factory shipping order, and the machine learning algorithm is used to learn the entity's past through the historical data. production mix, resulting in a forecasting model for predicting delivery dates. Furthermore, the data obtained through the enterprise resource planning system also includes the production data of the factory and the production time of the machines therein. The parameters of the material planning include the inventory of raw materials, components and/or semi-finished products required to produce a product; the parameters of the schedule planning include the delivery ratio and quantity of the products produced according to the date.
為使能更進一步瞭解本發明的特徵及技術內容,請參閱以下有關本發明的詳細說明與圖式,然而所提供的圖式僅用於提供參考與說明,並非用來對本發明加以限制。For a further understanding of the features and technical content of the present invention, please refer to the following detailed descriptions and drawings of the present invention. However, the drawings provided are only for reference and description, and are not intended to limit the present invention.
以下是通過特定的具體實施例來說明本發明的實施方式,本領域技術人員可由本說明書所公開的內容瞭解本發明的優點與效果。本發明可通過其他不同的具體實施例加以施行或應用,本說明書中的各項細節也可基於不同觀點與應用,在不悖離本發明的構思下進行各種修改與變更。另外,本發明的附圖僅為簡單示意說明,並非依實際尺寸的描繪,事先聲明。以下的實施方式將進一步詳細說明本發明的相關技術內容,但所公開的內容並非用以限制本發明的保護範圍。The following are specific embodiments to illustrate the embodiments of the present invention, and those skilled in the art can understand the advantages and effects of the present invention from the content disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the concept of the present invention. In addition, the drawings of the present invention are merely schematic illustrations, and are not drawn according to the actual size, and are stated in advance. The following embodiments will further describe the related technical contents of the present invention in detail, but the disclosed contents are not intended to limit the protection scope of the present invention.
應當可以理解的是,雖然本文中可能會使用到“第一”、“第二”、“第三”等術語來描述各種元件或者信號,但這些元件或者信號不應受這些術語的限制。這些術語主要是用以區分一元件與另一元件,或者一信號與另一信號。另外,本文中所使用的術語“或”,應視實際情況可能包括相關聯的列出項目中的任一個或者多個的組合。It should be understood that although terms such as "first", "second" and "third" may be used herein to describe various elements or signals, these elements or signals should not be limited by these terms. These terms are primarily used to distinguish one element from another element, or a signal from another signal. In addition, the term "or", as used herein, should include any one or a combination of more of the associated listed items, as the case may be.
為了提供企業能在產品製造之前提供客戶準確的交貨日期,說明書公開一種智能交期預測方法與系統,其中利用機器學習法(或配合深度學習法)學習企業提供的生產數據,產生預測交期的預測模型,模擬企業決策過程。特別的是,在模擬企業決策的過程中,採用了企業歷史數據,並可引用進成本等邊界條件,且不影響真實工廠的運作,運用機器人流程自動化(Robotic Process Automation,RPA)方法進行多次模擬,再用機器學習法優化預測模型。進一步地,當取得最終決策合適解時,還可反向推出合適的安全庫存解。In order to provide enterprises with accurate delivery dates for customers before products are manufactured, the specification discloses an intelligent delivery forecast method and system, in which the machine learning method (or with the deep learning method) is used to learn the production data provided by the enterprise to generate a forecast delivery date Predictive models that simulate corporate decision-making processes. In particular, in the process of simulating enterprise decision-making, the historical data of the enterprise is used, and boundary conditions such as cost can be introduced without affecting the operation of the real factory. Robotic Process Automation (RPA) method is used for many times Simulation, and then use machine learning to optimize the predictive model. Further, when a suitable solution for the final decision is obtained, a suitable safety stock solution can also be reversed.
根據智能交期預測方法的實施例,主體為一運行於計算機系統的軟體解決方案,其中包含應用人工智能(AI)多目標優化核心系統、一個參數化的生產目標決策系統,以及一個可針對生產線定製的生產排成模擬系統。當根據客戶需求給定生產參數後,智能交期預測系統可以模擬製造工藝中每一個工序的設備、物料、半成品與人員之關係,以計算出每一批次產品完工時間。智能交期預測系統可以先進規劃排程系統(Advanced Planning and Scheduling,APS)實現。實現所述方法的計算機系統還包含了對企業資源計劃系統(Enterprise Resource Planning,ERP)的接口,可取得企業之財務、人員、設備、物料等數據,以及對製造執行系統(Manufacturing Execution System,MES)之接口,用以讀取歷史生產數據,並且可由智能交期預測系統主動發起執行命令。According to the embodiment of the intelligent delivery forecast method, the main body is a software solution running on a computer system, which includes a multi-objective optimization core system using artificial intelligence (AI), a parameterized production target decision-making system, and a production line. Customized production lines up the simulation system. When the production parameters are given according to customer requirements, the intelligent delivery forecast system can simulate the relationship between equipment, materials, semi-finished products and personnel in each process in the manufacturing process to calculate the completion time of each batch of products. The intelligent delivery forecasting system can be realized by Advanced Planning and Scheduling (APS). The computer system implementing the method also includes an interface to an enterprise resource planning system (Enterprise Resource Planning, ERP), which can obtain data such as finance, personnel, equipment, materials, etc. ) interface to read historical production data, and the intelligent delivery forecast system can actively initiate the execution of commands.
