Intelligent manufacturing is facing significant challenges in adapting to the ever-changing equipment, instrumentation, process and economics. Such a trend together with the pressure to reliably control and contain production costs means that frequent adjusting decisions are required to adapt to incessant volatility imposed on manufacturing systems. Under this circumstance, cost-effective and quality-guaranteed manufacturing strategies would be the most logical route to reducing production costs. In this paper, a novel dynamical cost prediction and control (CPC) model is proposed to support collective decision-making in intelligent manufacturing, where the model output is the real-time prediction of possible manufacturing costs, while the inputs are generic manufacturing key performance indicators covering inventory, product quality, production efficiency, resource utilisation and environmental impact. This proposed CPC model distinguishes itself from existing ones for its capability to translate manufacturing data (at both the physical level and operation management level) into financial metrics that contribute to forming a common language between engineering, financial and administrative departments of an enterprise. The case study about the assembly line of optoelectronic devices demonstrates that, although different enterprise departments have different priorities, our CPC model helps them to achieve certain consensus on intended production that finally creates satisfactory profitability for the company at controlled manufacturing costs.

Article Hang Geng ^1,*, Alireza Mousavi ², Nikolaos Grigorios Markatos ², Kai Chen ¹, and Xuan Gou ^{1 1} School of Automation Engineering, University of Electronic Science and Technology of China, Chengdu 611731, China ² Department of Computer Science, Brunel University London, Uxbridge, Middlesex, UB8 3PH, United Kingdom ^* Correspondence: Hang.Geng@uestc.edu.cn Received: 10 July 2023 Accepted: 11 October 2023 Published: 26 March 2024 Abstract: Intelligent manufacturing is facing significant challenges in adapting to the ever-changing equipment, instrumentation, process and economics. Such a trend together with the pressure to reliably control and contain production costs means that frequent adjusting decisions are required to adapt to incessant volatility imposed on manufacturing systems. Under this circumstance, cost-effective and quality-guaranteed manufacturing strategies would be the most logical route to reducing production costs. In this paper, a novel dynamical cost prediction and control (CPC) model is proposed to support collective decision-making in intelligent manufacturing, where the model output is the real-time prediction of possible manufacturing costs, while the inputs are generic manufacturing key performance indicators covering inventory, product quality, production efficiency, resource utilisation and environmental impact. This proposed CPC model distinguishes itself from existing ones for its capability to translate manufacturing data (at both the physical level and operation management level) into financial metrics that contribute to forming a common language between engineering, financial and administrative departments of an enterprise. The case study about the assembly line of optoelectronic devices demonstrates that, although different enterprise departments have different priorities, our CPC model helps them to achieve certain consensus on intended production that finally creates satisfactory profitability for the company at controlled manufacturing costs. Intelligent manufacturing is a highly evolving and competitive industrial field, and a prerequisite to keep this evolvement and competitiveness is to reliably control the capital and operational expenditure while significantly reduce the final production cost [1−5]. To fulfill this goal, prediction and control of manufacturing costs becomes a major concern of both stakeholders and decision-makers, and once such prediction is obtained, business decisions and operation processes could be fully evaluated and further improved [6−10]. Generally, cost prediction and control (CPC) are vital in two aspects: 1) accurate CPC assists decision-makers in meeting business objectives such as customer satisfaction, waste reduction, and profit maximization; and 2) predicted costs can be used as manufacturing metrics to dissolve often conflicting objectives such as delivery time, production flow, energy efficiency and production quality so as to achieve balanced manufacture. This gives rise to extraordinary enthusiasm in formulating various CPC approaches whose focus is the appropriate translation of management data along with operation data into certain financial cost models to support timely yet economic decision-making of a manufacturing enterprise, see e.g. references [11−14]. Basically, even in the most mature and advanced manufacturing realms (e.g. optoelectronics), it is still a significant challenge to conduct real-time and non-stop CPC due to the colossal amount of data acquisition required by the cost analysis of every element with respect to the manufacturing process [15−18]. Moreover, most CPC approaches (e.g. the well-known activity-based CPC approach) carry out data actualization via a labor-intensive, case-based and one-off way, which are inappropriate for highly evolving production plans seeking for real-time adaptation [19−21]. In this paper, we intend to fill such a gap by building a dynamical CPC model capable of adapting itself to such possible variations of production plans. This is made possible via making full use of the capabilities (owned by the industrial internet of things) with respect to real-time tracking and tracing of manufacturing states, e.g. machine and product states. As decision-making under a complex manufacturing environment constitutes a major concern to the management department [22−24], we are also motivated to endow the established CPC model with the ability to support the decision-making process, where the key is to reach a balance between key performance parameters (KPIs), e.g. the inventory, productivity, product quality, energy efficiency and environmental impact, of the manufacturing processes [25, 26]. Once the balance is reached, we are able to boost productivity, product quality, energy efficiency, etc. and alleviate inventory and environmental impact through the reduction of manufacturing defects and wastes. 