可參考圖1所示揭露書提出的智能交期預測系統運作的概念示意圖,此圖顯示智能交期預測系統中以智能手段為核心,能夠根據從企業得到的歷史數據101建立模型,其中提出系統模型103包括了交貨模型131與產能模型132,也就是說,通過智能交期預測系統可以根據客戶需求與企業的資源預測出交貨日期(即交期)外,還可推估出產能與生產相關參數。Please refer to the conceptual schematic diagram of the operation of the intelligent delivery forecasting system proposed in the disclosure book shown in FIG. 1 . This figure shows that the intelligent delivery forecasting system takes intelligent means as the core, and can build a model according to the
在智能交期預測系統運作的概念下,其中交貨模型131依照企業提供的資訊得出一預期交貨日(即天數105),還可包括交貨日之達交機率。產能模型132依照過往歷史數據之生產組合選擇讓機器學習過去決策者選擇之生產組合,相關生產數據如產品分類,包括既有商品、既有數量、數量改變、類似新品等資訊提出一套生產標準107,之後依照生產中批量時間109與生產規則111,決策者進行生產選擇113,最終產生生產選擇建議115。Under the concept of the operation of the intelligent delivery forecast system, the
在此一提的是,當系統提供了生產選擇建議115,可讓企業決策者(如業務、老闆等)得知新訂單需求可能影響已有訂單之生產,並使其根據交期預測與生產組合決定欲接訂單之交貨日期以及物料與替代的各種可能,最終讓決策者選擇最終解決方案。It is mentioned here that when the system provides the
圖2顯示智能交期預測系統的架構實施例圖。FIG. 2 shows a schematic diagram of the architecture of the intelligent delivery forecasting system.
圖中顯示實現交期預測系統20與其週邊提供數據的系統架構示意圖,其中之一為企業資源規劃系統24,此為一種運行傳統常用於企業內資源規劃計算機系統,連結企業內各部門管理伺服器,統一管理並記錄各種資源與信息,紀錄了企業各種層面的銷售生產採購庫存等數據,形成龐大的歷史生產數據庫,這些數據即可作為所述智能交期預測系統20中學習的數據,以形成預測模型。The figure shows a schematic diagram of the system architecture for realizing the delivery
交期預測系統20中主要包括人工智能多目標優化核心21以及生產排程模擬系統22。人工智能多目標優化核心21提供多種智能手段,如算法模塊211所描述的一或多種智能演算法,其中示意表示有幾個模組,如基因算法A、粒子群優化算法B以及決策樹算法C,而實際運行卻不受到圖示的幾種算法限制。The delivery
算法模塊211中表示系統可應用的幾種智能演算法,例如基因算法(genetic algorithm,GA)A,基因算法A為一種電腦模擬方法,在系統中可用於生產時間排程的模擬運算,可解決實際生產製程的問題。The
算法模塊211提出一粒子群優化(Particle Swarm Optimization,PSO)算法B,粒子群優化算法B在所述交期預測系統20中用於根據歷史數據得出提供決策的最佳解,特別是針對生產原物料採購時的優化問題模擬出最佳解。The
算法模塊211可包括一決策樹型算法(RF, XGB)C,決策樹型算法C是一種機器學習演算法,利用分類和回歸方法在數據中進行抽樣與隨機選取其中特徵,得出數據特徵後,可以模擬出決策的結果,目的是可以產生提供決策者的多種決策方向。The
經算法模塊211中一或多個智能演算法針對企業資源規劃系統24提供的數據進行機器學習演算與建模後,可以提供排程參數、物料人員規劃參數以及多種交期規劃組合至智能交期預測系統20中生產排程模擬系統22。生產排程模擬系統22還自加上自製造執行系統27用於收集製程上的即時訊息,接收到實際生產線上的數據。生產排程模擬系統22也取得生產系統25中的生產設備251、物料252與產品253等數據。After one or more intelligent algorithms in the
如此,使得生產排程模擬系統22可以根據歷史數據演算得出的生產相關排程、原物料、交期規劃、實際歷史生產數據,以及生產線上的真實數據,生產排程模擬系統22產生的數據可以回饋給人工智能多目標優化核心21,提供多組態結果以提供優化其中建立的預測模型,通過預測模型提供決策系統23多組產品訂單交期、生產成本與分批次生產選項等多重解,由決策系統23做出最終決策,包括提供給業務端29向客戶交待的交貨日期,以及提供製造執行系統27生產所需資源與生產指令,以進行生產。當決策系統23做出最終的交期判斷,還可以回溯提供企業應該有的庫存最佳解。In this way, the production
通過智能交期預測系統20,使得企業決策者可以獲得模擬得出的交期以及生產資訊,主要目的之一是提供業務端29能夠快速且相對精準地回應客戶預期交貨日,並依照其中建立的預測模型伸出生產組合選擇模型,可協助工廠管理者迅速判斷如何規劃生產並估計出多重選擇下的成本變更可能。Through the intelligent
圖3顯示智能交期預測系統中以計算機系統搭配軟體手段實現智能交期預測系統20的各種功能手段,各種功能手段以圖中顯示的功能模組來描述。FIG. 3 shows various functional means of the intelligent
智能交期預測系統20中主要的智能手段包括數據收集模組301,實現數據收集的方式如計算機系統通過有線或無線通訊方式通過網路或是特定連線取得一實體(如公司、工廠)內數據,實施例如企業資源規劃系統(ERP)提供數據,數據主要為實體過去生產組合的歷史數據,還可包括工廠之生產數據與其中機台的生產時間,之後可在智能交期預測系統20建立針對此企業的客製化資料庫,再以模型訓練模組307利用一機器學習演算法學習所收集的數據,將數據透過編碼讓機器學習演算法依照提供之數據找出關聯規則,學習數據中各種信息的關聯性,最後可以形成預測交貨日期的預測模型。其中,當以上述數據收集模組301的通訊手段接收到生產需求的數據時,可通過預測模型預測出交貨日期,以及實現交貨日期的一或多種生產組合,用以提供企業決策者進行決策。The main intelligent means in the intelligent
根據智能交期預測系統20的實施例,還可包括其他數據處理與學習的軟體搭配硬體實現的功能模組,如一數據清洗模組303,實現的數據清洗手段用以清洗通過企業資源規劃系統取得的數據,以形成符合機器學習演算法學習所需的數據格式的數據,使得後續手段可依照此格式作為模型訓練的數據。According to the embodiment of the intelligent
智能交期預測系統20還可包括一自動優化模組309,當模型訓練模組307通過一或多個機器學習演算法學習數據得出多個預測模型時,可通過自動優化模組309實現的軟體手段相互比對各種模型,選擇或自行組成一表現較優的預測模型,作為預測交期的預測模型。The intelligent
當通過上述數據收集模組301取得新的數據時,系統利用自動學習模組305針對模型訓練模組307以機器學習演算法學習新的數據的結果,自動提取新的生產流程與學習新的數據,特別可根據學習新的數據得出其中各種信息的關聯性,進行微調,以優化預測模型。When new data is obtained through the above-mentioned
智能交期預測系統20可再設有一生產組合選擇模組311,這是在智能交期預測系統20得出預測交期後,可利用生產組合選擇的軟體手段利用另一機器學習演算法學習所述實體之決策者過去選擇之生產組合以及修正各生產組合中的參數之權重的歷史數據,之後形成一生產組合選擇模型,提供實體進行一生產重劃。根據一實施例,所形成的生產組合選擇模型用以提出實體部分交貨或挪單挪料的建議組合,形成新的物料規劃與/或新的排程規劃的新的生產組合。The intelligent
根據智能交期預測系統20可應用的一情境應用。企業執行生產時,業務人員需要在工廠於客戶下訂單時回覆客戶批量之預期交貨日,但常見是業務人員需要與產線人員與物料管理人員溝通並確認目前工廠產能狀況,並依照經驗給出預交日,但此預交日與真實交貨日往往相差許多。因此,揭露書提出之智能交期預測系統20採用了上述各功能模組的智能手段,自企業取得歷史數據,如客戶訂單、工廠出貨單等,之後利用機器學習演算法學習數據,訓練得出預測模型,以進行現有產品的交貨預測。而此預測模型因為參考了過去的數據,通過智能手段可以處理將來缺乏完整資料之新品或少量產品的情況,其中主要方式是可以蒐集企業的生產資訊與機台生產時間等資料,如物料前置時間、機台產能與參數、工序間運行時間等,學習得出符合企業需求的交期以及生產組合。According to a situational application that the intelligent
接著,圖4顯示智能交期預測方法的實施例之一流程圖,其中動作描述可參考圖3。Next, FIG. 4 shows a flow chart of an embodiment of the intelligent delivery date prediction method, wherein the action description can refer to FIG. 3 .
智能交期預測方法執行於一計算機系統中,通過計算機系統中的處理電路、記憶體、資料庫實現各種學習演算法,其中流程包括,一開始,計算機系統通過內部或外部網路取得企業資源規劃系統的數據,至少包括至少一客戶訂單與至少一工廠出貨單(步驟S401),並以其中軟體手段清洗數據,主要是清洗通過企業資源規劃系統取得的數據,以形成符合機器學習演算法學習所需的數據格式的數據,另還可過濾不必要、無效或會影響效能的數據(步驟S403),之後即開始利用機器學習演算法學習數據中信息關聯性(步驟S405),形成預測模型(步驟S407)。其中,若通過一或多個機器學習演算法學習數據得出多個預測模型時,計算機系統還可用以選擇或自行組成一表現較優的預測模型,作為預測交期的預測模型。之後,根據一實施例,若產生新的數據,可繼續以機器學習演算法學習新的數據,根據學習新的數據得出其中各種信息的關聯性,回饋給系統以優化預測模型。The intelligent delivery forecast method is implemented in a computer system, and various learning algorithms are implemented through the processing circuit, memory, and database in the computer system. The process includes, at the beginning, the computer system obtains the enterprise resource planning through internal or external network. The data of the system at least includes at least one customer order and at least one factory shipping order (step S401 ), and the data is cleaned by means of software, mainly cleaning the data obtained through the enterprise resource planning system, so as to form a machine learning algorithm for learning The data in the required data format can also filter unnecessary, invalid or performance-impacting data (step S403 ), and then start to use the machine learning algorithm to learn the information correlation in the data (step S405 ) to form a prediction model ( Step S407). Wherein, if multiple prediction models are obtained by learning data through one or more machine learning algorithms, the computer system can also be used to select or compose a prediction model with better performance as a prediction model for predicting the delivery date. Afterwards, according to an embodiment, if new data is generated, the machine learning algorithm can continue to learn the new data, and the correlation of various information in the new data can be obtained according to the learned new data, and fed back to the system to optimize the prediction model.