1.1. Literature Review Manufacturing-system-oriented CPC modeling is the integration of engineering principles, scientific theories and commercial practices for predicting the cost of manufacturing processes. Such kind of cost modeling is widely witnessed in a great variety of areas such as vehicle production, mechanical engineering, semiconductor manufacture and optoelectronic manufacture, see e.g. [27−30]. Basically, CPC techniques can be divided into qualitative and quantitative methods, where the former consists of intuitive and analogical CPC approaches, while the latter contains parametric and analytical approaches. In the intuitive CPC approach, past experience and expert knowledge are systemically leveraged to build a combination of rules and decisions to form a cost function with typical examples like the decision support systems (DSS), and case-based, rule-based, expert-based and fuzzy logic-based methodologies, see [31−33] and the references therein. The analogical CPC approach builds itself on similarity analysis that compares new products with old products of known costs in order to predict the costs of new products [34−36], while the parametric CPC approach relies on the mathematical/statistical relationship between product costs and associate parameters [37−39]. A noteworthy point is that in the analytical CPC approach, all costs of the material, labor, infrastructure, etc. are required to be added up to generate the final production costs. To be more specific, following the idea of intuitive CPC, a knowledge-based intelligent cost model has been built in [30] to estimate the manufacturing cost at the conceptual design stage, where both material cost and operation costs (of machines) have been considered. Using the analogical CPC approach, a novel cost model has been given in [40] to estimate the operational machining costs of the microelectronic discharge. However, in such a model, the variation of workpiece/electrode parameters would lead to the change of machining time, limiting its wide applications to broader industrial scenarios. Unlike the analogical CPC approach, the parametric CPC approach requires past/historical data, and is more suitable to be applied to the early design stage without any prior knowledge about manufacturing plans [41]. As one of the most famous analytical CPC approaches, the activity-based costing (ABC) methodology has been developed in [42] to address the increasing share problem caused by the indirect fixed costs (which have significant roles in cost models of manufactured products). Note that the ABC method is, in fact, a costing model that assigns resources/costs (demanded by each production activity) to all services and products at every stage of production, delivery and marketing. Consequently, the aim of ABC is to measure and price out all activity resources together with services for customers. Unfortunately, although the ABC method is of high CPC accuracy, it can only be implemented after the completion of production, which unavoidably fails to satisfy the demand of real-time manufacture. For the latest CPC development, interested readers are referred to [43−46]. 1.2. Statement of Contributions In this article, we consider the problem of real-time CPC in intelligent manufacture which is a challenging problem that has been rarely investigated in the literature. The aim is to translate manufacturing KPIs into a universal and singular metric. We focus on a general CPC problem that features a systematic yet easy-to-deploy cost framework and addresses shortcomings of traditional detail-oriented CPC methods. Our study leads to the following contributions. 1) A novel CPC is established that translates sustainability initiatives and production strategies into real financial costs, therefore enabling operation managers/engineers to communicate with high-level decision-makers about the production process. 2) An elaborate selection strategy is designed for a series of KPIs that represent the manufacturing performance with the help of industrial partners (from the optoelectronics field) and subsequently, such KPIs are adopted as the inputs to the established CPC model. 3) A model validation process is conducted (at plants of our partners) that verifies at the shop floor, and production costs can be instantaneously reduced by tactical optimization, giving rise to high sustainability; and at the management level, proper determination of manufacturing parameters dramatically reduces the product defects, material wastes, time consumption, gas emission and energy overhead. 1.3. Organization The remainder of this article is organized as follows. In Section 2, the CPC problem of intelligent manufacture is formulated. In Section 3, the KPI-based solution is elegantly explored with its contribution to the final production cost clearly elaborated. In Section 4, the case study on a laser device company is fully carried out whose final production cost is predicted using the proposed CPC model, where a detailed comparison is provided between the predicted cost and the real cost (data given by the financial department of the partner company) to demonstrate the applicability of the proposed CPC model. Section 5 supplies a few concluding remarks on the proposed CPC modeling method. In manufacturing operations, KPIs can be defined as a group of metrics that demonstrate how efficiently a company achieves its operational performance [47, 48]. KPIs are sometimes referred to as values which can be compared with internal or external targets to give certain indications of the operational performance. Hundreds of KPIs are frequently used in manufacturing industries ranging from the automobile industry to the construction industry. For instance, the emission, waste and resource utilisation (RU) in automotive fields; the quality, delivery, cost and flexibility in financial fields; the suppliers and employees in social fields; and the construction time and speed, time variation, quality, functionality, unit cost and user satisfaction in construction fields [49−51]. To sum up, each type of