完成預測模型後,即接收生產數據(步驟S409),例如業務端接收客戶訂單、產品品項、單位數量、訂單期望交期等數據,並可以同時接受多個訂單。其中細節包括各個產品品項製造所需要的物料、工序、每一工藝階段的半成品以及完成品之數據,這些數據同樣可由企業資源規劃系統(ERP)與製造執行系統(MES)導出。After the prediction model is completed, the production data is received (step S409 ). For example, the business end receives data such as customer orders, product items, unit quantity, and expected delivery date of the order, and can accept multiple orders at the same time. The details include the data of the materials, processes, semi-finished products and finished products required for the manufacture of each product item, which can also be derived from the Enterprise Resource Planning System (ERP) and Manufacturing Execution System (MES).
接著可以設定權重(步驟S411),預測模型提供權重設定功能,這是提供決策者調整生產參數,在此步驟中,可讓實體之決策者修正生產組合中物料規劃與/或排程規劃中的參數之權重改變生產順序。例如,如果這個客戶比其他客戶重要,或是這個半成品是很多其他產品的關鍵零組件,這些系統尚未考量到的參數可以透過調整權重來增加其優先順序。Then the weights can be set (step S411 ). The forecast model provides a weight setting function, which is to provide decision makers to adjust production parameters. In this step, the decision makers of the entity can be allowed to revise the material planning and/or scheduling planning in the production combination. The weights of the parameters change the production order. For example, if this customer is more important than other customers, or if this semi-finished product is a critical component of many other products, parameters not yet considered by the system can be weighted to increase their priority.
最後,通過預測模型得出交貨日期、生產組合等(步驟S413),其中,各生產組合可以記載物料規劃以及排程規劃,以及/或機台生產時間,而物料規劃的參數可包括生產一產品所需的原料、零組件以及/或半成品的庫存,排程規劃的參數則可包括依照日期生產該產品的交貨比例與數量。在一實施例中,之後,還可利用另一機器學習演算法學習實體之決策者過去選擇之生產組合,以及修正各生產組合中的參數之權重的歷史數據,用於形成一生產組合選擇模型,提供實體進行一生產重劃。Finally, the delivery date, production combination, etc. are obtained through the prediction model (step S413), wherein each production combination can record material planning and scheduling planning, and/or machine production time, and the parameters of material planning can include production-a Inventory of raw materials, components, and/or semi-finished products required for a product, and parameters for scheduling can include delivery ratios and quantities to produce the product by date. In one embodiment, another machine learning algorithm can also be used to learn the production combinations selected by the decision maker of the entity in the past, and to correct the historical data of the weights of the parameters in each production combination, so as to form a production combination selection model. , providing the entity to perform a production redistribution.
舉例來說,以鋁鍛造為例,鍛造分為粗鍛造與精密鍛造等兩個到數個工序,粗鍛造與精密鍛造使用的鍛造設備各自不同,而部分設備也可以被使用於不同的工序,鍛造的原物料為鋁材。另有可重複使用多次的模具,鍛造的半成品為經過粗鍛造之粗胚、完成品為精密鍛造形成之鋁鍛造件、每一個工序所需之工時、需要的人員配置、需用之原料、產生之成品、廢料與批量生產之成品率等數據,以及其對於不同設備之依賴性統計數據,均由以上所述之資料來源導入,如預測目標為系統執行過之品項,亦可以相同品項之歷史數據直接帶入,最終完整的生產模型數據經由工程與生管人員確認。For example, taking aluminum forging as an example, forging is divided into two or more processes such as rough forging and precision forging. The forging equipment used in rough forging and precision forging is different, and some equipment can also be used in different processes. The raw material for forging is aluminum. There are also molds that can be reused many times. The forged semi-finished products are rough forged blanks, and the finished products are aluminum forgings formed by precision forging. , data such as finished products, scraps, and mass production yields, as well as statistical data on their dependence on different equipment, are imported from the above-mentioned data sources. If the forecast target is the item that has been executed by the system, the same The historical data of the item is directly brought in, and the final complete production model data is confirmed by the engineering and production management personnel.
圖5顯示智能交期預測方法的另一實施例流程圖,此流程表達出所提出的智能交期預測系統針對現有產品與新產品的處理流程。第一部份針對企業現有產品,在方法中,先取得現有產品資訊(步驟S501),進行數據處理,包括數據清洗與過濾,目的是得出可建模的數據(步驟S503),再以機器學習演算法學習數據的特徵與關聯性(步驟S505),並建立預測模型(步驟S507)。另一方面為針對新產品,取得新產品資訊(步驟S509),包括引入進料時間、機台參數與生產時間、製程工序等數據(步驟S511),同樣地經過數據處理得出可建模的數據後(步驟S513),以機器學習演算法學習新數據中的特徵與關聯性(步驟S515),以建立預測模型(步驟S517)。FIG. 5 shows a flow chart of another embodiment of the intelligent delivery forecast method, which expresses the processing flow of the proposed intelligent delivery forecast system for existing products and new products. The first part is aimed at the existing products of the enterprise. In the method, the existing product information is obtained first (step S501 ), and data processing is performed, including data cleaning and filtering, in order to obtain modelable data (step S503 ), and then the machine The learning algorithm learns the features and correlations of the data (step S505 ), and establishes a prediction model (step S507 ). On the other hand, for new products, new product information is obtained (step S509 ), including the introduction of data such as feeding time, machine parameters and production time, and process steps (step S511 ). Similarly, through data processing, a modelable model is obtained. After the data is collected (step S513 ), a machine learning algorithm is used to learn the features and correlations in the new data (step S515 ) to establish a prediction model (step S517 ).
在此一提的是,針對新產品,這部份往往是缺乏如企業資源規劃系統提供的那樣完整的內容,因此可以採用相似原物料、機台設備、類似製程工法等之資料作為關聯學習標的,數據經清洗後,可對各項目各自建立預測模型,統合成一預測模型。It is mentioned here that, for new products, this part often lacks the complete content provided by the enterprise resource planning system, so the data of similar raw materials, machine equipment, and similar process engineering methods can be used as the relevant learning target. , After the data is cleaned, a prediction model can be established for each project and unified into a prediction model.
如此可知,系統針對現有產品與新的產品都形成了預測模型,並且可以是通過多個機器學習演算法得出多個預測模型,在步驟S519中,系統可繼續利用模型演算法得出各自結果,再選擇其中之一預測模型(步驟S521),預測模型用於產生交期(步驟S525),同樣地,新產生的數據將持續自動優化模型(步驟S523)。It can be seen that the system has formed prediction models for both existing products and new products, and can obtain multiple prediction models through multiple machine learning algorithms. In step S519, the system can continue to use the model algorithms to obtain their respective results. , and then select one of the prediction models (step S521 ). The prediction model is used to generate the due date (step S525 ). Similarly, the newly generated data will continue to automatically optimize the model (step S523 ).
實現以上方法流程的系統架構可繼續參考圖6所示實現智能交期預測方法的交貨模型與產能模型的架構實施例示意圖,此圖架構涵蓋了交貨與產能標準雙系統,圖示上半部示意顯示一交貨模型架構61,所在意的是時間數據611,根據企業資源規劃系統得到的客戶需求得到訂單資訊612、銷貨資訊613以及其他有關企業內人員等其他資訊614,進行數據清洗615,以及後續建模、訓練616等步驟,最終產生提供交貨時間的預測模型617。The system architecture for realizing the above method flow can continue to refer to the schematic diagram of the architecture example of the delivery model and the capacity model for realizing the intelligent delivery forecast method shown in FIG. The department shows a
另一方面如下半部描述的產能模型架構62,產能包括不容易變動的前置時間621以及生產所需的參數模型622,同樣地,都經過數據清洗(623,624)、建模、訓練(625,626)等步驟,建立針對前置處理的前置模型627以及處理生產參數的參數模型628,如此,開始規劃物料與排程629,導入上述預測模型後,產生交貨時間630。On the other hand, the production
圖7接著以示意圖表示智能交期預測方法中處理新舊產品數據的細節以及如何通過回饋建立數據以及優化模型的實施例。FIG. 7 then schematically shows the details of processing new and old product data in the intelligent delivery forecast method and an embodiment of how to establish data and optimize models through feedback.
根據圖示,針對現有產品71,企業資源規劃系統(ERP)提供了工廠數據711,並包括出貨單712與訂單713,經數據清洗後,得出可建模資料714,並據此學習與訓練,以建立模型715,預測模型用以預測交期716,直到系統得到真實交貨日717為止,這部份形成的預測交貨日與真實交貨日可以回饋到企業資源規劃系統,成為現有產品71中的工廠數據711的一部分。According to the figure, for the existing
針對新產品72,經客戶開出規格,得到產品數據721,還包括生產所需的各種廠內數據722,如機台723、出入料724與工序725等,各數據經數據清洗後,得出可建模資料(726, 727, 728),之後以機器學習演算法學習各項目數據,針對各項目建立模型(729, 730, 731),通過系統的軟體手段,可將各項目模型進行統合,建立統合模型732,能夠針對新產品提出新品交貨預測733,當得到真實交貨日734,產生的新品交貨預測733以及真實交貨日734都會形成系統中的新品數據,也是將來優化模型的依據。For the
最後,系統預測的交期還會比對真實交貨日,相關程序可參考圖8所示智能交期預測方法中採用企業數據的實施例圖。Finally, the delivery date predicted by the system will also be compared with the actual delivery date. For related procedures, please refer to the embodiment diagram of using enterprise data in the intelligent delivery date prediction method shown in FIG. 8 .