industry has its own indicators depending on the final business objectives. Quality and productivity, as the crucial factors of manufacturing sustainability, are always contradicted with each other [52, 53]. The current CPC methodologies aim to mitigate this conflict by providing a cost model allowing manufacturers to appropriately opt manufacturing parameters so as to maintain the productivity at certain acceptable level, meanwhile capable of implementing zero-defect manufacture. Here, we aim to build a novel dynamical cost model to support collective decision-making in intelligent manufacturing, where the model output is the real-time prediction of possible manufacturing costs, while the inputs are five generic manufacturing KPIs. Normally, the number of KPIs with respect to certain manufacturing environment is dependent on the business size, complexity as well as enterprise principles and priorities [54]. As suggested by our case study partner, we choose five KPIs (i.e. inventory, product quality, production efficiency, RU and environmental impact) as our focused parameters to demonstrate how effectively and efficiently the company operates. As a matter of fact, the function of the cost model is to translate the gain of each KPI into certain financial cost (i.e. the monetary value). In other words, given the desired quality and quantity of the order, the formulated model is able to output the final production cost. Figure 1 illustrates the prediction flowchart of how such a cost model can be used to build a bridge between the newly arrived customer order and the resultant production cost. To be specific, when a new customer order arrives at the manufacturing factory, the production planner employs an underlying cost model to identify the most suitable KPIs as well as the related manufacturing parameters (about this specific customer order) that are capable of providing both low defect rate and high productivity of the overall manufacturing process. Figure 1. Cost prediction flowchart. Notice that it is not a one-shot process to calculate the final product cost as the calculation consists of several steps associated with several computation formulas about elaborately selected KPIs. Inspired by [28], the CPC model proposed in this paper can be established as follows: where is the final production cost; is the operational cost which includes salaries of operators, machines utilization fees, maintenance and depreciation fees, set up fees and energy consumption fees; represents the raw material cost; defines the inventory cost; and denotes the environmental impact cost. The four cost terms are analysed/calculated according to the previously selected five KPIs and subsequently, summed up to generate the final production cost. In what follows, let us study how each KPI contributes to the above four cost terms as well as the final production cost. In this section, we will explore the relationships between the aforementioned four cost terms, the five KPIs (i.e. inventory, product quality, production efficiency, RU and environmental impact) and a series of manufacturing parameters in order to determine , , and . With these four cost terms in hand, we will be able to acquire the output of model (1) which is the most important factor to both customer and production manager. 3.1. Product Quality Contribution Product quality serves as a key factor in the manufacturing process, and exerts direct influence on the final production cost. Moreover, product quality could also affect customer-manufacturer relationships such as the amount of orders, the evaluation of products, and the loyalty to manufacturers. Basically, there are numerous factors affecting product quality. For example, in cases of equipment faults, machine breakdowns and unqualified products, production managers and engineers have to stop the production process to carry out certain maintenance, called the routine maintenance frequency (RMF), in order to guarantee product quality. Basically, a low (high) RMF often indicates few (many) equipment/machine breakdowns and therefore, good (bad) production quality. More specifically, in the production of laser devices, the quality of the laser device deteriorates as the RMF of the assembly machines is out of certain range. As such, a trade-off between the product quality and RMF must be reached according to each separate order and specification that are set by customers. Define the RMF (expressed in numbers) by and the non-defective percentage of the produced devices by . Then, the relation between and is characterized by the following quadratic relationship: where , and are case-based coefficients to be determined. In addition, we categorize defective products into repairable and non-repairable ones as follows: where denotes the percentage of reparable products; and and represent the repairable and non-repairable percentages of the defective products, respectively. 3.2. Productivity Contribution The proposed CPC model in (1) is a function of the RU with respect to the machine/equipment, and the RU (denoted by ) is a main measurable indicator of productivity. Generally, the RMF is the average opening/shutting frequency (required for machine/equipment maintenance per week), and the shift frequency (expressed in numbers) is the total opening/shutting frequency required for completing the assigned production workload per week. In this case, the RU can be formulated as follows: Without loss of generality, the cost of machine operation is assumed to be a quadratic function with respect to as in many other manufacturing cases, see [55, 56] and the references therein. Note that combines together the salaries of operators, machines utilization fees, maintenance and depreciation fees, set up fees and energy consumption fees. The detailed calculation formula of is given as follows: where are cased-based coefficients to be determined by making use of the collected manufacturing data. 3.3. Efficiency Contribution Production efficiency evaluates the performance level and the capability of the manufacturing system to not only create qualified products, but also achieve satisfactory operations along with financial performances. Generally, the lower efficiency a manufacturing system has, the higher cost a manufacturer pays. As such, production efficiency can serve as a primary index of manufacturing enterprises striving to satisfy highly-changing customer demands and occupy competitive market shares. The proposed CPC model characterizes the KPI efficiency in terms of the raw material cost and the operational cost . Denote as the normal material cost and as the normal inspection cost. Then, the calculation formulas of and can be established as follows: where is constituted by two terms, i.e. the normal material cost , and the additional material cost required to repair defected products which are reparable; and is constituted by four terms, i.e. the operational cost , the additional operational cost required to repair defected products which are reparable, the normal inspection cost required to inspect machine status, and the additional inspection cost required to inspect machine status when repairing the defective products. 