圖中顯示智能交期預測系統可以自企業資源規劃系統80取得的訂單數據包括預交日準確之訂單801、預交日準確之出貨單802、預交日不準確之訂單803以及預交日不準確之出貨單804,預交日準確與不準確數據都分別進行數據欣喜,分別得出可建模資料(805, 807),並建立預測模型806,經預測交期後,可比對真實交貨日808,以此作為確認模型809的依據。The figure shows that the order data that the intelligent delivery date prediction system can obtain from the enterprise
根據上述實施例可知,智能交期預測系統中採用人工智能多目標優化核心(可參考圖2,21)提供的數據,能夠提供生產排程模擬系統(可參考圖2,22)建構一個生產模型,生產模型考慮生產線協同工作,同時生產多個品項與批次,其中之設備與原料的分配使用,各個品項批次使用有限設備資源的優先順序,並關於生產訂單之拆分或合併、原物料取得的時間與成本等財務規劃,皆為多目標優化之可調適的參數。According to the above embodiment, the intelligent delivery forecast system adopts the data provided by the artificial intelligence multi-objective optimization core (refer to Fig. 2, 21) to provide the production scheduling simulation system (refer to Fig. 2, 22) to construct a production model , the production model considers the collaborative work of the production line to produce multiple items and batches at the same time, the allocation and use of equipment and raw materials, the priority of each item batch using limited equipment resources, and the split or merge of production orders, Financial planning such as time and cost of raw material acquisition are all adjustable parameters for multi-objective optimization.
舉例來說,在鋁鍛造的製程中,不同的工件所需要的預熱、加壓鍛造時間或有差異,同時還需要考慮操作人員之熟練度、各個工序設備加工之良品率等因素,經由以上過程產生的工藝參數以及歷史數據,可以得到工序耗時與用料的統計模型。根據實施例,採用統計模型之最壞、標準與最佳情況(Worst Case, Typical Case, Best Case)分析,或是採用更完整之統計採樣,可以平行計算蒙地卡羅方法(MC)進行模擬。For example, in the process of aluminum forging, the preheating and pressure forging time required for different workpieces may be different. At the same time, factors such as the proficiency of the operator and the yield rate of each process equipment need to be considered. Through the above The process parameters and historical data generated by the process can be used to obtain statistical models of process time consumption and materials. According to the embodiment, the Worst Case, Typical Case, Best Case (Worst Case, Typical Case, Best Case) analysis of the statistical model is adopted, or the more complete statistical sampling is adopted, and the Monte Carlo method (MC) can be used for simulation in parallel. .
而利多目標優化核心建構的模型,可生成多組可調適生產參數,經帶入生產排程模擬系統進行模擬,可以平行計算方法可以同時模擬大量的參數,以取得面向交期與財務的優化解。相關人工智能機器學習演算法例如圖2所提到的基因算法(GA)、粒子群算法(PSO)與決策樹型算法(RF, GB, XGB)等算法,可以在高維度的廣域參數空間中持續迭代,以極高的效率精準地收斂至最佳的參數組合。The model constructed by the multi-objective optimization core can generate multiple sets of adjustable production parameters, which can be brought into the production scheduling simulation system for simulation, and a large number of parameters can be simulated simultaneously by parallel computing methods to obtain optimal solutions for delivery and finance. . Related artificial intelligence machine learning algorithms, such as the genetic algorithm (GA), particle swarm algorithm (PSO) and decision tree algorithm (RF, GB, XGB) mentioned in Figure 2, can be used in high-dimensional wide-area parameter space. It iterates continuously in the middle, and accurately converges to the best parameter combination with extremely high efficiency.
智能交期預測系統最後預測交期經更新模型後將得到愈來愈接近真實交貨日的交期,使得智能交期預測系統提供的交期可以成為業務端與客戶協商訂單交期的準確依據,同時還可求得的生產參數,再經由工程與生管人員之確認後,基於智能交期預測系統與企業資源規劃系統、製造執行系統之接口,可以直接進行原料訂貨與執行生產命令,再確保最後完成之訂單交期與預測規劃之交期相符合。The intelligent delivery date prediction system finally predicts that the delivery date will be closer and closer to the real delivery date after updating the model, so that the delivery date provided by the intelligent delivery date prediction system can become the accurate basis for the business side and the customer to negotiate the order delivery date. , At the same time, the obtained production parameters can be confirmed by engineering and production management personnel. Based on the interface between the intelligent delivery forecast system, the enterprise resource planning system and the manufacturing execution system, it is possible to directly order raw materials and execute production orders. Make sure that the delivery date of the final completed order is in line with the delivery schedule of the forecast plan.
綜上所述,以上實施例描述一種智能交期預測方法與系統,所提出的智能交期預測系統先取得企業提供的數據,如過去生產的歷史數據以及目前物料與生產設備的相關數據,藉著智能手段學習數據,以形成預測交貨日期的預測模型,藉此模擬決策過程,而提供企業合適的解決方案,最終再由決策者做出生產規劃與交期判斷。如此,所述方法可以提供企業決策者與經營階層回答產能吃緊時又要滿足客戶訂單時,在不考慮短期內人工與增加機器設備時決策優先順序,並對挪單與替代料提供相對最適合的解。To sum up, the above embodiment describes an intelligent delivery date prediction method and system. The proposed intelligent delivery date prediction system first obtains the data provided by the enterprise, such as the historical data of past production and the related data of current materials and production equipment. It uses intelligent means to learn data to form a forecast model for predicting the delivery date, thereby simulating the decision-making process, and providing suitable solutions for the enterprise, and finally the decision makers make production planning and delivery judgments. In this way, the method can provide decision makers and management strata of the enterprise to answer customer orders when the production capacity is tight, without considering the short-term labor and the increase of machinery and equipment, the decision-making priority, and provide the most suitable for the replacement orders and substitute materials. solution.
以上所公開的內容僅為本發明的優選可行實施例,並非因此侷限本發明的申請專利範圍,所以凡是運用本發明說明書及圖式內容所做的等效技術變化,均包含於本發明的申請專利範圍內。The contents disclosed above are only preferred feasible embodiments of the present invention, and are not intended to limit the scope of the present invention. Therefore, any equivalent technical changes made by using the contents of the description and drawings of the present invention are included in the application of the present invention. within the scope of the patent.
101:歷史數據 103:系統模型 131:交貨模型 132:產能模型 105:天數 107:標準 109:批量時間 111:生產規則 113:生產選擇 115:生產選擇建議 24:企業資源規劃系統 20:交期預測系統 21:人工智能多目標優化核心 211:算法模塊 A:基因算法 B:粒子群優化算法 C:決策樹型算法 22:生產排程模擬系統 23:決策系統 25:生產系統 251:生產設備 252:物料 253:產品 27:製造執行系統 29:業務端 301:數據收集模組 303:數據清洗模組 305:自動學習模組 307:模型訓練模組 309:自動優化模組 311:生產組合選擇模組 61:交貨模型架構 611:時間數據 612:訂單資訊 613:銷貨資訊 614:其他資訊 615:數據清洗 616:建模、訓練 617:交貨時間預測模型 62:產能模型架構 621:前置時間 622:參數模型 623,624:數據清洗 625,626:建模、訓練 627:前置模型 628:參數模型 629:規劃物料與排程 630:交貨時間 71:現有產品 711:工廠數據 712:出貨單 713:訂單 714:可建模資料 715:建立模型 716:交貨預測 717:真實交貨日 72:新產品 721:產品數據 722:廠內數據 723:機台 724:出入料 725:工序 726, 727, 728:可建模資料 729, 730,731:建立模型 732:建立統合模型 733:新品交貨預測 734:真實交貨日 80:企業資源規劃系統 801:預交日準確之訂單 802:預交日準確之出貨單 803:預交日不準確之訂單 804:預交日不準確之出貨單 805, 807:可建模資料 806:建立預測模型 808:真實交貨日 809:確認模型 步驟S401~S413:智能交期預測實施例流程之一 步驟S501~S525:智能交期預測實施例流程之二 101: Historical data 103: System Models 131: Delivery Model 132: Capacity Model 105: days 107: Standard 109: Batch time 111: Production Rules 113: Production Options 115: Production Selection Recommendations 24: Enterprise Resource Planning Systems 20: Delivery Prediction System 21: AI multi-objective optimization core 211: Algorithm Module A: Genetic algorithm B: Particle Swarm Optimization Algorithm C: decision tree algorithm 22: Production scheduling simulation system 23: Decision System 25: Production System 251: Production Equipment 252:Material 253: Products 27: Manufacturing Execution Systems 29: Business side 301: Data Collection Module 303: Data cleaning module 305: Automatic Learning Module 307: Model training module 309: Automatic optimization module 311: Production combination selection module 61: Delivery Model Architecture 611: Time data 612: Order Information 613: Sales information 614: Other information 615: Data Cleaning 616: Modeling, training 617: Delivery Time Prediction Model 62: Capacity Model Architecture 621: lead time 622: Parametric Model 623,624: Data Cleaning 625, 626: Modeling, Training 627: Front Model 628: Parametric Model 629: Planning Materials and Scheduling 630: Delivery time 71: Existing Products 711: Factory Data 712: Shipper 713: Order 714: Modelable data 715: Modeling 716: Delivery Forecast 717: Real delivery date 72: New Products 721: Product Data 722: In-plant data 723: Machine 724: Loading and unloading 725: Process 726, 727, 728: Modelable data 729, 730, 731: Modeling 732: Build an integrated model 733: New Product Delivery Forecast 734: Real Delivery Date 80: Enterprise Resource Planning System 801: Order with accurate pre-delivery date 802: Shipper with accurate pre-delivery date 803: Order with inaccurate pre-delivery date 804: Shipper with inaccurate pre-delivery date 805, 807: Modelable data 806: Build predictive models 808: Real Delivery Date 809: Confirm model Steps S401-S413: One of the embodiments of intelligent delivery forecasting process Steps S501-S525: the second embodiment of the intelligent delivery forecasting process
圖1顯示智能交期預測系統運作的概念示意圖;Figure 1 shows a conceptual schematic diagram of the operation of the intelligent delivery forecasting system;
圖2顯示智能交期預測系統的架構實施例圖;Fig. 2 shows the structure embodiment diagram of the intelligent delivery forecast system;
圖3顯示智能交期預測系統的功能模組實施例圖;Fig. 3 shows the functional module embodiment diagram of the intelligent delivery forecast system;
圖4顯示智能交期預測方法的實施例之一流程圖;FIG. 4 shows a flow chart of one embodiment of the intelligent delivery forecast method;
圖5顯示智能交期預測方法的實施例之二流程圖;Fig. 5 shows the flow chart of the second embodiment of the intelligent delivery forecast method;
圖6顯示實現智能交期預測方法的交貨模型與產能模型的架構實施例示意圖;FIG. 6 shows a schematic diagram of an embodiment of the architecture of the delivery model and the capacity model for realizing the intelligent delivery forecast method;
圖7顯示智能交期預測方法中處理產品數據的實施例示意圖;以及FIG. 7 shows a schematic diagram of an embodiment of processing product data in an intelligent delivery forecast method; and
圖8顯示智能交期預測方法中採用企業數據的實施例圖。FIG. 8 shows an embodiment diagram of using enterprise data in the intelligent delivery forecast method.