3.4. Environmental Impact Contribution Environmental impact on manufacturing systems is becoming a major concern as most developing/developed countries are striving to reach carbon neutrality in the middle of this century. A typical approach to describing the carbon emission of manufacturing plants is the measurement of the associated energy consumption. However, due to various environmental rules/terms in different countries, environmental impact may have largely discrepant results on the final production costs. Such costs should cover the influence from qualified, reworked and unqualified production. Denote the normal penalty for energy consumption as and the normal penalty for used materials as . Then, the specific environmental impact onto the production cost is given as follows: where is the environmental cost, term evaluates the energy consumption and penalty with respect to both qualified and unqualified products, while term assesses the energy consumption and penalty with respect to previously defected but later reworked products. 3.5. Inventory Contribution The inventory cost represents all overheads linked to material storage (which gradually reduces as raw...

Machine learning has emerged as a key approach in wildfire risk prediction research. However, in practical applications, the scarcity of data for specific regions often hinders model performance, with models trained on region-specific data struggling to generalize due to differences in data distributions. While traditional methods based on expert knowledge tend to generalize better across regions, they are limited in leveraging multi-source data effectively, resulting in suboptimal predictive accuracy. This paper addresses this challenge by exploring how accumulated domain expertise in wildfire prediction can reduce model reliance on large volumes of high-quality data. An active learning algorithm is proposed based on XGBoost for wildfire risk assessment that autonomously identifies low-confidence predictions and seeks re-labeling through a human-in-the-loop or physics-based correction approach. This corrected data is reintegrated into the model, effectively preventing catastrophic forgetting. Experimental results demonstrate that the proposed human-in-the-loop approach significantly enhances labeling accuracy, predictive performance, and preserves the model’s ability to generalize. These findings highlight the value of incorporating human expertise into machine learning models, offering a practical solution to mitigate data quality challenges and improve model reliability in wildfire risk prediction.

As the core sensor of inertial measurement units, the reliability of Micro-Electro-Mechanical Systems (MEMS) gyroscopes is critical for long-term navigation and motion control applications. To bridge the mechanism-data gap in MEMS multi-mechanism degradation modeling, this paper proposes a physics-informed dual-indicator reliability assessment framework based on Wiener processes. Two degradation indicators under consideration are frequency-related degradation caused by stiffness degradation and Q-factor degradation caused by damping degradation, for which corresponding physics-embedded stochastic degradation models are formulated. The two indicators are normalized and fused through a generalized weighted limit state function, where failure is defined as gyroscope-level performance failure. Closed-form reliability expressions are derived for linear limit states, while Monte Carlo simulation is used for nonlinear cases. Reduced-order multiphysics simulation cases, including a double-ended fixed beam and a cantilevered MEMS mass block, are used to demonstrate the mechanism-to-indicator-to-reliability modeling procedure. The results show that the proposed dual-indicator framework provides more balanced reliability assessment than single-indicator analysis under the simulation setting. The proposed method offers an alternative mechanism-informed approach for reliability analysis and lifetime prediction of other MEMS devices.

To address the challenges of vegetation interference, background confusion, and road fragmentation caused by the narrow and elongated structures of forest roads in complex remote sensing imagery, this study proposed an improved U-Net-based model, namely HAA-UNet, for automatic forest road extraction. The proposed model integrates a VGG16 encoder, an Atrous Spatial Pyramid Pooling (ASPP) module, and a Hybrid Dilated Convolution (HDCC) module to enhance feature representation and spatial detail reconstruction of road targets under complex forest environments. Experimental results demonstrated that HAA-UNet achieved superior segmentation performance on the forest road dataset of Xichang City, Sichuan Province, with Precision, Recall, F1-score, and mIoU values of 87.24%, 87.83%, 87.53%, and 79.85%, respectively, outperforming all comparison models. These results indicate that the proposed method effectively improves road continuity and boundary delineation in complex forest scenes. Furthermore, the extracted road information was integrated into the Forest Fire Risk Index (FFRI) assessment framework, demonstrating that accurate road data can improve the spatial characterization of fire risk and provide reliable data support for forest fire risk assessment and forest resource management.