101:歷史數據 101: Historical data
103:系統模型 103: System Models
131:交貨模型 131: Delivery Model
132:產能模型 132: Capacity Model
105:天數 105: days
107:標準 107: Standard
109:批量時間 109: Batch time
111:生產規則 111: Production Rules
113:生產選擇 113: Production Options
115:生產選擇建議 115: Production Selection Recommendations
Claims (20)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW109146820A TW202226088A (en) | 2020-12-30 | 2020-12-30 | Method and system for anticipating delivery time |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW109146820A TW202226088A (en) | 2020-12-30 | 2020-12-30 | Method and system for anticipating delivery time |
Publications (1)
Publication Number | Publication Date |
---|---|
TW202226088A true TW202226088A (en) | 2022-07-01 |
Family
ID=83436747
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW109146820A TW202226088A (en) | 2020-12-30 | 2020-12-30 | Method and system for anticipating delivery time |
Country Status (1)
Country | Link |
---|---|
TW (1) | TW202226088A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116362410A (en) * | 2023-04-14 | 2023-06-30 | 无锡星智数服科技有限公司 | MES-based production time prediction method, system and storage medium |
-
2020
- 2020-12-30 TW TW109146820A patent/TW202226088A/en unknown
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116362410A (en) * | 2023-04-14 | 2023-06-30 | 无锡星智数服科技有限公司 | MES-based production time prediction method, system and storage medium |
CN116362410B (en) * | 2023-04-14 | 2023-10-31 | 无锡星智数服科技有限公司 | MES-based production time prediction method, system and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Yelles-Chaouche et al. | Reconfigurable manufacturing systems from an optimisation perspective: a focused review of literature | |
Sadeghi et al. | A fuzzy grey goal programming approach for aggregate production planning | |
CN107464025B (en) | Supply chain demand prediction and intelligent decision method based on big data | |
Deliktas et al. | A flexible job shop cell scheduling with sequence‐dependent family setup times and intercellular transportation times using conic scalarization method | |
CN106537270A (en) | Cascaded model predictive control (mpc) approach for plantwide control and optimization | |
Bastos et al. | A maintenance prediction system using data mining techniques | |
CN112907150B (en) | Production scheduling method based on genetic algorithm | |
Saldivar et al. | Identifying smart design attributes for Industry 4.0 customization using a clustering Genetic Algorithm | |
CN114298426A (en) | APS dynamic production plan scheduling system | |
Latif et al. | A simulation algorithm of a digital twin for manual assembly process | |
TW202226088A (en) | Method and system for anticipating delivery time | |
CN114707874A (en) | Scheduling method, equipment and storage medium applied to aluminum alloy production | |
Kuric et al. | Computer aided process planning in machinery industry | |
Hung et al. | Analysis of key success factors for industry 4.0 development | |
Feyzioğlu et al. | A simulation-based optimization approach to size manufacturing systems | |
TWM613466U (en) | System for anticipating delivery time | |
JP7261639B2 (en) | Production control system and production control method | |
CN114819239A (en) | Intelligent delivery period prediction method and system | |
CN116415780A (en) | Intelligent ordering method and system | |
Talibi et al. | The relevance study of adaptive kanban in a multicriteria constraints context using data-driven simulation method | |
Liu et al. | Manufacturing Process Optimization in the Process Industry | |
CN112506164A (en) | Layout planning method, device and platform for reconfigurable manufacturing system | |
Yu et al. | Product mix optimization under required cycle time for semiconductor wafer fabrication | |
CN112381396B (en) | Integrated management method and system for serving enterprise development planning | |
Paweł et al. | Integration of Digital Twin with Simulation in Order to Meet Factory Expectations |